elastix: a toolbox for intensity-based medical image registration

Improving lesion volume measurements on digital mammograms

Lesion volume is an important predictor for prognosis in breast cancer. However, it is currently impossible to compute lesion volumes accurately from digital mammography data, which is the most popular and readily available imaging modality for breast cancer. We make a step towards a more accurate lesion volume measurement on digital mammograms by developing a model that allows to estimate lesion volumes on processed mammogram. Processed mammograms are the images routinely used by radiologists in clinical practice as well as in breast cancer screening and are available in medical centers. Processed mammograms are obtained from raw mammograms, which are the X-ray data coming directly from the scanner, by applying certain vendor-specific non-linear transformations. At the core of our volume estimation method is a physics-based algorithm for measuring lesion volumes on raw mammograms. We subsequently extend this algorithm to processed mammograms via a deep learning image-to-image translation model that produces synthetic raw mammograms from processed mammograms in a multi-vendor setting. We assess the reliability and validity of our method using a dataset of 1778 mammograms with an annotated mass. Firstly, we investigate the correlations between lesion volumes computed from mediolateral oblique and craniocaudal views, with a resulting Pearson correlation of 0.93 [95% confidence interval (CI) 0.92 - 0.93]. Secondly, we compare the resulting lesion volumes from true and synthetic raw data, with a resulting Pearson correlation of 0.998 [95%CI 0.998 - 0.998] . Finally, for a subset of 100 mammograms with a malignant mass and concurrent MRI examination available, we analyze the agreement between lesion volume on mammography and MRI, resulting in an intraclass correlation coefficient of 0.81 [95%CI 0.73 - 0.87] for consistency and 0.78 [95%CI 0.66 - 0.86] for absolute agreement. In conclusion, we developed an algorithm to measure mammographic lesion volume that reached excellent reliability and good validity, when using MRI as ground truth. The algorithm may play a role in lesion characterization and breast cancer prognostication on mammograms.

Multi-grained contrastive representation learning for label-efficient lesion segmentation and onset time classification of acute ischemic stroke

Ischemic lesion segmentation and the time since stroke (TSS) onset classification from paired multi-modal MRI imaging of unwitnessed acute ischemic stroke (AIS) patients is crucial, which supports tissue plasminogen activator (tPA) thrombolysis decision-making. Deep learning methods demonstrate superiority in TSS classification. However, they often overfit task-irrelevant features due to insufficient paired labeled data, resulting in poor generalization. We observed that unpaired data are readily available and inherently carry task-relevant cues, but are less often considered and explored. Based on this, in this paper, we propose to fully excavate the potential of unpaired unlabeled data and use them to facilitate the downstream AIS analysis task. We first analyze the utility of features at the varied grain and propose a multi-grained contrastive learning (MGCL) framework to learn task-related prior representations from both coarse-grained and fine-grained levels. The former can learn global prior representations to enhance the location ability for the ischemic lesions and perceive the healthy surroundings, while the latter can learn local prior representations to enhance the perception ability for semantic relation between the ischemic lesion and other health regions. To better transfer and utilize the learned task-related representation, we designed a novel multi-task framework to simultaneously achieve ischemic lesion segmentation and TSS classification with limited labeled data. In addition, a multi-modal region-related feature fusion module is proposed to enable the feature correlation and synergy between multi-modal deep image features for more accurate TSS decision-making. Extensive experiments on the large-scale multi-center MRI dataset demonstrate the superiority of the proposed framework. Therefore, it is promising that it helps better stroke evaluation and treatment decision-making.

Uncertain prediction of deformable image registration on lung CT using multi-category features and supervised learning

The assessment of deformable registration uncertainty is an important task for the safety and reliability of registration methods in clinical applications. However, it is typically done by a manual and time-consuming procedure. We propose a novel automatic method to predict registration uncertainty based on multi-category features and supervised learning. Three types of features, including deformation field statistical features, deformation field physiologically realistic features, and image similarity features, are introduced and calculated to train the random forest regressor for local registration uncertain prediction. Deformation field statistical features represent the numerical stability of registration optimization, which are correlated to the uncertainty of deformation fields; deformation field physiologically realistic features represent the biomechanical properties of organ motions, which mathematically reflect the physiological reality of deformation; image similarity features reflect the similarity between the warped image and fixed image. The multi-category features comprehensively reflect the registration uncertainty. The strategy of spatial adaptive random perturbations is also introduced to accurately simulate spatial distribution of registration uncertainty, which makes deformation field statistical features more discriminative to the uncertainty of deformation fields. Experiments were conducted on three publicly available thoracic CT image datasets. Seventeen randomly selected image pairs are used to train the random forest model, and 9 image pairs are used to evaluate the prediction model. The quantitative experiments on lung CT images show that the proposed method outperforms the baseline method for uncertain prediction of classical iterative optimization-based registration and deep learning-based registration with different registration qualities. The proposed method achieves good performance for registration uncertain prediction, which has great potential in improving the accuracy of registration uncertain prediction.

Tumor habitat and peritumoral region evolution-based imaging features to assess risk categorization of thymomas

AIM

To develop an aggregate model that integrated clinical data, habitat characteristics, and intratumoral and peritumoral features to assess the risk categorization of thymomas.

MATERIALS AND METHODS

We retrospectively analyzed 140 thymoma patients (70 low-risk and 70 high-risk), including pathological data. The patients were randomly divided into training cohort (n = 114) and test cohort (n = 26). The k-means clustering was utilized to partition the primary tumor into habitats based on intratumoral radiomic features, 6 distinct habitats were identified. By expanding the region of interest (ROI) mask, 2 peritumoral regions were obtained. Finally, 7 clinical characteristics, 3 habitat values, 20 radiomic features were utilized to develop an aggregated model, to predict the risk of thymoma. Shapley additive explanations (SHAP) interpretation was used for features importance ranking. The accuracy and area under curve (AUC) were used to analyze the performance of the models.

RESULTS

The aggregated model, which utilized the XGBoost classifier, demonstrated the best performance with an AUC of 0.811 and an accuracy of 0.769. In comparison, the radiomic model produced an AUC of 0.654 and an accuracy of 0.692. Additionally, the Intratumoral + peritumoral model exhibited an AUC of 0.728 and an accuracy of 0.769.

CONCLUSION

Our study establishes a novel tool to predict the risk of thymoma with a good performance. If prospectively validated, the model may refine thymoma patient selection for risk-adaptative therapy and improve prognosis.

HaN-Seg: The head and neck organ-at-risk CT and MR segmentation challenge

BACKGROUND AND PURPOSE

To promote the development of auto-segmentation methods for head and neck (HaN) radiation treatment (RT) planning that exploit the information of computed tomography (CT) and magnetic resonance (MR) imaging modalities, we organized HaN-Seg: The Head and Neck Organ-at-Risk CT and MR Segmentation Challenge.

MATERIALS AND METHODS

The challenge task was to automatically segment 30 organs-at-risk (OARs) of the HaN region in 14 withheld test cases given the availability of 42 publicly available training cases. Each case consisted of one contrast-enhanced CT and one T1-weighted MR image of the HaN region of the same patient, with up to 30 corresponding reference OAR delineation masks. The performance was evaluated in terms of the Dice similarity coefficient (DSC) and 95-percentile Hausdorff distance (HD95), and statistical ranking was applied for each metric by pairwise comparison of the submitted methods using the Wilcoxon signed-rank test.

RESULTS

While 23 teams registered for the challenge, only seven submitted their methods for the final phase. The top-performing team achieved a DSC of 76.9 % and a HD95 of 3.5 mm. All participating teams utilized architectures based on U-Net, with the winning team leveraging rigid MR to CT registration combined with network entry-level concatenation of both modalities.

CONCLUSION

This challenge simulated a real-world clinical scenario by providing non-registered MR and CT images with varying fields-of-view and voxel sizes. Remarkably, the top-performing teams achieved segmentation performance surpassing the inter-observer agreement on the same dataset. These results set a benchmark for future research on this publicly available dataset and on paired multi-modal image segmentation in general.

MyoFold: Joint myocardial tissue composition and wall motion quantification via a highly folded sequence

PURPOSE

This study aims to develop and evaluate a novel cardiovascular MR sequence, MyoFold, designed for the simultaneous quantifications of myocardial tissue composition and wall motion.

METHODS

MyoFold is designed as a 2D single breathing-holding sequence, integrating joint T1/T2 mapping and cine imaging. The sequence uses a 2-fold accelerated balanced SSFP (bSSFP) for data readout and incorporates electrocardiogram synchronization to align with the cardiac cycle. MyoFold initially acquires six single-shot inversion-recovery images, completed during the diastole of six successive heartbeats. T2 preparation (T2-prep) is applied to introduce T2 weightings for the last three images. Subsequently, over the following six heartbeats, segmented bSSFP is performed for the movie of the entire cardiac cycle, synchronized with an electrocardiogram. A neural network trained using numerical simulations of MyoFold is used for T1 and T2 calculations. MyoFold was validated through phantom and in vivo experiments, with comparisons made against MOLLI, SASHA, T2-prep bSSFP, and the conventional cine.

RESULTS

In phantom studies, MyoFold exhibited a 10% overestimation in T1 measurements, whereas T2 measurements demonstrated high accuracy. In vivo experiments revealed that MyoFold T1 had comparable accuracy to SASHA and precision similar to MOLLI. MyoFold demonstrated good agreement with T2-prep bSSFP in myocardial T2 measurements. No significant differences were observed in the quantification of left-ventricle wall thickness and function between MyoFold and the conventional cine.

CONCLUSION

MyoFold presents as a rapid, simple, and multitasking approach for quantitative cardiovascular MR examinations, offering simultaneous assessment of tissue composition and wall motion. The sequence's multitasking capabilities make it a promising tool for comprehensive cardiac evaluations in clinical settings.

Anisotropy of T2 and T1ρ relaxation time in articular cartilage at 3 T

PURPOSE

The anisotropy of R2 and R1ρ relaxation rates in articular cartilage contains information about the collagenous structure of the tissue. Here we determine and study the anisotropic and isotropic components of T2 and T1ρ relaxation parameters in articular cartilage with a clinical 3T MRI device. Furthermore, a visual representation of the topographical variation in anisotropy is given via anisotropy mapping.

METHODS

Eight bovine stifle joints were imaged at 22 orientations with respect to the main magnetic field using T2, continuous-wave (CW) T1ρ, and adiabatic T1ρ mapping sequences. Relaxation rates were separated into isotropic and anisotropic relaxation components using a previously established relaxation anisotropy model. Pixel-wise anisotropy values were determined from the relaxation-time maps using Michelson contrast.

RESULTS

The relaxation rates obtained from the samples displayed notable variation depending on the sample orientation, magnetization preparation, and cartilage layer. R2 demonstrated significant anisotropy, whereas CW-R1ρ (300 Hz) and CW-R1ρ (500 Hz) displayed a low degree of anisotropy. Adiabatic R1ρ was largely isotropic. In the deep cartilage regions, relaxation rates were generally faster and more anisotropic than in the cartilage closer to the tissue surface. The isotropic relaxation rate components were found to have similar values regardless of measurement sequence.

CONCLUSIONS

The fitted relaxation model for T2 and T1ρ demonstrated varying amounts anisotropy, depending on magnetization preparation, and studied the articular cartilage layer. Anisotropy mapping of full joints showed varying amounts of anisotropy depending on the quantitative MRI parameter and topographical location, and in the case of T2, showed systematic changes in anisotropy across cartilage depth.

Contrast-Enhancing Lesion Segmentation in Multiple Sclerosis: A Deep Learning Approach Validated in a Multicentric Cohort

The detection of contrast-enhancing lesions (CELs) is fundamental for the diagnosis and monitoring of patients with multiple sclerosis (MS). This task is time-consuming and suffers from high intra- and inter-rater variability in clinical practice. However, only a few studies proposed automatic approaches for CEL detection. This study aimed to develop a deep learning model that automatically detects and segments CELs in clinical Magnetic Resonance Imaging (MRI) scans. A 3D UNet-based network was trained with clinical MRI from the Swiss Multiple Sclerosis Cohort. The dataset comprised 372 scans from 280 MS patients: 162 showed at least one CEL, while 118 showed no CELs. The input dataset consisted of T1-weighted before and after gadolinium injection, and FLuid Attenuated Inversion Recovery images. The sampling strategy was based on a white matter lesion mask to confirm the existence of real contrast-enhancing lesions. To overcome the dataset imbalance, a weighted loss function was implemented. The Dice Score Coefficient and True Positive and False Positive Rates were 0.76, 0.93, and 0.02, respectively. Based on these results, the model developed in this study might well be considered for clinical decision support.

Longitudinal registration of T1-weighted breast MRI: A registration algorithm (FLIRE) and clinical application

PURPOSE

MRI is commonly used to aid breast cancer diagnosis and treatment evaluation. For patients with breast cancer, neoadjuvant chemotherapy aims to reduce the tumor size and extent of surgery necessary. The current clinical standard to measure breast tumor response on MRI uses the longest tumor diameter. Radiologists also account for other tissue properties including tumor contrast or pharmacokinetics in their assessment. Accurate longitudinal image registration of breast tissue is critical to properly compare response to treatment at different timepoints.

METHODS

In this study, a deformable Fast Longitudinal Image Registration (FLIRE) algorithm was optimized for breast tissue. FLIRE was then compared to the publicly available software packages with high accuracy (DRAMMS) and fast runtime (Elastix). Patients included in the study received longitudinal T1-weighted MRI without fat saturation at two to six timepoints as part of asymptomatic screening (n = 27) or throughout neoadjuvant chemotherapy treatment (n = 32). T1-weighted images were registered to the first timepoint with each algorithm.

RESULTS

Alignment and runtime performance were compared using two-way repeated measure ANOVAs (P < 0.05). Across all patients, Pearson's correlation coefficient across the entire image volume was slightly higher with statistical significance and had less variance for FLIRE (0.98 ± 0.01 stdev) compared to DRAMMS (0.97 ± 0.03 stdev) and Elastix (0.95 ± 0.03 stdev). Additionally, FLIRE runtime (10.0 mins) was 9.0 times faster than DRAMMS (89.6 mins) and 1.5 times faster than Elastix (14.5 mins) on a Linux workstation.

CONCLUSION

FLIRE demonstrates promise for time-sensitive clinical applications due to its accuracy, robustness across patients and timepoints, and speed.

Novel spin-lock time sampling strategies for improved reproducibility in quantitative T1ρ mapping

This study aimed to optimize the sampling of spin-lock times (TSLs) in quantitative T1ρ mapping for improved reproducibility. Two new TSL sampling schemes were proposed: (i) reproducibility-guided random sampling (RRS) and (ii) reproducibility-guided optimal sampling (ROS). They were compared to the existing linear sampling (LS) and precision-guided sampling (PS) schemes for T1ρ reproducibility through numerical simulations, phantom experiments, and volunteer studies. Each study evaluated the four sampling schemes with three commonly used T1ρ preparations based on composite and balanced spin-locking. Additionally, the phantom and volunteer studies investigated the impact of B0 and B1 field inhomogeneities on T1ρ reproducibility, respectively. The reproducibility was assessed using the coefficient of variation (CoV) by repeating the T1ρ measurements eight times for phantom experiments and five times for volunteer studies. Numerical simulations resulted in lower mean CoVs for the proposed RRS (1.74%) and ROS (0.68%) compared to LS (2.93%) and PS (3.68%). In the phantom study, the mean CoVs were also lower for RRS (2.7%) and ROS (2.6%) compared to LS (4.1%) and PS (3.1%). Furthermore, the mean CoVs of the proposed RRS and ROS were statistically lower (P < 0.001) compared to existing LS and PS schemes at a B1 offset of 20%. In the volunteer study, consistently lower mean CoVs were observed in bilateral thigh muscles for RRS (9.3%) and ROS (9.2%) compared to LS (10.9%) and PS (10.2%), and the difference was more prominent at B0 offsets higher than 50 Hz. The proposed sampling schemes improve the reproducibility of quantitative T1ρ mapping by optimizing the selection of TSLs. This improvement is especially beneficial for longitudinal studies that track and monitor disease progression and treatment response.

Multicompartment imaging of the brain using a comprehensive MR imaging protocol

In this study, we describe a comprehensive 3D magnetic resonance imaging (MRI) protocol designed to assess major tissue and fluid components in the brain. The protocol comprises four different sequences: 1) magnetization transfer prepared Cones (MT-Cones) for two-pool MT modeling to quantify macromolecular content; 2) short-TR adiabatic inversion-recovery prepared Cones (STAIR-Cones) for myelin water imaging; 3) proton-density weighted Cones (PDw-Cones) for total water imaging; and 4) highly T2 weighted Cones (T2w-Cones) for free water imaging. By integrating these techniques, we successfully mapped key brain components-namely macromolecules, myelin water, intra/extracellular water, and free water-in ten healthy volunteers and five patients with multiple sclerosis (MS) using a 3T clinical scanner. Brain macromolecular proton fraction (MMPF), myelin water proton fraction (MWPF), intra/extracellular water proton fraction (IEWPF), and free water proton fraction (FWPF) values were generated in white matter (WM), grey matter (GM), and MS lesions. Excellent repeatability of the protocol was demonstrated with high intra-class correlation coefficient (ICC) values. In MS patients, the MMPF and MWPF values of the lesions and normal-appearing WM (NAWM) were significantly lower than those in normal WM (NWM) in healthy volunteers. Moreover, we observed significantly higher FWPF values in MS lesions compared to those in NWM and NAWM regions. This study demonstrates the capability of our technique to volumetrically map major brain components. The technique may have particular value in providing a comprehensive assessment of neuroinflammatory and neurodegenerative diseases of the brain.

Subtraction of single-photon emission computed tomography (SPECT) in radioembolization: a comparison of four methods

BACKGROUND

Subtraction of single-photon emission computed tomography (SPECT) images has a number of clinical applications in e.g. foci localization in ictal/inter-ictal SPECT and defect detection in rest/stress cardiac SPECT. In this work, we investigated the technical performance of SPECT subtraction for the purpose of quantifying the effect of a vasoconstricting drug (angiotensin-II, or AT2) on the Tc-99m-MAA liver distribution in hepatic radioembolization using an innovative interventional hybrid C-arm scanner. Given that subtraction of SPECT images is challenging due to high noise levels and poor resolution, we compared four methods to obtain a difference image in terms of image quality and quantitative accuracy. These methods included (i) image subtraction: subtraction of independently reconstructed SPECT images, (ii) projection subtraction: reconstruction of a SPECT image from subtracted projections, (iii) projection addition: reconstruction by addition of projections as a background term during the iterative reconstruction, and (iv) image addition: simultaneous reconstruction of the difference image and the subtracted image.

RESULTS

Digital simulations (XCAT) and phantom studies (NEMA-IQ and anthropomorphic torso) showed that all four methods were able to generate difference images but their performance on specific metrics varied substantially. Image subtraction had the best quantitative performance (activity recovery coefficient) but had the worst visual quality (contrast-to-noise ratio) due to high noise levels. Projection subtraction showed a slightly better visual quality than image subtraction, but also a slightly worse quantitative accuracy. Projection addition had a substantial bias in its quantitative accuracy which increased with less counts in the projections. Image addition resulted in the best visual image quality but had a quantitative bias when the two images to subtract contained opposing features.

CONCLUSION

All four investigated methods of SPECT subtraction demonstrated the capacity to generate a feasible difference image from two SPECT images. Image subtraction is recommended when the user is only interested in quantitative values, whereas image addition is recommended when the user requires the best visual image quality. Since quantitative accuracy is most important for the dosimetric investigation of AT2 in radioembolization, we recommend using the image subtraction method for this purpose.

Human cerebrospinal fluid monoclonal CASPR2 autoantibodies induce changes in electrophysiology, functional MRI, and behavior in rodent models

Anti-contactin associated protein receptor 2 (CASPR2) encephalitis is a severe autoimmune encephalitis with a variable clinical phenotype including behavioral abnormalities, cognitive decline, epileptic seizures, peripheral nerve hyperexcitability and neuropathic pain. The detailed mechanisms of how CASPR2 autoantibodies lead to synaptic dysfunction and clinical symptoms are largely unknown. Aiming for analyses from the molecular to the clinical level, we isolated antibody-secreting cells from the cerebrospinal fluid of two patients with CASPR2 encephalitis. From these we cloned four anti-CASPR2 human monoclonal autoantibodies (mAbs) with strong binding to brain and peripheral nerves. All were highly hypermutated and mainly of the IgG4 subclass. Mutagenesis studies determined selective binding to the discoidin domain of CASPR2. Surface plasmon resonance revealed affinities with dissociation constants KD in the pico- to nanomolar range. CASPR2 mAbs interrupted the interaction of CASPR2 with its binding partner contactin 2 in vitro and were internalized after binding to CASPR2-expressing cells. Electrophysiological recordings of rat hippocampal slices after stereotactic injection of CASPR2 mAbs showed characteristic afterpotentials following electrical stimulation. In vivo experiments with intracerebroventricular administration of human CASPR2 mAbs into mice and rats showed EEG-recorded brain hyperexcitability but no spontaneous recurrent seizures. Behavioral assessment of infused mice showed a subtle clinical phenotype, mainly affecting sociability. Mouse brain MRI exhibited markedly reduced resting-state functional connectivity without short-term structural changes. Together, the experimental data support the direct pathogenicity of CASPR2 autoantibodies. The minimally invasive EEG and MRI techniques applied here may serve as novel objective, quantifiable tools for improved animal models, in particular for subtle neuropsychiatric phenotypes or repeated measurements.

On-board synthetic 4D MRI generation from 4D CBCT for radiotherapy of abdominal tumors: A feasibility study

BACKGROUND

Magnetic resonance-guided radiotherapy with an MR-guided LINAC represents potential clinical benefits in abdominal treatments due to the superior soft-tissue contrast compared to kV-based images in conventional treatment units. However, due to the high cost associated with this technology, only a few centers have access to it. As an alternative, synthetic 4D MRI generation based on artificial intelligence methods could be implemented. Nevertheless, appropriate MRI texture generation from CT images might be challenging and prone to hallucinations, compromising motion accuracy.

PURPOSE

To evaluate the feasibility of on-board synthetic motion-resolved 4D MRI generation from prior 4D MRI, on-board 4D cone beam CT (CBCT) images, motion modeling information, and deep learning models using the digital anthropomorphic phantom XCAT.

METHODS

The synthetic 4D MRI corresponds to phases from on-board 4D CBCT. Each synthetic MRI volume in the 4D MRI was generated by warping a reference 3D MRI (MRIref, end of expiration phase from a prior 4D MRI) with a deformation field map (DFM) determined by (I) the eigenvectors from the principal component analysis (PCA) motion-modeling of the prior 4D MRI, and (II) the corresponding eigenvalues predicted by a convolutional neural network (CNN) model using the on-board 4D CBCT images as input. The CNN was trained with 1000 deformations of one reference CT (CTref, same conditions as MRIref) generated by applying 1000 DFMs computed by randomly sampling the original eigenvalues from the prior 4D MRI PCA model. The evaluation metrics for the CNN model were root-mean-square error (RMSE) and mean absolute error (MAE). Finally, different on-board 4D-MRI generation scenarios were assessed by changing the respiratory period, the amplitude of the diaphragm, and the chest wall motion of the 4D CBCT using normalized root-mean-square error (nRMSE) and structural similarity index measure (SSIM) for image-based evaluation, and volume dice coefficient (VDC), volume percent difference (VPD), and center-of-mass shift (COMS) for contour-based evaluation of liver and target volumes.

RESULTS

The RMSE and MAE values of the CNN model reported 0.012 ± 0.001 and 0.010 ± 0.001, respectively for the first eigenvalue predictions. SSIM and nRMSE were 0.96 ± 0.06 and 0.22 ± 0.08, respectively. VDC, VPD, and COMS were 0.92 ± 0.06, 3.08 ± 3.73 %, and 2.3 ± 2.1 mm, respectively, for the target volume. The more challenging synthetic 4D-MRI generation scenario was for one 4D-CBCT with increased chest wall motion amplitude, reporting SSIM and nRMSE of 0.82 and 0.51, respectively.

CONCLUSIONS

On-board synthetic 4D-MRI generation based on predicting actual treatment deformation from on-board 4D-CBCT represents a method that can potentially improve the treatment-setup localization in abdominal radiotherapy treatments with a conventional kV-based LINAC.

Association of medullary reticular formation ventral part with spasticity in mice suffering from photothrombotic stroke

Strokes cause spasticity via stretch reflex hyperexcitability in the spinal cord, and spastic paralysis due to involuntary muscle contraction in the hands and fingers can severely restrict skilled hand movements. However, the underlying neurological mechanisms remain unknown. Using a mouse model of spasticity after stroke, we demonstrate changes in neuronal activity with and without electrostimulation of the afferent nerve to induce the stretch reflex, measured using quantitative activation-induced manganese-enhanced magnetic resonance imaging. Neuronal activity increased within the ventral medullary reticular formation (MdV) in the contralesional brainstem during the acute post-stroke phase, and this increase was characterised by activation of circuits involved in spasticity. Interestingly, ascending electrostimulation inhibited the MdV activity on the stimulation side in normal conditions. Moreover, immunohistochemical staining showed that, in the acute phase, the density of GluA1, one of the α-amino-3 hydroxy‑5 methyl -4 isoxazolepropionic acid receptor (AMPAR) subunits, at the synapses of MdV neurons was significantly increased. In addition, the GluA1/GluA2 ratio in these receptors was altered at 2 weeks post-stroke, confirming homeostatic plasticity as the underlying mechanisms of spasticity. These results provide new insights into the relationship between impaired skilled movements and spasticity at the acute post-stroke phase.

MACG-Net: Multi-axis cross gating network for deformable medical image registration

Deformable Image registration is a fundamental yet vital task for preoperative planning, intraoperative information fusion, disease diagnosis and follow-ups. It solves the non-rigid deformation field to align an image pair. Latest approaches such as VoxelMorph and TransMorph compute features from a simple concatenation of moving and fixed images. However, this often leads to weak alignment. Moreover, the convolutional neural network (CNN) or the hybrid CNN-Transformer based backbones are constrained to have limited sizes of receptive field and cannot capture long range relations while full Transformer based approaches are computational expensive. In this paper, we propose a novel multi-axis cross grating network (MACG-Net) for deformable medical image registration, which combats these limitations. MACG-Net uses a dual stream multi-axis feature fusion module to capture both long-range and local context relationships from the moving and fixed images. Cross gate blocks are integrated with the dual stream backbone to consider both independent feature extractions in the moving-fixed image pair and the relationship between features from the image pair. We benchmark our method on several different datasets including 3D atlas-based brain MRI, inter-patient brain MRI and 2D cardiac MRI. The results demonstrate that the proposed method has achieved state-of-the-art performance. The source code has been released at https://github.com/Valeyards/MACG.

Fast SPECT/CT planar bone imaging enabled by deep learning enhancement

BACKGROUND

The application of deep learning methods in rapid bone scintigraphy is increasingly promising for minimizing the duration of SPECT examinations. Recent works showed several deep learning models based on simulated data for the synthesis of high-count bone scintigraphy images from low-count counterparts. Few studies have been conducted and validated on real clinical pairs due to the misalignment inherent in multiple scan procedures.

PURPOSE

To generate high quality whole-body bone images from 2× and 3× fast scans using deep learning based enhancement method.

MATERIALS AND METHODS

Seventy-six cases who underwent whole-body bone scans were enrolled in this prospective study. All patients went through a standard scan at a speed of 20 cm/min, which followed by fast scans consisting of 2× and 3× accelerations at speeds of 40 and 60 cm/min. A content-attention image restoration approach based on Residual-in-Residual Dense Block (RRDB) is introduced to effectively recover high-quality images from fast scans with fine-details and less noise. Our approach is robust with misalignment introduced from patient's metabolism, and shows valid count-level consistency. Learned Perceptual Image Patch Similarity (LPIPS) and Fréchet Inception Distance (FID) are employed in evaluating the similarity to the standard bone images. To further prove our method practical in clinical settings, image quality of the anonymous images was evaluated by two experienced nuclear physicians on a 5-point Likert scale (5 =  excellent) .

RESULTS

The proposed method reaches the state-of-the-art performance on FID and LPIPS with 0.583 and 0.176 for 2× fast scans and 0.583 and 0.185 for 3× fast scans. Clinic evaluation further demonstrated the restored images had a significant improvement compared to fast scan in image quality, technetium 99m-methyl diphosphonate (Tc-99 m MDP) distribution, artifacts, and diagnostic confidence.

CONCLUSIONS

Our method was validated for accelerating whole-body bone scans by introducing real clinical data. Confirmed by nuclear medicine physicians, the proposed method can effectively enhance image diagnostic value, demonstrating potential for efficient high-quality fast bone imaging in practical settings.

Engineering Toxoplasma gondii secretion systems for intracellular delivery of multiple large therapeutic proteins to neurons

Delivering macromolecules across biological barriers such as the blood-brain barrier limits their application in vivo. Previous work has demonstrated that Toxoplasma gondii, a parasite that naturally travels from the human gut to the central nervous system (CNS), can deliver proteins to host cells. Here we engineered T. gondii's endogenous secretion systems, the rhoptries and dense granules, to deliver multiple large (>100 kDa) therapeutic proteins into neurons via translational fusions to toxofilin and GRA16. We demonstrate delivery in cultured cells, brain organoids and in vivo, and probe protein activity using imaging, pull-down assays, scRNA-seq and fluorescent reporters. We demonstrate robust delivery after intraperitoneal administration in mice and characterize 3D distribution throughout the brain. As proof of concept, we demonstrate GRA16-mediated brain delivery of the MeCP2 protein, a putative therapeutic target for Rett syndrome. By characterizing the potential and current limitations of the system, we aim to guide future improvements that will be required for broader application.

Multi-modal 3-Dimensional Visualization of Pediatric Neuroblastoma: Aiding Surgical Planning Beyond Anatomical Information

BACKGROUND

Patient-specific 3D models of neuroblastoma and relevant anatomy are useful tools for surgical planning. However, these models do not represent the heterogenous biology of neuroblastoma. This heterogeneity is visualized with the ADC and 123I-MIGB-SPECT-CT imaging. Combining these multi-modal data into preoperative 3D heatmaps, may allow differentiation of the areas of vital and non-vital tumor tissue. We developed a workflow to create multi-modal preoperative 3D models for neuroblastoma surgery.

METHODS

We included 7 patients who underwent neuroblastoma surgery between 2022 and 2023. We developed 3D models based on the contrast enhanced T1-weighted MRI scans. Subsequently, we aligned the corresponding ADC and 123I-MIBG-SPECT-CT images using rigid transformation. We estimated registration precision using the Dice score and the target registration error (TRE). 3D heatmaps were computed based on ADC and 123I-MIBG uptake.

RESULTS

The registration algorithm had a median Dice score of 0.81 (0.75-0.90) for ADC and 0.77 (0.65-0.91) for 123I-MIBG-SPECT. For the ADC registration, the median TRE of renal vessels was 4.90 mm (0.86-10.18) and of the aorta 4.67 mm (1.59-12.20). For the 123I -MIBG-SPECT imaging the TRE of the renal vessels was 5.52 mm (1.71-10.97) and 5.28 mm (3.33-16.77) for the aorta.

CONCLUSIONS

We successfully developed a registration workflow to create multi-modal 3D models which allows the surgeon to visualize the tumor and its biological behavior in relation to the surrounding tissue. Future research will include linking of pathological results to imaging data, to validate these multi-modal 3D models.

LEVEL OF EVIDENCE

Level IV.

TYPE OF STUDY

Clinical Research.

Deep learning-based white matter lesion volume on CT is associated with outcome after acute ischemic stroke

BACKGROUND

Intravenous thrombolysis (IVT) before endovascular treatment (EVT) for acute ischemic stroke might induce intracerebral hemorrhages which could negatively affect patient outcomes. Measuring white matter lesions size using deep learning (DL-WML) might help safely guide IVT administration. We aimed to develop, validate, and evaluate a DL-WML volume on CT compared to the Fazekas scale (WML-Faz) as a risk factor and IVT effect modifier in patients receiving EVT directly after IVT.

METHODS

We developed a deep-learning model for WML segmentation on CT and validated with internal and external test sets. In a post hoc analysis of the MR CLEAN No-IV trial, we associated DL-WML volume and WML-Faz with symptomatic-intracerebral hemorrhage (sICH) and 90-day functional outcome according to the modified Rankin Scale (mRS). We used multiplicative interaction terms between WML measures and IVT administration to evaluate IVT treatment effect modification. Regression models were used to report unadjusted and adjusted common odds ratios (cOR/acOR).

RESULTS

In total, 516 patients from the MR CLEAN No-IV trial (male/female, 291/225; age median, 71 [IQR, 62-79]) were analyzed. Both DL-WML volume and WML-Faz are associated with sICH (DL-WML volume acOR, 1.78 [95%CI, 1.17; 2.70]; WML-Faz acOR, 1.53 95%CI [1.02; 2.31]) and mRS (DL-WML volume acOR, 0.70 [95%CI, 0.55; 0.87], WML-Faz acOR, 0.73 [95%CI 0.60; 0.88]). Only in the unadjusted IVT effect modification analysis WML-Faz was associated with more sICH if IVT was given (p = 0.046). Neither WML measure was associated with worse mRS if IVT was given.

CONCLUSION

DL-WML volume and WML-Faz had a similar relationship with functional outcome and sICH. Although more sICH might occur in patients with more severe WML-Faz receiving IVT, no worse functional outcome was observed.

CLINICAL RELEVANCE STATEMENT

White matter lesion severity on baseline CT in acute ischemic stroke patients has a similar predictive value if measured with deep learning or the Fazekas scale. Safe administration of intravenous thrombolysis using white matter lesion severity should be further studied.

KEY POINTS

White matter damage is a predisposing risk factor for intracranial hemorrhage in patients with acute ischemic stroke but remains difficult to measure on CT. White matter lesion volume on CT measured with deep learning had a similar association with symptomatic intracerebral hemorrhages and worse functional outcome as the Fazekas scale. A patient-level meta-analysis is required to study the benefit of white matter lesion severity-based selection for intravenous thrombolysis before endovascular treatment.

CMAN: Cascaded Multi-scale Spatial Channel Attention-guided Network for large 3D deformable registration of liver CT images

Deformable image registration is an essential component of medical image analysis and plays an irreplaceable role in clinical practice. In recent years, deep learning-based registration methods have demonstrated significant improvements in convenience, robustness and execution time compared to traditional algorithms. However, registering images with large displacements, such as those of the liver organ, remains underexplored and challenging. In this study, we present a novel convolutional neural network (CNN)-based unsupervised learning registration method, Cascaded Multi-scale Spatial-Channel Attention-guided Network (CMAN), which addresses the challenge of large deformation fields using a double coarse-to-fine registration approach. The main contributions of CMAN include: (i) local coarse-to-fine registration in the base network, which generates the displacement field for each resolution and progressively propagates these local deformations as auxiliary information for the final deformation field; (ii) global coarse-to-fine registration, which stacks multiple base networks for sequential warping, thereby incorporating richer multi-layer contextual details into the final deformation field; (iii) integration of the spatial-channel attention module in the decoder stage, which better highlights important features and improves the quality of feature maps. The proposed network was trained using two public datasets and evaluated on another public dataset as well as a private dataset across several experimental scenarios. We compared CMAN with four state-of-the-art CNN-based registration methods and two well-known traditional algorithms. The results show that the proposed double coarse-to-fine registration strategy outperforms other methods in most registration evaluation metrics. In conclusion, CMAN can effectively handle the large-deformation registration problem and show potential for application in clinical practice. The source code is made publicly available at https://github.com/LocPham263/CMAN.git.

Attention-Aware Non-Rigid Image Registration for Accelerated MR Imaging

Accurate motion estimation at high acceleration factors enables rapid motion-compensated reconstruction in Magnetic Resonance Imaging (MRI) without compromising the diagnostic image quality. In this work, we introduce an attention-aware deep learning-based framework that can perform non-rigid pairwise registration for fully sampled and accelerated MRI. We extract local visual representations to build similarity maps between the registered image pairs at multiple resolution levels and additionally leverage long-range contextual information using a transformer-based module to alleviate ambiguities in the presence of artifacts caused by undersampling. We combine local and global dependencies to perform simultaneous coarse and fine motion estimation. The proposed method was evaluated on in-house acquired fully sampled and accelerated data of 101 patients and 62 healthy subjects undergoing cardiac and thoracic MRI. The impact of motion estimation accuracy on the downstream task of motion-compensated reconstruction was analyzed. We demonstrate that our model derives reliable and consistent motion fields across different sampling trajectories (Cartesian and radial) and acceleration factors of up to 16x for cardiac motion and 30x for respiratory motion and achieves superior image quality in motion-compensated reconstruction qualitatively and quantitatively compared to conventional and recent deep learning-based approaches. The code is publicly available at https://github.com/lab-midas/GMARAFT.

Automatic Segmentation and Quantification of Nigrosome-1 Neuromelanin and Iron in MRI: A Candidate Biomarker for Parkinson's Disease

BACKGROUND

There is a lack of automated tools for the segmentation and quantification of neuromelanin (NM) and iron in the nigrosome-1 (N1). Existing tools evaluate the N1 sign, i.e., the presence or absence of the "swallow-tail" in iron-sensitive MRI, or globally analyze the MRI signal in an area containing the N1, without providing a volumetric delineation.

PURPOSE

Present an automated method to segment the N1 and quantify differences in N1's NM and iron content between Parkinson's disease (PD) patients and healthy controls (HCs). Study whether N1 degeneration is clinically related to PD and could be used as a biomarker of the disease.

STUDY TYPE

Prospective.

SUBJECTS

Seventy-one PD (65.3 ± 10.3 years old, 34 female/37 male); 30 HC (62.7 ± 7.8 years old, 17 female/13 male).

FIELD STRENGTH/SEQUENCE

3 T Anatomical T1-weighted MPRAGE, NM-MRI T1-weighted gradient with magnetization transfer, susceptibility-weighted imaging (SWI).

ASSESSMENT

N1 was automatically segmented in SWI images using a multi-image atlas, populated with healthy N1 structures manually annotated by a neurologist. Relative NM and iron content were quantified and their diagnostic performance assessed and compared with the substantia nigra pars compacta (SNc). The association between image parameters and clinically relevant variables was studied.

STATISTICAL TESTS

Nonparametric tests were used (Mann-Whitney's U, chi-square, and Friedman tests) at P = 0.05.

RESULTS

N1's relative NM content decreased and relative iron content increased in PD patients compared with HCs (NM-CRHC = 22.55 ± 1.49; NM-CRPD = 19.79 ± 1.92; NM-nVolHC = 2.69 × 10-5 ± 1.02 × 10-5; NM-nVolPD = 1.18 × 10-5 ± 0.96 × 10-5; Iron-CRHC = 10.51 ± 2.64; Iron-CRPD = 19.35 ± 7.88; Iron-nVolHC = 0.72 × 10-5 ± 0.81 × 10-5; Iron-nVolPD = 2.82 × 10-5 ± 2.04 × 10-5). Binary logistic regression analyses combining N1 and SNc image parameters yielded a top AUC = 0.955. Significant correlation was found between most N1 parameters and both disease duration (ρNM-CR = -0.31; ρiron-CR = 0.43; ρiron-nVol = 0.46) and the motor status (ρNM-nVol = -0.27; ρiron-CR = 0.33; ρiron-nVol = 0.28), suggesting NM reduction along with iron accumulation in N1 as the disease progresses.

DATA CONCLUSION

This method provides a fully automatic N1 segmentation, and the analyses performed reveal that N1 relative NM and iron quantification improves diagnostic performance and suggest a relative NM reduction along with a relative iron accumulation in N1 as the disease progresses.

EVIDENCE LEVEL

1 TECHNICAL EFFICACY: Stage 1.

Deep learning-based automatic segmentation of bone graft material after maxillary sinus augmentation

OBJECTIVES

To investigate the accuracy and reliability of deep learning in automatic graft material segmentation after maxillary sinus augmentation (SA) from cone-beam computed tomography (CBCT) images.

MATERIALS AND METHODS

One hundred paired CBCT scans (a preoperative scan and a postoperative scan) were collected and randomly allocated to training (n = 82) and testing (n = 18) subsets. The ground truths of graft materials were labeled by three observers together (two experienced surgeons and a computer engineer). A deep learning model including a 3D V-Net and a 3D Attention V-Net was developed. The overall performance of the model was assessed through the testing data set. The comparative accuracy and inference time consumption of the model-driven and manual segmentation (by two surgeons with 3 years of experience in dental implant surgery) were conducted on 10 CBCT scans from the test samples.

RESULTS

The deep learning model had a Dice coefficient (Dice) of 90.36 ± 2.53%, a 95% Hausdorff distance (HD) of 1.59 ± 0.82 mm, and an average surface distance (ASD) of 0.38 ± 0.11 mm. The proposed model only needed 7.2 s, while the surgeon took 19.15 min on average to complete a segmentation task. The overall performances of the model were significantly superior to those of surgeons.

CONCLUSIONS

The proposed deep learning model yielded a more accurate and efficient performance of automatic segmentation of graft material after SA than that of the two surgeons. The proposed model could facilitate a powerful system for volumetric change evaluation, dental implant planning, and digital dentistry.

Influence of learned landmark correspondences on lung CT registration

BACKGROUND

Disease or injury may cause a change in the biomechanical properties of the lungs, which can alter lung function. Image registration can be used to measure lung ventilation and quantify volume change, which can be a useful diagnostic aid. However, lung registration is a challenging problem because of the variation in deformation along the lungs, sliding motion of the lungs along the ribs, and change in density.

PURPOSE

Landmark correspondences have been used to make deformable image registration robust to large displacements.

METHODS

To tackle the challenging task of intra-patient lung computed tomography (CT) registration, we extend the landmark correspondence prediction model deep convolutional neural network-Match by introducing a soft mask loss term to encourage landmark correspondences in specific regions and avoid the use of a mask during inference. To produce realistic deformations to train the landmark correspondence model, we use data-driven synthetic transformations. We study the influence of these learned landmark correspondences on lung CT registration by integrating them into intensity-based registration as a distance-based penalty.

RESULTS

Our results on the public thoracic CT dataset COPDgene show that using learned landmark correspondences as a soft constraint can reduce median registration error from approximately 5.46 to 4.08 mm compared to standard intensity-based registration, in the absence of lung masks.

CONCLUSIONS

We show that using landmark correspondences results in minor improvements in local alignment, while significantly improving global alignment.

Phase-Resolved Functional Lung (PREFUL) MRI to Quantify Ventilation: Feasibility and Physiological Relevance in Severe Asthma

RATIONALE AND OBJECTIVES

Emergent evidence in several respiratory diseases supports translational potential for Phase-Resolved Functional Lung (PREFUL) MRI to spatially quantify ventilation but its feasibility and physiological relevance have not been demonstrated in patients with asthma. This study compares PREFUL-derived ventilation defect percent (VDP) in severe asthma patients to healthy controls and measures its responsiveness to bronchodilator therapy and relation to established measures of airways disease.

MATERIALS AND METHODS

Forty-one adults with severe asthma and seven healthy controls performed same-day free-breathing 1H MRI, 129Xe MRI, spirometry, and oscillometry. A subset of participants (n = 23) performed chest CT and another subset of participants with asthma (n = 19) repeated 1H MRI following the administration of a bronchodilator. VDP was calculated for both PREFUL and 129Xe MRI. Additionally, the percent of functional small airways disease was determined from CT parametric response maps (PRMfSAD).

RESULTS

PREFUL VDP measured pre-bronchodilator (19.1% [7.4-43.3], p = 0.0002) and post-bronchodilator (16.9% [6.1-38.4], p = 0.0007) were significantly greater than that of healthy controls (7.5% [3.7-15.5]) and was significantly decreased post-bronchodilator (from 21.9% [10.1-36.9] to 16.9% [6.1-38.4], p = 0.0053). PREFUL VDP was correlated with spirometry (FEV1%pred: r = -0.46, p = 0.0023; FVC%pred: r = -0.35, p = 0.024, FEV1/FVC: r = -0.46, p = 0.0028), 129Xe MRI VDP (r = 0.39, p = 0.013), and metrics of small airway disease (CT PRMfSAD: r = 0.55, p = 0.021; Xrs5 Hz: r = -0.44, p = 0.0046, and AX: r = 0.32, p = 0.044).

CONCLUSION

PREFUL-derived VDP is responsive to bronchodilator therapy in asthma and is associated with measures of airflow obstruction and small airway dysfunction. These findings validate PREFUL VDP as a physiologically relevant and accessible ventilation imaging outcome measure in asthma.

Adaptive assessment based on fractional CBCT images for cervical cancer

PURPOSE

Anatomical and other changes during radiotherapy will cause inaccuracy of dose distributions, therefore the expectation for online adaptive radiation therapy (ART) is high in effectively reducing uncertainties due to intra-variation. However, ART requires extensive time and effort. This study investigated an adaptive assessment workflow based on fractional cone-beam computed tomography (CBCT) images.

METHODS

Image registration, synthetic CT (sCT) generation, auto-segmentation, and dose calculation were implemented and integrated into ArcherQA Adaptive Check. The rigid registration was based on ITK open source. The deformable image registration (DIR) method was based on a 3D multistage registration network, and the sCT generation method was performed based on a 2D cycle-consistent adversarial network (CycleGAN). The auto-segmentation of organs at risk (OARs) on sCT images was finished by a deep learning-based auto-segmentation software, DeepViewer. The contours of targets were obtained by the structure-guided registration. Finally, the dose calculation was based on a GPU-based Monte Carlo (MC) dose code, ArcherQA.

RESULTS

The dice similarity coefficient (DSCs) were over 0.86 for target volumes and over 0.79 for OARs. The gamma pass rate of ArcherQA versus Eclipse treatment planning system was more than 99% at the 2%/2 mm criterion with a low-dose threshold of 10%. The time for the whole process was less than 3 min. The dosimetric results of ArcherQA Adaptive Check were consistent with the Ethos scheduled plan, which can effectively identify the fractions that need the implementation of the Ethos adaptive plan.

CONCLUSION

This study integrated AI-based technologies and GPU-based MC technology to evaluate the dose distributions using fractional CBCT images, demonstrating remarkably high efficiency and precision to support future ART processes.

Reinforcement learning-based anatomical maps for pancreas subregion and duct segmentation

BACKGROUND

The pancreas is a complex abdominal organ with many anatomical variations, and therefore automated pancreas segmentation from medical images is a challenging application.

PURPOSE

In this paper, we present a framework for segmenting individual pancreatic subregions and the pancreatic duct from three-dimensional (3D) computed tomography (CT) images.

METHODS

A multiagent reinforcement learning (RL) network was used to detect landmarks of the head, neck, body, and tail of the pancreas, and landmarks along the pancreatic duct in a selected target CT image. Using the landmark detection results, an atlas of pancreases was nonrigidly registered to the target image, resulting in anatomical probability maps for the pancreatic subregions and duct. The probability maps were augmented with multilabel 3D U-Net architectures to obtain the final segmentation results.

RESULTS

To evaluate the performance of our proposed framework, we computed the Dice similarity coefficient (DSC) between the predicted and ground truth manual segmentations on a database of 82 CT images with manually segmented pancreatic subregions and 37 CT images with manually segmented pancreatic ducts. For the four pancreatic subregions, the mean DSC improved from 0.38, 0.44, and 0.39 with standard 3D U-Net, Attention U-Net, and shifted windowing (Swin) U-Net architectures, to 0.51, 0.47, and 0.49, respectively, when utilizing the proposed RL-based framework. For the pancreatic duct, the RL-based framework achieved a mean DSC of 0.70, significantly outperforming the standard approaches and existing methods on different datasets.

CONCLUSIONS

The resulting accuracy of the proposed RL-based segmentation framework demonstrates an improvement against segmentation with standard U-Net architectures.

A multi-task generative model for simultaneous post-contrast MR image synthesis and brainstem glioma segmentation

OBJECTIVES

This study aims to generate post-contrast MR images reducing the exposure of gadolinium-based contrast agents (GBCAs) for brainstem glioma (BSG) detection, simultaneously delineating the BSG lesion, and providing high-resolution contrast information.

METHODS

A retrospective cohort of 30 patients diagnosed with brainstem glioma was included. Multi-contrast images, including pre-contrast T1 weighted (pre-T1w), T2 weighted (T2w), arterial spin labeling (ASL) and post-contrast T1w images, were collected. A multi-task generative model was developed to synthesize post-contrast T1w images and simultaneously segment BSG masks from the multi-contrast inputs. Performance evaluation was conducted using peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), and mean absolute error (MAE) metrics. A perceptual study was also undertaken to assess diagnostic quality.

RESULTS

The proposed model achieved SSIM of 0.86 ± 0.04, PSNR of 26.33 ± 0.05 and MAE of 57.20 ± 20.50 for post-contrast T1w image synthesis. Automated delineation of the BSG lesions achieved Dice similarity coefficient (DSC) score of 0.88 ± 0.27.

CONCLUSIONS

The proposed model can synthesize high-quality post-contrast T1w images and accurately segment the BSG region, yielding satisfactory DSC scores.

CLINICAL RELEVANCE STATEMENT

The synthesized post-contrast MR image presented in this study has the potential to reduce the usage of gadolinium-based contrast agents, which may pose risks to patients. Moreover, the automated segmentation method proposed in this paper aids radiologists in accurately identifying the brainstem glioma lesion, facilitating the diagnostic process.

Spatial correlation between in vivo imaging and immunohistochemical biomarkers: A methodological study

In this study, we present a method that enables voxel-by-voxel comparison of in vivo imaging to immunohistochemistry (IHC) biomarkers. As a proof of concept, we investigated the spatial correlation between dynamic contrast enhanced (DCE-)CT parameters and IHC biomarkers Ki-67 (proliferation), HIF-1α (hypoxia), and CD45 (immune cells). 54 whole-mount tumor slices of 15 laryngeal and hypopharyngeal carcinomas were immunohistochemically stained and digitized. Heatmaps of biomarker positivity were created and registered to DCE-CT parameter maps. The adiabatic approximation to the tissue homogeneity model was used to fit the following DCE parameters: Ktrans (transfer constant), Ve (extravascular and extracellular space), and Vi (intravascular space). Both IHC and DCE maps were downsampled to 4 × 4 × 3 mm[3] voxels. The mean values per tumor were used to calculate the between-subject correlations between parameters. For the within-subject (spatial) correlation, values of all voxels within a tumor were compared using the repeated measures correlation (rrm). No between-subject correlations were found between IHC biomarkers and DCE parameters, whereas we found multiple significant within-subject correlations: Ve and Ki-67 (rrm = -0.17, P < .001), Ve and HIF-1α (rrm = -0.12, P < .001), Ktrans and CD45 (rrm = 0.13, P < .001), Vi and CD45 (rrm = 0.16, P < .001), and Vi and Ki-67 (rrm = 0.08, P = .003). The strongest correlation was found between IHC biomarkers Ki-67 and HIF-1α (rrm = 0.35, P < .001). This study shows the technical feasibility of determining the 3 dimensional spatial correlation between histopathological biomarker heatmaps and in vivo imaging. It also shows that between-subject correlations do not reflect within-subject correlations of parameters.

Specific and comprehensive genetic targeting reveals brain-wide distribution and synaptic input patterns of GABAergic axo-axonic interneurons

Axo-axonic cells (AACs), also called chandelier cells (ChCs) in the cerebral cortex, are the most distinctive type of GABAergic interneurons described in the neocortex, hippocampus, and basolateral amygdala (BLA). AACs selectively innervate glutamatergic projection neurons (PNs) at their axon initial segment (AIS), thus may exert decisive control over PN spiking and regulate PN functional ensembles. However, the brain-wide distribution, synaptic connectivity, and circuit function of AACs remain poorly understood, largely due to the lack of specific and reliable experimental tools. Here, we have established an intersectional genetic strategy that achieves specific and comprehensive targeting of AACs throughout the mouse brain based on their lineage (Nkx2.1) and molecular (Unc5b, Pthlh) markers. We discovered that AACs are deployed across essentially all the pallium-derived brain structures, including not only the dorsal pallium-derived neocortex and medial pallium-derived hippocampal formation, but also the lateral pallium-derived claustrum-insular complex, and the ventral pallium-derived extended amygdaloid complex and olfactory centers. AACs are also abundant in anterior olfactory nucleus, taenia tecta, and lateral septum. AACs show characteristic variations in density across neocortical areas and layers and across subregions of the hippocampal formation. Neocortical AACs comprise multiple laminar subtypes with distinct dendritic and axonal arborization patterns. Retrograde monosynaptic tracing from AACs across neocortical, hippocampal, and BLA regions reveal shared as well as distinct patterns of synaptic input. Specific and comprehensive targeting of AACs facilitates the study of their developmental genetic program and circuit function across brain structures, providing a ground truth platform for understanding the conservation and variation of a bona fide cell type across brain regions and species.

Microstructural characterization of multiple sclerosis lesion phenotypes using multiparametric longitudinal analysis

BACKGROUND AND OBJECTIVES

In multiple sclerosis (MS), slowly expanding lesions were shown to be associated with worse disability and prognosis. Their timely detection from cross-sectional data at early disease stages could be clinically relevant to inform treatment planning. Here, we propose to use multiparametric, quantitative MRI to allow a better cross-sectional characterization of lesions with different longitudinal phenotypes.

METHODS

We analysed T1 and T2 relaxometry maps from a longitudinal cohort of MS patients. Lesions were classified as enlarging, shrinking, new or stable based on their longitudinal volumetric change using a newly developed automated technique. Voxelwise deviations were computed as z-scores by comparing individual patient data to T1, T2 and T2/T1 normative values from healthy subjects. We studied the distribution of microstructural properties inside lesions and within perilesional tissue.

RESULTS AND CONCLUSIONS

Stable lesions exhibited the highest T1 and T2 z-scores in lesion tissue, while the lowest values were observed for new lesions. Shrinking lesions presented the highest T1 z-scores in the first perilesional ring while enlarging lesions showed the highest T2 z-scores in the same region. Finally, a classification model was trained to predict the longitudinal lesion type based on microstructural metrics and feature importance was assessed. Z-scores estimated in lesion and perilesional tissue from T1, T2 and T2/T1 quantitative maps carry discriminative and complementary information to classify longitudinal lesion phenotypes, hence suggesting that multiparametric MRI approaches are essential for a better understanding of the pathophysiological mechanisms underlying disease activity in MS lesions.

Quantitative muscle MRI in sporadic inclusion body myositis (sIBM): A prospective cohort study

Background

Sporadic inclusion body myositis (sIBM) is the predominant idiopathic inflammatory myopathy (IIM) in older people. Limitations of classical clinical assessments have been discussed as possible explanations for failed clinical trials, underlining the need for more sensitive outcome measures. Quantitative muscle MRI (qMRI) is a promising candidate for evaluating and monitoring sIBM.

Objective

Longitudinal assessment of qMRI in sIBM patients.

Methods

We evaluated fifteen lower extremity muscles of 12 sIBM patients (5 females, mean age 69.6, BMI 27.8) and 12 healthy age- and gender-matched controls. Seven patients and matched controls underwent a follow-up evaluation after one year. Clinical assessment included testing for muscle strength with Quick Motor Function Measure (QMFM), IBM functional rating scale (IBM-FRS), and gait analysis (6-minute walking distance). 3T-MRI scans of the lower extremities were performed, including a Dixon-based sequence, T2 mapping and Diffusion Tensor Imaging. The qMRI-values fat-fraction (FF), water T2 relaxation time (wT2), fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (λ1), and radial diffusivity (RD) were analysed.

Results

Compared to healthy controls, significant differences for all qMRI parameters averaged over all muscles were found in sIBM using a MANOVA (p < 0.001). In low-fat muscles (FF < 10% ), a significant increase of wT2 and FA with an accompanying decrease of MD, λ1, and RD was observed (p≤0.020). The highest correlation with clinical assessments was found for wT2 values in thigh muscles (r≤-0.634). Significant changes of FF (+3.0% ), wT2 (+0.6 ms), MD (-0.04 10 - 3mm2/s), λ1 (-0.05 10 - 3mm2/s), and RD (-0.03 10 - 3mm2/s) were observed in the longitudinal evaluation of sIBM patients (p≤0.001). FA showed no significant change (p = 0.242).

Conclusion

qMRI metrics correlate with clinical findings and can reflect different ongoing pathophysiological mechanisms. While wT2 is an emerging marker of disease activity, the role of diffusion metrics, possibly reflecting changes in fibre size and intracellular deposits, remains subject to further investigations.

Cardiorespiratory motion characteristics and their dosimetric impact on cardiac stereotactic body radiotherapy

BACKGROUND

Cardiac stereotactic body radiotherapy (CSBRT) is an emerging and promising noninvasive technique for treating refractory arrhythmias utilizing highly precise, single or limited-fraction high-dose irradiations. This method promises to revolutionize the treatment of cardiac conditions by delivering targeted therapy with minimal exposure to surrounding healthy tissues. However, the dynamic nature of cardiorespiratory motion poses significant challenges to the precise delivery of dose in CSBRT, introducing potential variabilities that can impact treatment efficacy. The complexities of the influence of cardiorespiratory motion on dose distribution are compounded by interplay and blurring effects, introducing additional layers of dose uncertainty. These effects, critical to the understanding and improvement of the accuracy of CSBRT, remain unexplored, presenting a gap in current clinical literature.

PURPOSE

To investigate the cardiorespiratory motion characteristics in arrhythmia patients and the dosimetric impact of interplay and blurring effects induced by cardiorespiratory motion on CSBRT plan quality.

METHODS

The position and volume variations in the substrate target and cardiac substructures were evaluated in 12 arrhythmia patients using displacement maximum (DMX) and volume metrics. Moreover, a four-dimensional (4D) dose reconstruction approach was employed to examine the dose uncertainty of the cardiorespiratory motion.

RESULTS

Cardiac pulsation induced lower DMX than respiratory motion but increased the coefficient of variation and relative range in cardiac substructure volumes. The mean DMX of the substrate target was 0.52 cm (range: 0.26-0.80 cm) for cardiac pulsation and 0.82 cm (range: 0.32-2.05 cm) for respiratory motion. The mean DMX of the cardiac structure ranged from 0.15 to 1.56 cm during cardiac pulsation and from 0.35 to 1.89 cm during respiratory motion. Cardiac pulsation resulted in an average deviation of -0.73% (range: -4.01%-4.47%) in V25 between the 3D and 4D doses. The mean deviations in the homogeneity index (HI) and gradient index (GI) were 1.70% (range: -3.10%-4.36%) and 0.03 (range: -0.14-0.11), respectively. For cardiac substructures, the deviations in D50 due to cardiac pulsation ranged from -1.88% to 1.44%, whereas the deviations in Dmax ranged from -2.96% to 0.88% of the prescription dose. By contrast, the respiratory motion led to a mean deviation of -1.50% (range: -10.73%-4.23%) in V25. The mean deviations in HI and GI due to respiratory motion were 4.43% (range: -3.89%-13.98%) and 0.18 (range: -0.01-0.47) (p < 0.05), respectively. Furthermore, the deviations in D50 and Dmax in cardiac substructures for the respiratory motion ranged from -0.28% to 4.24% and -4.12% to 1.16%, respectively.

CONCLUSIONS

Cardiorespiratory motion characteristics vary among patients, with the respiratory motion being more significant. The intricate cardiorespiratory motion characteristics and CSBRT plan complexity can induce substantial dose uncertainty. Therefore, assessing individual motion characteristics and 4D dose reconstruction techniques is critical for implementing CSBRT without compromising efficacy and safety.

Cell atlas of the regenerating human liver after portal vein embolization

The liver has the remarkable capacity to regenerate. In the clinic, regeneration is induced by portal vein embolization, which redirects portal blood flow, resulting in liver hypertrophy in locations with increased blood supply, and atrophy of embolized segments. Here, we apply single-cell and single-nucleus transcriptomics on healthy, hypertrophied, and atrophied patient-derived liver samples to explore cell states in the regenerating liver. Our data unveils pervasive upregulation of genes associated with developmental processes, cellular adhesion, and inflammation in post-portal vein embolization liver, disrupted portal-central hepatocyte zonation, and altered cell subtype composition of endothelial and immune cells. Interlineage crosstalk analysis reveals mesenchymal cells as an interaction hub between immune and endothelial cells, and highlights the importance of extracellular matrix proteins in liver regeneration. Moreover, we establish tissue-scale iterative indirect immunofluorescence imaging for high-dimensional spatial analysis of perivascular microenvironments, uncovering changes to tissue architecture in regenerating liver lobules. Altogether, our data is a rich resource revealing cellular and histological changes in human liver regeneration.

GZMK+CD8+ T cells Target A Specific Acinar Cell Type in Sjögren's Disease

Sjögren's Disease (SjD) is a systemic autoimmune disease without a clear etiology or effective therapy. Utilizing unbiased single-cell and spatial transcriptomics to analyze human minor salivary glands in health and disease we developed a comprehensive understanding of the cellular landscape of healthy salivary glands and how that landscape changes in SjD patients. We identified novel seromucous acinar cell types and identified a population of PRR4+CST3+WFDC2- seromucous acinar cells that are particularly targeted in SjD. Notably, GZMK+CD8 T cells, enriched in SjD, exhibited a cytotoxic phenotype and were physically associated with immune-engaged epithelial cells in disease. These findings shed light on the immune response's impact on transitioning acinar cells with high levels of secretion and explain the loss of this specific cell population in SjD. This study explores the complex interplay of varied cell types in the salivary glands and their role in the pathology of Sjögren's Disease.

Prospective longitudinal cohort study of quantitative muscle magnetic resonance imaging in a healthy control population

Quantitative muscle magnetic resonance imaging (qMRI) is a valuable methodology for assessing muscular injuries and neuromuscular disorders. Notably, muscle diffusion tensor imaging (DTI) gives insights into muscle microstructural and macrostructural characteristics. However, the long-term reproducibility and robustness of these measurements remain relatively unexplored. The purpose of this prospective longitudinal cohort study was to assess the long-term robustness and range of variation of qMRI parameters, especially DTI metrics, in the lower extremity muscles of healthy controls under real-life conditions. Twelve volunteers (seven females, age 44.1 ± 12.1 years, body mass index 23.3 ± 2.0 kg/m2) underwent five leg muscle MRI sessions every 20 ± 4 weeks over a total period of 1.5 years. A multiecho gradient-echo Dixon-based sequence, a multiecho spin-echo T2-mapping sequence, and a spin-echo echo planar imaging diffusion-weighted sequence were acquired bilaterally with a Philips 3-T Achieva MR System using a 16-channel torso coil. Fifteen leg muscles were segmented in both lower extremities. qMRI parameters, including fat fraction (FF), water T2 relaxation time, and the diffusion metrics fractional anisotropy (FA) and mean diffusivity (MD), were evaluated. Coefficients of variance (wsCV) and intraclass correlation coefficients (ICCs) were calculated to assess the reproducibility of qMRI parameters. The standard error of measurement (SEM) and the minimal detectable change (MDC) were calculated to determine the range of variation. All tests were applied to all muscles and, subsequently, to each muscle separately. wsCV showed good reproducibility (≤ 10%) for all qMRI parameters in all muscles. The ICCs revealed excellent agreement between time points (FF = 0.980, water T2 = 0.941, FA = 0.952, MD = 0.948). Random measurement errors assessed by SEM and the MDC were low (< 12%). In conclusion, in this study, we showed that qMRI parameters in healthy volunteers living normal lives are stable over 18 months, thereby defining a benchmark for the expected range of variation over time.

Control of innate olfactory valence by segregated cortical amygdala circuits

Animals perform innate behaviors that are stereotyped responses to specific evolutionarily relevant stimuli in the absence of prior learning or experience. These behaviors can be reduced to an axis of valence, whereby specific odors evoke approach or avoidance. The cortical amygdala (plCoA) mediates innate attraction and aversion to odor. However, little is known about how this brain area gives rise to behaviors of opposing motivational valence. Here, we sought to define the circuit features of plCoA that give rise to innate olfactory behaviors of valence. We characterized the physiology, gene expression, and projections of this structure, identifying a divergent, topographic organization that selectively controls innate attraction and avoidance to odor. First, we examined odor-evoked responses in these areas and found sparse encoding of odor identity, but not valence. We next considered a topographic organization and found that optogenetic stimulation of the anterior and posterior domains of plCoA elicits attraction and avoidance, respectively, suggesting a functional axis for valence. Using single cell and spatial RNA sequencing, we identified the molecular cell types in plCoA, revealing an anteroposterior gradient in cell types, whereby anterior glutamatergic neurons preferentially express Slc17a6 and posterior neurons express Slc17a7. Activation of these respective cell types recapitulates appetitive and aversive valence behaviors, and chemogenetic inhibition reveals partial necessity for valence responses to innate appetitive or aversive odors. Finally, we identified topographically organized circuits defined by projections, whereby anterior neurons preferentially project to medial amygdala, and posterior neurons preferentially project to nucleus accumbens, which are respectively sufficient and necessary for innate negative and positive olfactory valence. Together, these data advance our understanding of how the olfactory system generates stereotypic, hardwired attraction and avoidance, and supports a model whereby distinct, topographically distributed plCoA populations direct innate olfactory valence responses by signaling to divergent valence-specific targets, linking upstream olfactory identity to downstream valence behaviors, through a population code. This represents a novel circuit motif in which valence encoding is represented not by the firing properties of individual neurons, but by population level identity encoding that is routed through divergent targets to mediate distinct valence.

Multi-modality imaging parameters that predict rapid tumor regression in head and neck radiotherapy

Background and purpose

Volume regression during radiotherapy can indicate patient-specific treatment response. We aimed to identify pre-treatment multimodality imaging (MMI) metrics from positron emission tomography (PET), magnetic resonance imaging (MRI), and computed tomography (CT) that predict rapid tumor regression during radiotherapy in human papilloma virus (HPV) associated oropharyngeal carcinoma.

Materials and methods

Pre-treatment FDG PET-CT, diffusion-weighted MRI (DW-MRI), and intra-treatment (at 1, 2, and 3 weeks) MRI were acquired in 72 patients undergoing chemoradiation therapy for HPV+ oropharyngeal carcinoma. Nodal gross tumor volumes were delineated on longitudinal images to measure intra-treatment volume changes. Pre-treatment PET standardized uptake value (SUV), CT Hounsfield Unit (HU), and non-gaussian intravoxel incoherent motion DW-MRI metrics were computed and correlated with volume changes. Intercorrelations between MMI metrics were also assessed using network analysis. Validation was carried out on a separate cohort (N = 64) for FDG PET-CT.

Results

Significant correlations with volume loss were observed for baseline FDG SUVmean (Spearman ρ = 0.46, p < 0.001), CT HUmean (ρ = 0.38, p = 0.001), and DW-MRI diffusion coefficient, Dmean (ρ = -0.39, p < 0.001). Network analysis revealed 41 intercorrelations between MMI and volume loss metrics, but SUVmean remained a statistically significant predictor of volume loss in multivariate linear regression (p = 0.01). Significant correlations were also observed for SUVmean in the validation cohort in both primary (ρ = 0.30, p = 0.02) and nodal (ρ = 0.31, p = 0.02) tumors.

Conclusions

Multiple pre-treatment imaging metrics were correlated with rapid nodal gross tumor volume loss during radiotherapy. FDG-PET SUV in particular exhibited significant correlations with volume regression across the two cohorts and in multivariate analysis.

Validation of an organ mapping antibody panel for cyclical immunofluorescence microscopy on normal human kidneys

The lack of standardization in antibody validation remains a major contributor to irreproducibility of human research. To address this, we have applied a standardized approach to validate a panel of antibodies to identify 18 major cell types and 5 extracellular matrix compartments in the human kidney by immunofluorescence (IF) microscopy. We have used these to generate an organ mapping antibody panel for two-dimensional (2-D) and three-dimensional (3-D) cyclical IF (CyCIF) to provide a more detailed method for evaluating tissue segmentation and volumes using a larger panel of markers than would normally be possible using standard fluorescence microscopy. CyCIF also makes it possible to perform multiplexed IF microscopy of whole slide images, which is a distinct advantage over other multiplexed imaging technologies that are applicable to limited fields of view. This enables a broader view of cell distributions across larger anatomical regions, allowing a better chance to capture localized regions of dysfunction in diseased tissues. These methods are broadly accessible to any laboratory with a fluorescence microscope, enabling spatial cellular phenotyping in normal and disease states. We also provide a detailed solution for image alignment between CyCIF cycles that can be used by investigators to perform these studies without programming experience using open-sourced software. This ability to perform multiplexed imaging without specialized instrumentation or computational skills opens the door to integration with more highly dimensional molecular imaging modalities such as spatial transcriptomics and imaging mass spectrometry, enabling the discovery of molecular markers of specific cell types, and how these are altered in disease.NEW & NOTEWORTHY We describe here validation criteria used to define on organ mapping panel of antibodies that can be used to define 18 cell types and five extracellular matrix compartments using cyclical immunofluorescence (CyCIF) microscopy. As CyCIF does not require specialized instrumentation, and image registration required to assemble CyCIF images can be performed by any laboratory without specialized computational skills, this technology is accessible to any laboratory with access to a fluorescence microscope and digital scanner.

Motion-Compensated MR CINE Reconstruction With Reconstruction-Driven Motion Estimation

In cardiac CINE, motion-compensated MR reconstruction (MCMR) is an effective approach to address highly undersampled acquisitions by incorporating motion information between frames. In this work, we propose a novel perspective for addressing the MCMR problem and a more integrated and efficient solution to the MCMR field. Contrary to state-of-the-art (SOTA) MCMR methods which break the original problem into two sub-optimization problems, i.e. motion estimation and reconstruction, we formulate this problem as a single entity with one single optimization. Our approach is unique in that the motion estimation is directly driven by the ultimate goal, reconstruction, but not by the canonical motion-warping loss (similarity measurement between motion-warped images and target images). We align the objectives of motion estimation and reconstruction, eliminating the drawbacks of artifacts-affected motion estimation and therefore error-propagated reconstruction. Further, we can deliver high-quality reconstruction and realistic motion without applying any regularization/smoothness loss terms, circumventing the non-trivial weighting factor tuning. We evaluate our method on two datasets: 1) an in-house acquired 2D CINE dataset for the retrospective study and 2) the public OCMR cardiac dataset for the prospective study. The conducted experiments indicate that the proposed MCMR framework can deliver artifact-free motion estimation and high-quality MR images even for imaging accelerations up to 20x, outperforming SOTA non-MCMR and MCMR methods in both qualitative and quantitative evaluation across all experiments. The code is available at https://github.com/JZPeterPan/MCMR-Recon-Driven-Motion.

Unsupervised lung CT image registration via stochastic decomposition of deformation fields

We address the problem of lung CT image registration, which underpins various diagnoses and treatments for lung diseases. The main crux of the problem is the large deformation that the lungs undergo during respiration. This physiological process imposes several challenges from a learning point of view. In this paper, we propose a novel training scheme, called stochastic decomposition, which enables deep networks to effectively learn such a difficult deformation field during lung CT image registration. The key idea is to stochastically decompose the deformation field, and supervise the registration by synthetic data that have the corresponding appearance discrepancy. The stochastic decomposition allows for revealing all possible decompositions of the deformation field. At the learning level, these decompositions can be seen as a prior to reduce the ill-posedness of the registration yielding to boost the performance. We demonstrate the effectiveness of our framework on Lung CT data. We show, through extensive numerical and visual results, that our technique outperforms existing methods.

Investigating the use of comprehensive motion monitoring for intrafraction 3D drift assessment of hypofractionated prostate cancer patients on a 1.5T magnetic resonance imaging radiotherapy system

This work investigates the use of a multi-2D cine magnetic resonance imaging-based comprehensive motion monitoring (CMM) system for the assessment of prostate intrafraction 3D drifts. The data of six healthy volunteers were analyzed and the values of a clinically-relevant registration quality factor metric exported by CMM were presented. Additionally, the CMM-derived prostate motion was compared to a 3D-based reference and the 2D-3D tracking agreement was reported. Due to the low quality of SI motion tracking (often > 2 mm tracking mismatch between anatomical planes) we conclude that further improvements are desirable prior to clinical introduction of CMM for prostate drift corrections.

System-level time computation and representation in the suprachiasmatic nucleus revealed by large-scale calcium imaging and machine learning

The suprachiasmatic nucleus (SCN) is the mammalian central circadian pacemaker with heterogeneous neurons acting in concert while each neuron harbors a self-sustained molecular clockwork. Nevertheless, how system-level SCN signals encode time of the day remains enigmatic. Here we show that population-level Ca2+ signals predict hourly time, via a group decision-making mechanism coupled with a spatially modular time feature representation in the SCN. Specifically, we developed a high-speed dual-view two-photon microscope for volumetric Ca2+ imaging of up to 9000 GABAergic neurons in adult SCN slices, and leveraged machine learning methods to capture emergent properties from multiscale Ca2+ signals as a whole. We achieved hourly time prediction by polling random cohorts of SCN neurons, reaching 99.0% accuracy at a cohort size of 900. Further, we revealed that functional neuron subtypes identified by contrastive learning tend to aggregate separately in the SCN space, giving rise to bilaterally symmetrical ripple-like modular patterns. Individual modules represent distinctive time features, such that a module-specifically learned time predictor can also accurately decode hourly time from random polling of the same module. These findings open a new paradigm in deciphering the design principle of the biological clock at the system level.

UDRSNet: An unsupervised deformable registration module based on image structure similarity

BACKGROUND

Image registration is a challenging problem in many clinical tasks, but deep learning has made significant progress in this area over the past few years. Real-time and robust registration has been made possible by supervised transformation estimation. However, the quality of registrations using this framework depends on the quality of ground truth labels such as displacement field.

PURPOSE

To propose a simple and reliable method for registering medical images based on image structure similarity in a completely unsupervised manner.

METHODS

We proposed a deep cascade unsupervised deformable registration approach to align images without reliable clinical data labels. Our basic network was composed of a displacement estimation module (ResUnet) and a deformation module (spatial transformer layers). We adoptedl 2 $l_2$ -norm to regularize the deformation field instead of the traditionall 1 $l_1$ -norm regularization. Additionally, we utilized structural similarity (ssim) estimation during the training stage to enhance the structural consistency between the deformed images and the reference images.

RESULTS

Experiments results indicated that by incorporating ssim loss, our cascaded methods not only achieved higher dice score of 0.9873, ssim score of 0.9559, normalized cross-correlation (NCC) score of 0.9950, and lower relative sum of squared difference (SSD) error of 0.0313 on CT images, but also outperformed the comparative methods on ultrasound dataset. The statistical t $t$ -test results also proved that these improvements of our method have statistical significance.

CONCLUSIONS

In this study, the promising results based on diverse evaluation metrics have demonstrated that our model is simple and effective in deformable image registration (DIR). The generalization ability of the model was also verified through experiments on liver CT images and cardiac ultrasound images.

Improving sensitivity through data augmentation with synthetic lymph node metastases for AI-based analysis of PSMA PET-CT images

BACKGROUND

We developed a fully automated artificial intelligence (AI)AI-based-based method for detecting suspected lymph node metastases in prostate-specific membrane antigen (PSMA)(PSMA) positron emission tomography-computed tomography (PET-CT)(PET-CT) images of prostate cancer patients by using data augmentation that adds synthetic lymph node metastases to the images to expand the training set.

METHODS

Synthetic data were derived from original training images to which new synthetic lymph node metastases were added. Thus, the original training set from a previous study (n = 420) was expanded by one synthetic image for every original image (n = 840), which was used to train an AI model. The performance of the AI model was compared to that of nuclear medicine physicians and a previously developed AI model. The human readers were alternately used as a reference and compared to either another reading or AI model.

RESULTS

The new AI model had an average sensitivity of 84% for detecting lymph node metastases compared with 78% for human readings. Our previously developed AI method without synthetic data had an average sensitivity of 79%. The number of false positive lesions were slightly higher for the new AI model (average 3.3 instances per patient) compared to human readings and the previous AI model (average 2.8 instances per patient), while the number of false negative lesions was lower.

CONCLUSIONS

Creating synthetic lymph node metastases, as a form of data augmentation, on [18F]PSMA-1007F]PSMA-1007 PETPET-CT-CT images improved the sensitivity of an AI model for detecting suspected lymph node metastases. However, the number of false positive lesions increased somewhat.

Two-Stage Training Framework Using Multicontrast MRI Radiomics for IDH Mutation Status Prediction in Glioma

Purpose To develop a radiomics framework for preoperative MRI-based prediction of isocitrate dehydrogenase (IDH) mutation status, a crucial glioma prognostic indicator. Materials and Methods Radiomics features (shape, first-order statistics, and texture) were extracted from the whole tumor or the combination of nonenhancing, necrosis, and edema regions. Segmentation masks were obtained via the federated tumor segmentation tool or the original data source. Boruta, a wrapper-based feature selection algorithm, identified relevant features. Addressing the imbalance between mutated and wild-type cases, multiple prediction models were trained on balanced data subsets using random forest or XGBoost and assembled to build the final classifier. The framework was evaluated using retrospective MRI scans from three public datasets (The Cancer Imaging Archive [TCIA, 227 patients], the University of California San Francisco Preoperative Diffuse Glioma MRI dataset [UCSF, 495 patients], and the Erasmus Glioma Database [EGD, 456 patients]) and internal datasets collected from the University of Texas Southwestern Medical Center (UTSW, 356 patients), New York University (NYU, 136 patients), and University of Wisconsin-Madison (UWM, 174 patients). TCIA and UTSW served as separate training sets, while the remaining data constituted the test set (1617 or 1488 testing cases, respectively). Results The best performing models trained on the TCIA dataset achieved area under the receiver operating characteristic curve (AUC) values of 0.89 for UTSW, 0.86 for NYU, 0.93 for UWM, 0.94 for UCSF, and 0.88 for EGD test sets. The best performing models trained on the UTSW dataset achieved slightly higher AUCs: 0.92 for TCIA, 0.88 for NYU, 0.96 for UWM, 0.93 for UCSF, and 0.90 for EGD. Conclusion This MRI radiomics-based framework shows promise for accurate preoperative prediction of IDH mutation status in patients with glioma. Keywords: Glioma, Isocitrate Dehydrogenase Mutation, IDH Mutation, Radiomics, MRI Supplemental material is available for this article. Published under a CC BY 4.0 license. See also commentary by Moassefi and Erickson in this issue.

Deep learning reveals lung shape differences on baseline chest CT between mild and severe COVID-19: A multi-site retrospective study

Severe COVID-19 can lead to extensive lung disease causing lung architectural distortion. In this study we employed machine learning and statistical atlas-based approaches to explore possible changes in lung shape among COVID-19 patients and evaluated whether the extent of these changes was associated with COVID-19 severity. On a large multi-institutional dataset (N = 3443), three different populations were defined; a) healthy (no COVID-19), b) mild COVID-19 (no ventilator required), c) severe COVID-19 (ventilator required), and the presence of lung shape differences between them were explored using baseline chest CT. Significant lung shape differences were observed along mediastinal surfaces of the lungs across all severity of COVID-19 disease. Additionally, differences were seen on basal surfaces of the lung when compared between healthy and severe COVID-19 patients. Finally, an AI model (a 3D residual convolutional network) characterizing these shape differences coupled with lung infiltrates (ground-glass opacities and consolidation regions) was found to be associated with COVID-19 severity.

Embryonically active piriform cortex neurons promote intracortical recurrent connectivity during development

Neuronal activity plays a critical role in the maturation of circuits that propagate sensory information into the brain. How widely does early activity regulate circuit maturation across the developing brain? Here, we used targeted recombination in active populations (TRAP) to perform a brain-wide survey for prenatally active neurons in mice and identified the piriform cortex as an abundantly TRAPed region. Whole-cell recordings in neonatal slices revealed preferential interconnectivity within embryonically TRAPed piriform neurons and their enhanced synaptic connectivity with other piriform neurons. In vivo Neuropixels recordings in neonates demonstrated that embryonically TRAPed piriform neurons exhibit broad functional connectivity within piriform and lead spontaneous synchronized population activity during a transient neonatal period, when recurrent connectivity is strengthening. Selectively activating or silencing these neurons in neonates enhanced or suppressed recurrent synaptic strength, respectively. Thus, embryonically TRAPed piriform neurons represent an interconnected hub-like population whose activity promotes recurrent connectivity in early development.

Prior frequency guided diffusion model for limited angle (LA)-CBCT reconstruction

Objective.Cone-beam computed tomography (CBCT) is widely used in image-guided radiotherapy. Reconstructing CBCTs from limited-angle acquisitions (LA-CBCT) is highly desired for improved imaging efficiency, dose reduction, and better mechanical clearance. LA-CBCT reconstruction, however, suffers from severe under-sampling artifacts, making it a highly ill-posed inverse problem. Diffusion models can generate data/images by reversing a data-noising process through learned data distributions; and can be incorporated as a denoiser/regularizer in LA-CBCT reconstruction. In this study, we developed a diffusion model-based framework, prior frequency-guided diffusion model (PFGDM), for robust and structure-preserving LA-CBCT reconstruction.Approach.PFGDM uses a conditioned diffusion model as a regularizer for LA-CBCT reconstruction, and the condition is based on high-frequency information extracted from patient-specific prior CT scans which provides a strong anatomical prior for LA-CBCT reconstruction. Specifically, we developed two variants of PFGDM (PFGDM-A and PFGDM-B) with different conditioning schemes. PFGDM-A applies the high-frequency CT information condition until a pre-optimized iteration step, and drops it afterwards to enable both similar and differing CT/CBCT anatomies to be reconstructed. PFGDM-B, on the other hand, continuously applies the prior CT information condition in every reconstruction step, while with a decaying mechanism, to gradually phase out the reconstruction guidance from the prior CT scans. The two variants of PFGDM were tested and compared with current available LA-CBCT reconstruction solutions, via metrics including peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM).Main results.PFGDM outperformed all traditional and diffusion model-based methods. The mean(s.d.) PSNR/SSIM were 27.97(3.10)/0.949(0.027), 26.63(2.79)/0.937(0.029), and 23.81(2.25)/0.896(0.036) for PFGDM-A, and 28.20(1.28)/0.954(0.011), 26.68(1.04)/0.941(0.014), and 23.72(1.19)/0.894(0.034) for PFGDM-B, based on 120°, 90°, and 30° orthogonal-view scan angles respectively. In contrast, the PSNR/SSIM was 19.61(2.47)/0.807(0.048) for 30° for DiffusionMBIR, a diffusion-based method without prior CT conditioning.Significance. PFGDM reconstructs high-quality LA-CBCTs under very-limited gantry angles, allowing faster and more flexible CBCT scans with dose reductions.