User-Guided Segmentation of Multi-modality Medical Imaging Datasets with ITK-SNAP

Diffusion-Weighted Imaging Fluid-Attenuated Inversion Recovery Mismatch on Portable, Low-Field Magnetic Resonance Imaging Among Acute Stroke Patients

OBJECTIVE

For stroke patients with unknown time of onset, mismatch between diffusion-weighted imaging (DWI) and fluid-attenuated inversion recovery (FLAIR) magnetic resonance imaging (MRI) can guide thrombolytic intervention. However, access to MRI for hyperacute stroke is limited. Here, we sought to evaluate whether a portable, low-field (LF)-MRI scanner can identify DWI-FLAIR mismatch in acute ischemic stroke.

METHODS

Eligible patients with a diagnosis of acute ischemic stroke underwent LF-MRI acquisition on a 0.064-T scanner within 24 h of last known well. Qualitative and quantitative metrics were evaluated. Two trained assessors determined the visibility of stroke lesions on LF-FLAIR. An image coregistration pipeline was developed, and the LF-FLAIR signal intensity ratio (SIR) was derived.

RESULTS

The study included 71 patients aged 71 ± 14 years and a National Institutes of Health Stroke Scale of 6 (interquartile range 3-14). The interobserver agreement for identifying visible FLAIR hyperintensities was high (κ = 0.85, 95% CI 0.70-0.99). Visual DWI-FLAIR mismatch had a 60% sensitivity and 82% specificity for stroke patients <4.5 h, with a negative predictive value of 93%. LF-FLAIR SIR had a mean value of 1.18 ± 0.18 <4.5 h, 1.24 ± 0.39 4.5-6 h, and 1.40 ± 0.23 >6 h of stroke onset. The optimal cut-point for LF-FLAIR SIR was 1.15, with 85% sensitivity and 70% specificity. A cut-point of 6.6 h was established for a FLAIR SIR <1.15, with an 89% sensitivity and 62% specificity.

INTERPRETATION

A 0.064-T portable LF-MRI can identify DWI-FLAIR mismatch among patients with acute ischemic stroke. Future research is needed to prospectively validate thresholds and evaluate a role of LF-MRI in guiding thrombolysis among stroke patients with uncertain time of onset. ANN NEUROL 2024;96:321-331.

Ultrasound-based deep learning radiomics nomogram for differentiating mass mastitis from invasive breast cancer

BACKGROUND

The purpose of this study is to develop and validate the potential value of the deep learning radiomics nomogram (DLRN) based on ultrasound to differentiate mass mastitis (MM) and invasive breast cancer (IBC).

METHODS

50 cases of MM and 180 cases of IBC with ultrasound Breast Imaging Reporting and Data System 4 category were recruited (training cohort, n = 161, validation cohort, n = 69). Based on PyRadiomics and ResNet50 extractors, radiomics and deep learning features were extracted, respectively. Based on supervised machine learning methods such as logistic regression, random forest, and support vector machine, as well as unsupervised machine learning methods using K-means clustering analysis, the differences in features between MM and IBC were analyzed to develop DLRN. The performance of DLRN had been evaluated by receiver operating characteristic curve, calibration, and clinical practicality.

RESULTS

Supervised machine learning results showed that compared with radiomics models, especially random forest models, deep learning models were better at recognizing MM and IBC. The area under the curve (AUC) of the validation cohort was 0.84, the accuracy was 0.83, the sensitivity was 0.73, and the specificity was 0.83. Compared to radiomics or deep learning models, DLRN even further improved discrimination ability (AUC of 0.90 and 0.90, accuracy of 0.83 and 0.88 for training and validation cohorts), which had better clinical benefits and good calibratability. In addition, the information heterogeneity of deep learning features in MM and IBC was validated again through unsupervised machine learning clustering analysis, indicating that MM had a unique features phenotype.

CONCLUSION

The DLRN developed based on radiomics and deep learning features of ultrasound images has potential clinical value in effectively distinguishing between MM and IBC. DLRN breaks through visual limitations and quantifies more image information related to MM based on computers, further utilizing machine learning to effectively utilize this information for clinical decision-making. As DLRN becomes an autonomous screening system, it will improve the recognition rate of MM in grassroots hospitals and reduce the possibility of incorrect treatment and overtreatment.

Laminar dissociation of feedforward and feedback in high-level ventral visual cortex during imagery and perception

Visual imagery and perception share neural machinery but rely on different information flow. While perception is driven by the integration of sensory feedforward and internally generated feedback information, imagery relies on feedback only. This suggests that although imagery and perception may activate overlapping brain regions, they do so in informationally distinctive ways. Using lamina-resolved MRI at 7 T, we measured the neural activity during imagery and perception of faces and scenes in high-level ventral visual cortex at the mesoscale of laminar organization that distinguishes feedforward from feedback signals. We found distinctive laminar profiles for imagery and perception of scenes and faces in the parahippocampal place area and the fusiform face area, respectively. Our findings provide insight into the neural basis of the phenomenology of visual imagery versus perception and shed new light into the mesoscale organization of feedforward and feedback information flow in high-level ventral visual cortex.

Efficacy and safety of hepatic arterial infusion chemotherapy combined with tyrosine kinase inhibitors and immune checkpoint inhibitors in the treatment of advanced hepatocellular carcinoma with portal vein tumor thrombosis in the main trunk

Purpose

To evaluate the efficacy and safety of mFOLFOX-based hepatic arterial infusion chemotherapy (HAIC) combined with tyrosine kinase inhibitors (TKIs) and immune checkpoint inhibitors (ICIs) in the treatment of advanced hepatocellular carcinoma (HCC) with portal vein tumor thrombosis (PVTT).

Methods

This retrospective study included patients who received mFOLFOX-based HAIC combined with TKIs and ICIs from January 2021 to January 2023. The primary outcome was the objective response rate of PVTT response, and the secondary outcomes were 6-month, 1-year survival rate, overall survival (OS), and corresponding adverse events and complications were also evaluated. PVTT responses were assessed using ITK-SNAP software.

Results

A total of 37 patients were included in the analysis, 18.92% achieved a complete response and 56.76% achieved a partial response in PVTT response. The objective response rate (ORR) of PVTT was 75.68%. The 6-month survival rate was 89%, the 1-year survival rate was 66%, and the median OS was 15.8 months. In univariate analysis, Child-Pugh score (P=0.010) was important factor for predicting OS; in multivariate analysis, Child-Pugh score (P=0.015, HR= 3.089, 95%CI: 1.250-7.633) was the important factor for predicting OS. In terms of adverse reactions, the most common adverse reactions associated with HAIC are pain and thrombocytopenia associated with oxaliplatin.

Conclusion

FOLFOX-based HAIC combined with TKIs and ICIs induced an objective response rate of 75.68% in PVTT.

Clinical signicance

FOLFOX-based HAIC combined with TKIs and ICIs provides more treatment options for PVTT.

Quantitative susceptibility mapping in amyotrophic lateral sclerosis: automatic quantification of the magnetic susceptibility in the subcortical nuclei

Objective: Previous studies have suggested a link between dysregulation of cortical iron levels and neuronal loss in amyotrophic lateral sclerosis (ALS) patients. However, few studies have reported differences in quantitative susceptibility mapping (QSM) values in subcortical nuclei between patients with ALS and healthy controls (HCs). Methods: MRI was performed using a 3 Tesla Prisma scanner (64-channel head coil), including 3D T1-MPRAGE and multi-echo 3D GRE for QSM reconstruction. Automated QSM segmentation was used to measure susceptibility values in the subcortical nuclei, which were compared between the groups. Correlations with clinical scales were analyzed. Group comparisons were performed using independent t-tests, with p < 0.05 considered significant. Correlations were assessed using Pearson's correlation, with p < 0.05 considered significant. Cohen's d was reported to compare the standardized mean difference (SMD) of QSM. Results: Twelve patients with limb-onset ALS (mean age 48.7 years, 75% male) and 13 age-, sex-, and handedness-matched HCs (mean age 44.6 years, 69% male) were included. Compared to HCs, ALS patients demonstrated significantly lower susceptibility in the left caudate nucleus (CN) (SMD = -0.845), right CN (SMD = -0.851), whole CN (SMD = -1.016), and left subthalamic nucleus (STN) (SMD = -1.000). Susceptibility in the left putamen (SMD = -0.857), left thalamus (SMD = -1.081), and whole thalamus (SMD = -0.968) was significantly higher in the patients. The susceptibility of the substantia nigra (SN), CN, and pulvinar was positively correlated with disease duration. Conclusions: QSM detects abnormal iron accumulation patterns in the subcortical gray matter of ALS patients, which correlates with disease characteristics, supporting its potential as a neuroimaging biomarker.

Enlarged perivascular spaces in the basal ganglia are associated with arteries not veins

Enlarged perivascular spaces (EPVS) are common in cerebral small vessel disease (CSVD) and have been identified as a marker of dysfunctional brain clearance. However, it remains unknown if the enlargement occurs predominantly around arteries or veins. We combined in vivo ultra-high-resolution MRI and histopathology to investigate the spatial relationship of veins and arteries with EPVS within the basal ganglia (BG). Furthermore, we assessed the relationship between the EPVS and measures of blood-flow (blood-flow velocity, pulsatility index) in the small arteries of the BG. Twenty-four healthy controls, twelve non-CAA CSVD patients, and five probable CAA patients underwent a 3 tesla [T] and 7T MRI-scan, and EPVS, arteries, and veins within the BG were manually segmented. Furthermore, the scans were co-registered. Six autopsy-cases were also assessed. In the BG, EPVS were significantly closer to and overlapped more frequently with arteries than with veins. Histological analysis showed a higher proportion of BG EPVS surrounding arteries than veins. Finally, the pulsatility index of BG arteries correlated with EPVS volume. Our results are in line with previous works and establish a pathophysiological relationship between arteries and EPVS, contributing to elucidating perivascular clearance routes in the human brain.

Brain network topology and its cognitive impact in adult glioma survivors

Structural brain network topology can be altered in case of a brain tumor, due to both the tumor itself and its treatment. In this study, we explored the role of structural whole-brain and nodal network metrics and their association with cognitive functioning. Fifty WHO grade 2-3 adult glioma survivors (> 1-year post-therapy) and 50 matched healthy controls underwent a cognitive assessment, covering six cognitive domains. Raw cognitive assessment scores were transformed into w-scores, corrected for age and education. Furthermore, based on multi-shell diffusion-weighted MRI, whole-brain tractography was performed to create weighted graphs and to estimate whole-brain and nodal graph metrics. Hubs were defined based on nodal strength, betweenness centrality, clustering coefficient and shortest path length in healthy controls. Significant differences in these metrics between patients and controls were tested for the hub nodes (i.e. n = 12) and non-hub nodes (i.e. n = 30) in two mixed-design ANOVAs. Group differences in whole-brain graph measures were explored using Mann-Whitney U tests. Graph metrics that significantly differed were ultimately correlated with the cognitive domain-specific w-scores. Bonferroni correction was applied to correct for multiple testing. In survivors, the bilateral putamen were significantly less frequently observed as a hub (pbonf < 0.001). These nodes' assortativity values were positively correlated with attention (r(90) > 0.573, pbonf < 0.001), and proxy IQ (r(90) > 0.794, pbonf < 0.001). Attention and proxy IQ were significantly more often correlated with assortativity of hubs compared to non-hubs (pbonf < 0.001). Finally, the whole-brain graph measures of clustering coefficient (r = 0.685), global (r = 0.570) and local efficiency (r = 0.500) only correlated with proxy IQ (pbonf < 0.001). This study demonstrated potential reorganization of hubs in glioma survivors. Assortativity of these hubs was specifically associated with cognitive functioning, which could be important to consider in future modeling of cognitive outcomes and risk classification in glioma survivors.

Age-related decline in thickness and surface area in the cortical surface and hippocampus: lifespan trajectories and decade-by-decade analyses

Previous studies on age-related changes in cortical and hippocampal morphology were not designed or able to reveal the complex spatial patterns of changes across the lifespan. To this end, the current study examined these changes in a decade-by-decade manner by comparing consecutive age decades at the vertex-wise level. Additionally, the lifespan trajectories of cortical/hippocampal mean thickness and total surface area were modeled and plotted out to provide an overview of their age-related changes. Using two lifespan datasets (Ntotal = 1378; 18 ≤ age ≤ 100), vertex-wise thickness and surface area measurements were extracted from the cortical and unfolded hippocampal surfaces and analyzed using whole-brain/hippocampus vertex-wise analyses. Lifespan trajectories of cortical/hippocampal mean thickness and total surface area were modeled with generalized additive models for location, scale, and shape. These models revealed fairly linear declines in both cortical measures and inverted U-shaped trajectories for both hippocampal measures. Across the different age decades, the sizes and locations of cortical thinning clusters were highly variable across the age decades. No significant clusters of cortical surface area changes were observed across the age decades. Significant clusters of hippocampal surface area and thickness reduction were not observed until the 70s. Generally, the agreement between datasets on the hippocampal findings was much higher than those of the cortical surface. These findings revealed important nuances in the age-related changes of cortical and hippocampal morphology and cautioned against using lifespan trajectories to infer decade-by-decade changes in the cortical surface and the hippocampus.

Imaging Genomics and Multiomics: A Guide for Beginners Starting Radiomics-Based Research

Radiomics uses advanced mathematical analysis of pixel-level information from radiologic images to extract existing information in traditional imaging algorithms. It is intended to find imaging biomarkers related to the genomics of tumors or disease patterns that improve medical care by advanced detection of tumor response patterns in tumors and to assess prognosis. Radiomics expands the paradigm of medical imaging to help with diagnosis, management of diseases and prognostication, leveraging image features by extracting information that can be used as imaging biomarkers to predict prognosis and response to treatment. Radiogenomics is an emerging area in radiomics that investigates the association between imaging characteristics and gene expression profiles. There are an increasing number of research publications using different radiomics approaches without a clear consensus on which method works best. We aim to describe the workflow of radiomics along with a guide of what to expect when starting a radiomics-based research project.

Neurodesk: an accessible, flexible and portable data analysis environment for reproducible neuroimaging

Neuroimaging research requires purpose-built analysis software, which is challenging to install and may produce different results across computing environments. The community-oriented, open-source Neurodesk platform ( https://www.neurodesk.org/ ) harnesses a comprehensive and growing suite of neuroimaging software containers. Neurodesk includes a browser-accessible virtual desktop, command-line interface and computational notebook compatibility, allowing for accessible, flexible, portable and fully reproducible neuroimaging analysis on personal workstations, high-performance computers and the cloud.

Comparative validation of automated presurgical tractography based on constrained spherical deconvolution and diffusion tensor imaging with direct electrical stimulation

OBJECTIVES

Accurate presurgical brain mapping enables preoperative risk assessment and intraoperative guidance. This cross-sectional study investigated whether constrained spherical deconvolution (CSD) methods were more accurate than diffusion tensor imaging (DTI)-based methods for presurgical white matter mapping using intraoperative direct electrical stimulation (DES) as the ground truth.

METHODS

Five different tractography methods were compared (three DTI-based and two CSD-based) in 22 preoperative neurosurgical patients undergoing surgery with DES mapping. The corticospinal tract (CST, N = 20) and arcuate fasciculus (AF, N = 7) bundles were reconstructed, then minimum distances between tractograms and DES coordinates were compared between tractography methods. Receiver-operating characteristic (ROC) curves were used for both bundles. For the CST, binary agreement, linear modeling, and posthoc testing were used to compare tractography methods while correcting for relative lesion and bundle volumes.

RESULTS

Distance measures between 154 positive (functional response, pDES) and negative (no response, nDES) coordinates, and 134 tractograms resulted in 860 data points. Higher agreement was found between pDES coordinates and CSD-based compared to DTI-based tractograms. ROC curves showed overall higher sensitivity at shorter distance cutoffs for CSD (8.5 mm) compared to DTI (14.5 mm). CSD-based CST tractograms showed significantly higher agreement with pDES, which was confirmed by linear modeling and posthoc tests (PFWE < .05).

CONCLUSIONS

CSD-based CST tractograms were more accurate than DTI-based ones when validated using DES-based assessment of motor and sensory function. This demonstrates the potential benefits of structural mapping using CSD in clinical practice.

The Role of Treatment-Related Parameters and Brain Morphology in the Lesion Volume of Magnetic-Resonance-Guided Focused Ultrasound Thalamotomy in Patients with Tremor-Dominant Neurological Conditions

PURPOSE

To determine the best predictor of lesion volume induced by magnetic resonance (MR)-guided focused ultrasound (MRgFUS) thalamotomy in patients with tremor-dominant symptoms in Parkinson's disease (PD) and essential tremor (ET) patients.

METHODS

Thirty-six neurological patients with medication-refractory tremor (n°19 PD; n°17 ET) were treated using a commercial MRgFUS brain system (Exablate Neuro 4000, Insightec) integrated with a 1.5 T MRI unit (Sigma HDxt; GE Medical System). Linear regression analysis was used to determine how the demographic, clinical, radiological (Fazekas scale), volumetric (total GM/WM/CSF volume, cortical thickness), and MRgFUS-related parameters [Skull Density Ratio (SDR), n° of transducer elements, n° of sonications, skull area, maximal energy delivered (watt), maximal power delivered (joule), maximal sonication time delivered, maximal mean temperature reached (T°C_max), accumulated thermal dose (ATD)] impact on ventral intermediate (VIM)-thalamotomy-related 3D volumetric lesions of necrosis and edema.

RESULTS

The VIM thalamotomy was clinically efficacious in improving the tremor symptoms of all the patients as measured at 1 week after treatment. Multiple regression analysis revealed that T°C_max and n° of transducer elements were the best predictors of the necrosis and edema volumes. Moreover, total WM volume also predicted the size of necrosis.

CONCLUSIONS

Our study provides new insights into the clinical MRgFUS procedures that can be used to forecast brain lesion size and improve treatment outcomes.

Repeated blood-brain barrier opening with a nine-emitter implantable ultrasound device in combination with carboplatin in recurrent glioblastoma: a phase I/II clinical trial

Here, the results of a phase 1/2 single-arm trial (NCT03744026) assessing the safety and efficacy of blood-brain barrier (BBB) disruption with an implantable ultrasound system in recurrent glioblastoma patients receiving carboplatin are reported. A nine-emitter ultrasound implant was placed at the end of tumor resection replacing the bone flap. After surgery, activation to disrupt the BBB was performed every four weeks either before or after carboplatin infusion. The primary objective of the Phase 1 was to evaluate the safety of escalating numbers of ultrasound emitters using a standard 3 + 3 dose escalation. The primary objective of the Phase 2 was to evaluate the efficacy of BBB opening using magnetic resonance imaging (MRI). The secondary objectives included safety and clinical efficacy. Thirty-three patients received a total of 90 monthly sonications with carboplatin administration and up to nine emitters activated without observed DLT. Grade 3 procedure-related adverse events consisted of pre syncope (n = 3), fatigue (n = 1), wound infection (n = 2), and pain at time of device connection (n = 7). BBB opening endpoint was met with 90% of emitters showing BBB disruption on MRI after sonication. In the 12 patients who received carboplatin just prior to sonication, the progression-free survival was 3.1 months, the 1-year overall survival rate was 58% and median overall survival was 14.0 months from surgery.

Magnetic Resonance Imaging-Guided Frameless Stereotactic Injections of the Bilateral Cerebellar Dentate Nuclei in Nonhuman Primates: Technical Note

BACKGROUND AND OBJECTIVES

Nonhuman primates (NHPs) are important preclinical models for evaluating therapeutics because of their anatomophysiological similarities to humans, and can be especially useful for testing new delivery targets. With the growing promise of cell and gene therapies for the treatment of neurological diseases, it is important to ensure the accurate and safe delivery of these agents to target structures in the brain. However, a standard guideline or method has not been developed for stereotactic targeting in NHPs. In this article, we describe the safe use of a magnetic resonance imaging-guided frameless stereotactic system to target bilateral cerebellar dentate nuclei for accurate, real-time delivery of viral vector in NHPs.

METHODS

Seventeen rhesus macaques (Macaca mulatta) underwent stereotactic surgery under real-time MRI guidance using the ClearPoint® system. Bilateral cerebellar dentate nuclei were targeted through a single parietal entry point with a transtentorial approach. Fifty microliters of contrast-impregnated infusate was delivered to each dentate nucleus, and adjustments were made as necessary according to real-time MRI monitoring of delivery. Perioperative clinical outcomes and postoperative volumes of distribution were recorded.

RESULTS

All macaques underwent bilateral surgery successfully. Superficial pin site infection occurred in 4/17 (23.5%) subjects, which resolved with antibiotics. Two episodes of transient neurological deficit (anisocoria and unilateral weakness) were recorded, which did not require additional postoperative treatment and resolved over time. Volume of distribution of infusate achieved satisfactory coverage of target dentate nuclei, and only 1 incidence (2.9%) of cerebrospinal fluid penetration was recorded. Mean volume of distribution was 161.22 ± 39.61 mm3 (left, 173.65 ± 48.29; right, 148.80 ± 23.98).

CONCLUSION

MRI-guided frameless stereotactic injection of bilateral cerebellar dentate nuclei in NHPs is safe and feasible. The use of this technique enables real-time modification of the surgical plan to achieve adequate target coverage and can be readily translated to clinical use.

Review of the Free Research Software for Computer-Assisted Interventions

Research software is continuously developed to facilitate progress and innovation in the medical field. Over time, numerous research software programs have been created, making it challenging to keep abreast of what is available. This work aims to evaluate the most frequently utilized software by the computer-assisted intervention (CAI) research community. The software assessments encompass a range of criteria, including load time, stress load, multi-tasking, extensibility and range of functionalities, user-friendliness, documentation, and technical support. A total of eight software programs were selected: 3D Slicer, Elastix, ITK-SNAP, MedInria, MeVisLab, MIPAV, and Seg3D. While none of the software was found to be perfect on all evaluation criteria, 3D Slicer and ITK-SNAP emerged with the highest rankings overall. These two software programs could frequently complement each other, as 3D Slicer has a broad and customizable range of features, while ITK-SNAP excels at performing fundamental tasks in an efficient manner. Nonetheless, each software had distinctive features that may better fit the requirements of certain research projects. This review provides valuable information to CAI researchers seeking the best-suited software to support their projects. The evaluation also offers insights for the software development teams, as it highlights areas where the software can be improved.

Prediction of postoperative liver metastasis in pancreatic ductal adenocarcinoma based on multiparametric magnetic resonance radiomics combined with serological markers: a cohort study of machine learning

OBJECTIVE

To construct and validate a multi-dimensional model based on multiple machine leaning algorithms to predict PCLM using multi-parameter magnetic resonance (MRI) sequences with clinical and imaging parameters.

METHODS

A total of 148 PDAC retrospectively examined patients were classified as metastatic or non-metastatic based on results at 3 months after surgery. The radiomics features of the primary tumor were extracted from T2WI images, followed by dimension reduction. Then, multiple machine learning methods were used to construct models. Independent predictors were also screened using multifactor logistic regression and a nomogram was constructed in combination with the radiomics model. Area under the receiver operating characteristic curve (AUC) and decision curve analysis (DCA) were used to assess the accuracy and reliability of the nomogram.

RESULTS

The diagnostic efficacy of the radiomics model in the training and test set was 0.822 and 0.803, sensitivity was 0.742 and 0.692, and specificity was 0.792 and 0.875, respectively. The diagnostic efficacy of the nomogram in the training and test set was 0.866 and 0.832.

CONCLUSION

A radiomics nomogram based on machine learning improved the accuracy of predicting PCLM and may be useful for early preoperative diagnosis.

Arachidonic and docosahexaenoic acid supplementation and brain maturation in preterm infants; a double blind RCT

BACKGROUND

Arachidonic acid (ARA) and docosahexaenoic acid (DHA) are important structural components of neural cellular membranes and possess anti-inflammatory properties. Very preterm infants are deprived of the enhanced placental supply of these fatty acids, but the benefit of postnatal supplementation on brain development is uncertain. The aim of this study was to test the hypothesis that early enteral supplementation with ARA and DHA in preterm infants improves white matter (WM) microstructure assessed by diffusion-weighted MRI at term equivalent age.

METHODS

In this double-blind, randomized controlled trial, infants born before 29 weeks gestational age were allocated to either 100 mg/kg ARA and 50 mg/kg DHA (ARA:DHA group) or medium chain triglycerides (control). Supplements were started on the second day of life and provided until 36 weeks postmenstrual age. The primary outcome was brain maturation assessed by diffusion tensor imaging (DTI) using Tract-Based Spatial Statistics (TBSS) analysis.

RESULTS

We included 120 infants (60 per group) in the trial; mean (range) gestational age was 26+3 (22+6 - 28+6) weeks and postmenstrual age at scan was 41+3 (39+1 - 47+0) weeks. Ninety-two infants underwent MRI imaging, and of these, 90 had successful T1/T2 weighted MR images and 74 had DTI data of acceptable quality. TBSS did not show significant differences in mean or axial diffusivity between the groups, but demonstrated significantly higher fractional anisotropy in several large WM tracts in the ARA:DHA group, including corpus callosum, the anterior and posterior limb of the internal capsula, inferior occipitofrontal fasciculus, uncinate fasciculus, and the inferior longitudinal fasciculus. Radial diffusivity was also significantly lower in several of the same WM tracts in the ARA:DHA group.

CONCLUSION

This study suggests that supplementation with ARA and DHA at doses matching estimated fetal accretion rates improves WM maturation compared to control treatment, but further studies are needed to ascertain any functional benefit.

CLINICAL TRIAL REGISTRATION

www.

CLINICALTRIALS

gov; ID:NCT03555019.

Percutaneous microwave ablation of hepatic tumors: is there an impact of cirrhotic liver parenchyma upon the volume and short-term assessment of the ablation zone?

OBJECTIVE

To retrospectively compare and evaluate ablation zone volume and its reduction from baseline to 1 month follow-up post-percutaneous microwave ablation (MWA) between healthy and cirrhotic liver parenchyma.

METHODS

Institutional database research identified 84 patients (118 hepatic tumors) who underwent percutaneous MWA with the same system. Caudal-right lobe ratio was applied to distinguish cirrhotic (n = 51) and healthy (n = 67) group; ITK-SNAP software was used to quantify ablation zone volume. Long (LAD) and short 1 (SAD-1) and 2 (SAD-2) axis, tumor size diameter (mm) and volume (cm³) of the ablation zones were evaluated for each treated lesion in both groups at baseline (immediately post-ablation) and at 1 month follow-up.

RESULTS

There was no significant difference comparing ablation zone volumes at baseline (healthy group: mean ablation volume 14.84 cm³ vs cirrhotic group: mean ablation volume 17.85 cm³, p = 0.31) and 1 month post-ablation (healthy group: mean ablation volume 9.15 cm³ vs cirrhotic group: mean ablation volume 11.58 cm³, p = 0.24). When both "healthy" and "cirrhotic" liver group were evaluated independently, there was a significant difference of ablation volumes reduction (p-value < 0.001) from baseline to 1 month follow-up. When both groups were compared based on reduction (35.12-38.34%) there was no significant difference in ablation zone volumes (p-value = 0.77).

CONCLUSION

Percutaneous MWA results in ablation zones of a comparable volume in both healthy and cirrhotic liver parenchyma. Both cirrhotic and healthy liver parenchyma experience a similar significant reduction of ablation zone volume at 1 month post-therapy.

ADVANCES IN KNOWLEDGE STATEMENT

This study evaluates and compares the volume of the ablation zone after MWA between healthy and cirrhotic liver parenchyma from baseline to 1 month follow-up and attempts to identify potential differences. It is the first study to demonstrate significant shrinkage of ablation volumes in healthy livers as compared to cirrhotic livers after 4 weeks of follow-up. The results of this study can help us understand the effect of MWA when applied in different backgrounds of liver parenchyma, which could lead to different treatment planning.

SlicerSALT: From Medical Images to Quantitative Insights of Anatomy

Three-dimensional (3D) shape lies at the core of understanding the physical objects that surround us. In the biomedical field, shape analysis has been shown to be powerful in quantifying how anatomy changes with time and disease. The Shape AnaLysis Toolbox (SALT) was created as a vehicle for disseminating advanced shape methodology as an open source, free, and comprehensive software tool. We present new developments in our shape analysis software package, including easy-to-interpret statistical methods to better leverage the quantitative information contained in SALT's shape representations. We also show SlicerPipelines, a module to improve the usability of SALT by facilitating the analysis of large-scale data sets, automating workflows for non-expert users, and allowing the distribution of reproducible workflows.

Comparative Evaluation of a Dynamic Navigation System versus a Three-dimensional Microscope in Retrieving Separated Endodontic Files: An In Vitro Study

INTRODUCTION

This study aimed to compare the effectiveness of a dynamic navigation system and a three-dimensional microscope in retrieving broken rotary Nickel-Titanium files when using trepan burs and the extractor system.

MATERIALS AND METHODS

Thirty maxillary first bicuspids with 60 separate roots were split into 2 comparable groups based on a comprehensive cone beam computed tomography analysis of the root length and curvature. After standardized access opening, glide paths, and patency attainment with the K file (sizes 10 and 15), the teeth were arranged on 3D models (three per quadrant, six per model). Subsequently, controlled-memory heat-treated Nickel-Titanium rotary files (#25/0.04) were notched 4 mm from the tips and fractured at the apical third of the roots. The C-FR1 Endo file removal system was employed under both guidance to retrieve the fragments, and the success rate, canal aberration, treatment time, and volumetric changes were measured. The statistical analysis was performed using IBM SPSS software at a significance level of 0.05.

RESULTS

The microscope-guided group had a higher success rate than the dynamic navigation system guidance, but the difference was insignificant (P > .05). In addition, the microscope-guided drills resulted in a substantially lower proportion of canal aberration, shorter time to retrieve the fragments and less change in the root canal volume (P < .05).

CONCLUSION

Although dynamically guided trephining with the extractor can retrieve separated instruments, it is inferior to three-dimensional microscope guidance regarding treatment time, procedural errors, and volume change.

Intra- and peritumoral radiomics for predicting malignant BiRADS category 4 breast lesions on contrast-enhanced spectral mammography: a multicenter study

OBJECTIVE

To construct and test a nomogram based on intra- and peritumoral radiomics and clinical factors for predicting malignant BiRADS 4 lesions on contrast-enhanced spectral mammography.

METHODS

A total of 884 patients with BiRADS 4 lesions were enrolled from two centers. For each lesion, five ROIs were defined using the intratumoral region (ITR), peritumoral regions (PTRs) of 5 and 10 mm around the tumor, and ITR plus PTRs of 5 mm and 10 mm. Five radiomics signatures were established by LASSO after selecting features. A nomogram was built using selected signatures and clinical factors by multivariable logistic regression analysis. The performance of the nomogram was assessed with the AUC, decision curve analysis, and calibration curves, and also compared with the radiomics model, clinical model, and radiologists.

RESULTS

The nomogram built by three radiomics signatures (constructed from ITR, 5 mm PTR, and ITR + 10 mm PTR) and two clinical factors (age and BiRADS category) showed powerful predictive ability in internal and external test sets with AUCs of 0.907 and 0.904, respectively. The calibration curves, decision curve analysis, showed favorable predictive performance of the nomogram. In addition, radiologists improved the diagnostic performance with the help of nomogram.

CONCLUSION

The nomogram established via intratumoral and peritumoral radiomics features and clinical risk factors had the best performance in distinguishing benign and malignant BiRADS 4 lesions, which could help radiologists improve diagnostic capabilities.

KEY POINTS

• Radiomics features from peritumoral regions in contrast-enhanced spectral mammography images may provide valuable information for the diagnosis of benign and malignant breast imaging reporting and data system category 4 breast lesions. • The nomogram incorporated intra- and peritumoral radiomics features and clinical variables have good application prospects in assisting clinical decision-makers.

Associations of phosphorylated tau pathology with whole-hemisphere ex vivo morphometry in 7 tesla MRI

INTRODUCTION

Neurodegenerative disorders are associated with different pathologies that often co-occur but cannot be measured specifically with in vivo methods.

METHODS

Thirty-three brain hemispheres from donors with an Alzheimer's disease (AD) spectrum diagnosis underwent T2-weighted magnetic resonance imaging (MRI). Gray matter thickness was paired with histopathology from the closest anatomic region in the contralateral hemisphere.

RESULTS

Partial Spearman correlation of phosphorylated tau and cortical thickness with TAR DNA-binding protein 43 (TDP-43) and α-synuclein scores, age, sex, and postmortem interval as covariates showed significant relationships in entorhinal and primary visual cortices, temporal pole, and insular and posterior cingulate gyri. Linear models including Braak stages, TDP-43 and α-synuclein scores, age, sex, and postmortem interval showed significant correlation between Braak stage and thickness in the parahippocampal gyrus, entorhinal cortex, and Broadman area 35.

CONCLUSION

We demonstrated an association of measures of AD pathology with tissue loss in several AD regions despite a limited range of pathology in these cases.

HIGHLIGHTS

Neurodegenerative disorders are associated with co-occurring pathologies that cannot be measured specifically with in vivo methods. Identification of the topographic patterns of these pathologies in structural magnetic resonance imaging (MRI) may provide probabilistic biomarkers. We demonstrated the correlation of the specific patterns of tissue loss from ex vivo brain MRI with underlying pathologies detected in postmortem brain hemispheres in patients with Alzheimer's disease (AD) spectrum disorders. The results provide insight into the interpretation of in vivo structural MRI studies in patients with AD spectrum disorders.

Neurodesk: An accessible, flexible, and portable data analysis environment for reproducible neuroimaging

Neuroimaging data analysis often requires purpose-built software, which can be challenging to install and may produce different results across computing environments. Beyond being a roadblock to neuroscientists, these issues of accessibility and portability can hamper the reproducibility of neuroimaging data analysis pipelines. Here, we introduce the Neurodesk platform, which harnesses software containers to support a comprehensive and growing suite of neuroimaging software (https://www.neurodesk.org/). Neurodesk includes a browser-accessible virtual desktop environment and a command line interface, mediating access to containerized neuroimaging software libraries on various computing platforms, including personal and high-performance computers, cloud computing and Jupyter Notebooks. This community-oriented, open-source platform enables a paradigm shift for neuroimaging data analysis, allowing for accessible, flexible, fully reproducible, and portable data analysis pipelines.

Investigating the hemodynamic implications of triangular cross-sectioned superior sagittal sinus vessels and the errors associated with idealised modelling

The superior sagittal sinus (SSS) is a blood vessel that is often observed to be approximately triangular in cross-section, due to how the venous wall attaches to the surrounding tissue. Despite this, the vessel has been assumed to be circular, when models are generated without patient-specific data. In this study, the differences between the cerebral hemodynamics of one circular, three triangular and five patient-specific cross-sectional models of a SSS were conducted. The errors associated with using circular cross-sectioned flow extensions were also determined. Computational fluid dynamics (CFD) models were generated from these geometries, with a population mean transient blood flow profile incorporated. The maximal helicity of the fluid flow was found to be elevated in the triangular cross-section, compared to the circular, with a higher wall shear stress (WSS) observed over a smaller, more concentrated region on the posterior sinus wall. The errors associated with using a circular cross-section were detailed, with the cross-sectional area appearing to have a greater influence on the hemodynamic parameters than the triangularity or circularity of the cross-section. This highlighted the importance of exhibiting caution when incorporating idealised modelling, especially when commenting on the true hemodynamics of these models. Errors were also found to be induced when using a circular cross-sectioned flow extension, for a geometry which was non-circular. This study highlights the importance of understanding the human anatomy when modelling blood vessels.

MRI detection of brain gadolinium retention in multiple sclerosis: Magnetization transfer vs. T1-weighted imaging

BACKGROUND AND PURPOSE

Evidence of brain gadolinium retention has affected gadolinium-based contrast agent usage. It is, however, unclear to what extent macrocyclic agents are retained and whether their in vivo detection may necessitate nonconventional MRI. Magnetization transfer (MT) could prove suitable to detect gadolinium-related signal changes since dechelated gadolinium ions bind to macromolecules. Therefore, this study aimed to investigate associations of prior gadolinium administrations with MT and T1 signal abnormalities.

METHODS

A cohort of 23 persons with multiple sclerosis (MS) (18 females, 5 males, 57 ± 8.0 years) with multiple past gadolinium administrations (median 6, range 3-12) and 23 age- and sex-matched healthy controls underwent 1.5 Tesla MRI with MT, T1-weighted 2-dimensional spin echo, and T1-weighted 3-dimensional gradient echo. The signal intensity index was assessed by MRI in gadolinium retention predilection sites.

RESULTS

There were dose-dependent associations of the globus pallidus signal on gradient echo (r = .55, p < .001) and spin echo (r = .38, p = .013) T1-weighted imaging, but not on MT. Relative to controls, MS patients had higher signal intensity index in the dentate nucleus on T1-weighted gradient echo (1.037 ± 0.040 vs. 1.016 ± 0.023, p = .04) with a similar trend in the globus pallidus on T1-weighted spin echo (1.091 ± 0.034 vs. 1.076 ± 0.014, p = .06). MT detected no group differences.

CONCLUSIONS

Conventional T1-weighted imaging provided dose-dependent associations with gadolinium administrations in MS, while these could not be detected with 2-dimensional MT. Future studies could explore newer MT techniques like 3D and inhomogenous MT. Notably, these associations were identified with conventional MRI even though most patients had not received gadolinium administrations in the preceding 9 years, suggestive of long-term retention.

Exploring structural connectomes in children with unilateral cerebral palsy using graph theory

We explored structural brain connectomes in children with spastic unilateral cerebral palsy (uCP) and its relation to sensory-motor function using graph theory. In 46 children with uCP (mean age = 10 years 7 months ± 2 years 9 months; Manual Ability Classification System I = 15, II = 16, III = 15) we assessed upper limb somatosensory and motor function. We collected multi-shell diffusion-weighted, T1-weighted and T2-FLAIR MRI and identified the corticospinal tract (CST) wiring pattern using transcranial magnetic stimulation. Structural connectomes were constructed using Virtual Brain Grafting-modified FreeSurfer parcellations and multi-shell multi-tissue constrained spherical deconvolution-based anatomically-constrained tractography. Graph metrics (characteristic path length, global/local efficiency and clustering coefficient) of the whole brain, the ipsilesional/contralesional hemisphere, and the full/ipsilesional/contralesional sensory-motor network were compared between lesion types (periventricular white matter (PWM) = 28, cortical and deep gray matter (CDGM) = 18) and CST-wiring patterns (ipsilateral = 14, bilateral = 14, contralateral = 12, unknown = 6) using ANCOVA with age as covariate. Using elastic-net regularized regression we investigated how graph metrics, lesion volume, lesion type, CST-wiring pattern and age predicted sensory-motor function. In both the whole brain and subnetworks, we observed a hyperconnectivity pattern in children with CDGM-lesions compared with PWM-lesions, with higher clustering coefficient (p = [<.001-.047], η p 2 $$ {\eta}_p^2 $$ =[0.09-0.27]), characteristic path length (p = .003, η p 2 $$ {\eta}_p^2 $$ =0.19) and local efficiency (p = [.001-.02], η p 2 $$ {\eta}_p^2 $$ =[0.11-0.21]), and a lower global efficiency with age (p = [.01-.04], η p 2 $$ {\eta}_p^2 $$ =[0.09-0.15]). No differences were found between CST-wiring groups. Overall, good predictions of sensory-motor function were obtained with elastic-net regression (R²  = .40-.87). CST-wiring pattern was the strongest predictor for motor function. For somatosensory function, all independent variables contributed equally to the model. In conclusion, we demonstrated the potential of structural connectomics in understanding disease severity and brain development in children with uCP.

High-resolution imaging in studies of alcohol effect on prenatal development

Fetal alcohol syndrome represents the leading known preventable cause of mental retardation. FAS is on the most severe side of fetal alcohol spectrum disorders that stem from the deleterious effects of prenatal alcohol exposure. Affecting as many as 1 to 5 out of 100 children, FASD most often results in brain abnormalities that extend to structure, function, and cerebral hemodynamics. The present review provides an analysis of high-resolution imaging techniques that are used in animals and human subjects to characterize PAE-driven changes in the developing brain. Variants of magnetic resonance imaging such as magnetic resonance microscopy, magnetic resonance spectroscopy, diffusion tensor imaging, along with positron emission tomography, single-photon emission computed tomography, and photoacoustic imaging, are modalities that are used to study the influence of PAE on brain structure and function. This review briefly describes the aforementioned imaging modalities, the main findings that were obtained using each modality, and touches upon the advantages/disadvantages of each imaging approach.

Telomerase reverse transcriptase promoter mutation and histologic grade in IDH wild-type histological lower-grade gliomas: The value of perfusion-weighted image, diffusion-weighted image, and 18F-FDG-PET

PURPOSE

The telomerase reverse transcriptase promoter (TERTp) mutation is an unfavorable prognostic factor in isocitrate dehydrogenase-wildtype (IDHwt) histologically lower-grade astrocytoma (LGA), which was incorporated as a key component in the WHO 2021 classification of IDHwt LGA, replacing histologic grades in the WHO 2016 classification. The purpose of this study was to identify the imaging characteristics predictive of TERTp mutations in IDHwt LGA.

METHODS

This retrospective study was approved by our institutional review board. This single-center study retrospectively included 59 patients with pathologically confirmed IDHwt LGA with known TERTp mutation status. In addition to clinical information and morphological characteristics, semi-quantitative imaging biomarkers such as the tumor-to-normal ratio (T/N ratio) on ¹⁸F-FDG-PET, normalized apparent diffusion coefficient (nADC), and histogram parameters from normalized relative cerebral blood volume (nrCBV) maps were compared between (a) TERTp-wildtype and TERTp-mutant tumors or (b) grade II and grade III astrocytoma. A p value < 0.05 was considered significant.

RESULTS

There were no significant differences in the conventional imaging findings, T/N ratio on FDG-PET, nrCBV or ADC histogram metrics between IDHwt LGA with TERTp mutations and those without. Grade III IDHwt astrocytomas exhibited significantly higher nrCBV values, T/N ratio and lower ADC parameters than grade II IDHwt astrocytoma.

CONCLUSIONS

In patients with IDHwt LGA, T/N ratio, nrCBV values and nADC may be surrogate markers for predicting histologic grade, but are not useful for predicting TERTp mutations.

Radiomics-Based Machine Learning to Predict Recurrence in Glioma Patients Using Magnetic Resonance Imaging

OBJECTIVE

Recurrence is a major factor in the poor prognosis of patients with glioma. The aim of this study was to predict glioma recurrence using machine learning based on radiomic features.

METHODS

We recruited 77 glioma patients, consisting of 57 newly diagnosed patients and 20 patients with recurrence. After extracting the radiomic features from T2-weighted images, the data set was randomly divided into training (58 patients) and testing (19 patients) cohorts. An automated machine learning method (the Tree-based Pipeline Optimization Tool) was applied to generate 10 independent recurrence prediction models. The final model was determined based on the area under the curve (AUC) and average specificity. Moreover, an independent validation set of 20 patients with glioma was used to verify the model performance.

RESULTS

Recurrence in glioma patients was successfully predicting by machine learning using radiomic features. Among the 10 recurrence prediction models, the best model achieved an accuracy of 0.81, an AUC value of 0.85, and a specificity of 0.69 in the testing cohort, but an accuracy of 0.75 and an AUC value of 0.87 in the independent validation set.

CONCLUSIONS

Our algorithm that is generated by machine learning exhibits promising power and may predict recurrence noninvasively, thereby offering potential value for the early development of interventions to delay or prevent recurrence in glioma patients.

Efficient contour-based annotation by iterative deep learning for organ segmentation from volumetric medical images

PURPOSE

Training deep neural networks usually require a large number of human-annotated data. For organ segmentation from volumetric medical images, human annotation is tedious and inefficient. To save human labour and to accelerate the training process, the strategy of annotation by iterative deep learning recently becomes popular in the research community. However, due to the lack of domain knowledge or efficient human-interaction tools, the current AID methods still suffer from long training time and high annotation burden.

METHODS

We develop a contour-based annotation by iterative deep learning (AID) algorithm which uses boundary representation instead of voxel labels to incorporate high-level organ shape knowledge. We propose a contour segmentation network with a multi-scale feature extraction backbone to improve the boundary detection accuracy. We also developed a contour-based human-intervention method to facilitate easy adjustments of organ boundaries. By combining the contour-based segmentation network and the contour-adjustment intervention method, our algorithm achieves fast few-shot learning and efficient human proofreading.

RESULTS

For validation, two human operators independently annotated four abdominal organs in computed tomography (CT) images using our method and two compared methods, i.e. a traditional contour-interpolation method and a state-of-the-art (SOTA) convolutional network (CNN) method based on voxel label representation. Compared to these methods, our approach considerably saved annotation time and reduced inter-rater variabilities. Our contour detection network also outperforms the SOTA nnU-Net in producing anatomically plausible organ shape with only a small training set.

CONCLUSION

Taking advantage of the boundary shape prior and the contour representation, our method is more efficient, more accurate and less prone to inter-operator variability than the SOTA AID methods for organ segmentation from volumetric medical images. The good shape learning ability and flexible boundary adjustment function make it suitable for fast annotation of organ structures with regular shape.

CrossMoDA 2021 challenge: Benchmark of cross-modality domain adaptation techniques for vestibular schwannoma and cochlea segmentation

Domain Adaptation (DA) has recently been of strong interest in the medical imaging community. While a large variety of DA techniques have been proposed for image segmentation, most of these techniques have been validated either on private datasets or on small publicly available datasets. Moreover, these datasets mostly addressed single-class problems. To tackle these limitations, the Cross-Modality Domain Adaptation (crossMoDA) challenge was organised in conjunction with the 24th International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI 2021). CrossMoDA is the first large and multi-class benchmark for unsupervised cross-modality Domain Adaptation. The goal of the challenge is to segment two key brain structures involved in the follow-up and treatment planning of vestibular schwannoma (VS): the VS and the cochleas. Currently, the diagnosis and surveillance in patients with VS are commonly performed using contrast-enhanced T1 (ceT₁) MR imaging. However, there is growing interest in using non-contrast imaging sequences such as high-resolution T2 (hrT₂) imaging. For this reason, we established an unsupervised cross-modality segmentation benchmark. The training dataset provides annotated ceT₁ scans (N=105) and unpaired non-annotated hrT₂ scans (N=105). The aim was to automatically perform unilateral VS and bilateral cochlea segmentation on hrT₂ scans as provided in the testing set (N=137). This problem is particularly challenging given the large intensity distribution gap across the modalities and the small volume of the structures. A total of 55 teams from 16 countries submitted predictions to the validation leaderboard. Among them, 16 teams from 9 different countries submitted their algorithm for the evaluation phase. The level of performance reached by the top-performing teams is strikingly high (best median Dice score - VS: 88.4%; Cochleas: 85.7%) and close to full supervision (median Dice score - VS: 92.5%; Cochleas: 87.7%). All top-performing methods made use of an image-to-image translation approach to transform the source-domain images into pseudo-target-domain images. A segmentation network was then trained using these generated images and the manual annotations provided for the source image.

Validation and comparison of volume measurements using 1 multidetector computed tomography and 5 cone-beam computed tomography protocols: An in vitro study

Purpose

The purpose of this study was to compare volume measurements obtained using 2 image software packages on Digital Imaging and Communications in Medicine (DICOM) images acquired from 1 multidetector computed tomography and 5 cone-beam computed tomography devices, using different protocols for physical volume measurements.

Materials and Methods

Four pieces of bovine leg were prepared. Marrow was removed from 3 pieces, leaving cortical bone exposed. The resulting space of 1 piece was filled with water, another was filled with propylene glycol, and the third was left unfilled. The marrow in the fourth sample was left fully intact. Volume measurements were obtained after importing DICOM images into the Dolphin Imaging 11.95 and ITK-SNAP software programs. Data were analyzed using 3-way analysis of variance with a generalized linear model to determine the effects of voxel size, software, and content on percentage mean volume differences between tomographic protocols. A significance level of 0.05 was used.

Results

The intraclass correlation coefficients for intraobserver and interobserver reliability were, respectively, 0.915 and 0.764 for the Dolphin software and 0.894 and 0.766 for the ITK-SNAP software. Three sources of statistically significant variation were identified: the interaction between software and content (P=0.001), the main effect of content (P=0.014), and the main effect of software (P=0.001). Voxel size was not associated with statistically significant differences in volume measurements.

Conclusion

Both content and software influenced the accuracy of volume measurements, especially when the content had gray values similar to those of the adjacent tissues.

Cholesteatoma has an altered microbiota with a higher abundance of Staphylococcus species

Objective

To compare the microbiota between cholesteatoma and chronic suppurative otitis media (COM) and to identify potential pathogens that explain the relevant phenotypes of cholesteatoma.

Study Design

Prospective cohort study.

Methods

Surgical specimens collected from 20 cholesteatomas and nine COMs were treated to dissolve biofilms and subjected to 16S ribosomal RNA (rRNA) gene sequencing and amplicon sequence variant-level analysis for microbiota profiling and quantitative comparison. Correlations between the relative abundance of potential pathogens and the volume of the primary resected cholesteatomas were examined.

Results

Differences in bacterial composition (beta diversity) were observed between cholesteatomas and COM (p = .002), with a higher abundance of Staphylococcus in cholesteatomas than in COM (p = .005). Common genera in the external auditory canal (EAC) flora, such as Staphylococcus, Corynebacterium, and Cutibacterium, were predominant in both cholesteatoma and COM; Staphylococcus aureus and Pseudomonas aeruginosa were increased in both diseases compared with the EAC flora. Furthermore, coagulase-negative staphylococci (CoNS) were more abundant in cholesteatomas than in COM (p = 0.002). Linear discriminant analysis coupled with effect size measurements (LEfSe) identified four CoNS as potential biomarkers for cholesteatoma. The relative abundance of S. aureus, a potential pathogen, was positively correlated with cholesteatoma volume (r = .60, p = .02).

Conclusion

The microbiota of cholesteatoma and COM originated from EAC flora, but the bacterial composition was largely altered. Our results suggested that S. aureus infection is involved in cholesteatoma progression.

Level of Evidence

3b.

Automated multimodal segmentation of acute ischemic stroke lesions on clinical MR images

Magnetic resonance (MR) imaging (MRI) is commonly used to diagnose, assess and monitor stroke. Accurate and timely segmentation of stroke lesions provides the anatomico-structural information that can aid physicians in predicting prognosis, as well as in decision making and triaging for various rehabilitation strategies. To segment stroke lesions, MR protocols, including diffusion-weighted imaging (DWI) and T2-weighted fluid attenuated inversion recovery (FLAIR) are often utilized. These imaging sequences are usually acquired with different spatial resolutions due to time constraints. Within the same image, voxels may be anisotropic, with reduced resolution along slice direction for diffusion scans in particular. In this study, we evaluate the ability of 2D and 3D U-Net Convolutional Neural Network (CNN) architectures to segment ischemic stroke lesions using single contrast (DWI) and dual contrast images (T2w FLAIR and DWI). The predicted segmentations correlate with post-stroke motor outcome measured by the National Institutes of Health Stroke Scale (NIHSS) and Fugl-Meyer Upper Extremity (FM-UE) index based on the lesion loads overlapping the corticospinal tracts (CST), which is a neural substrate for motor movement and function. Although the four methods performed similarly, the 2D multimodal U-Net achieved the best results with a mean Dice of 0.737 (95% CI: 0.705, 0.769) and a relatively high correlation between the weighted lesion load and the NIHSS scores (both at baseline and at 90 days). A monotonically constrained quintic polynomial regression yielded R2 = 0.784 and 0.875 for weighted lesion load versus baseline and 90-Days NIHSS respectively, and better corrected Akaike information criterion (AICc) scores than those of the linear regression. In addition, using the quintic polynomial regression model to regress the weighted lesion load to the 90-Days FM-UE score results in an R2 of 0.570 with a better AICc score than that of the linear regression. Our results suggest that the multi-contrast information enhanced the accuracy of the segmentation and the prediction accuracy for upper extremity motor outcomes. Expanding the training dataset to include different types of stroke lesions and more data points will help add a temporal longitudinal aspect and increase the accuracy. Furthermore, adding patient-specific data may improve the inference about the relationship between imaging metrics and functional outcomes.

The University of Pennsylvania glioblastoma (UPenn-GBM) cohort: advanced MRI, clinical, genomics, & radiomics

Glioblastoma is the most common aggressive adult brain tumor. Numerous studies have reported results from either private institutional data or publicly available datasets. However, current public datasets are limited in terms of: a) number of subjects, b) lack of consistent acquisition protocol, c) data quality, or d) accompanying clinical, demographic, and molecular information. Toward alleviating these limitations, we contribute the "University of Pennsylvania Glioblastoma Imaging, Genomics, and Radiomics" (UPenn-GBM) dataset, which describes the currently largest publicly available comprehensive collection of 630 patients diagnosed with de novo glioblastoma. The UPenn-GBM dataset includes (a) advanced multi-parametric magnetic resonance imaging scans acquired during routine clinical practice, at the University of Pennsylvania Health System, (b) accompanying clinical, demographic, and molecular information, (d) perfusion and diffusion derivative volumes, (e) computationally-derived and manually-revised expert annotations of tumor sub-regions, as well as (f) quantitative imaging (also known as radiomic) features corresponding to each of these regions. This collection describes our contribution towards repeatable, reproducible, and comparative quantitative studies leading to new predictive, prognostic, and diagnostic assessments.

DBB - A Distorted Brain Benchmark for Automatic Tissue Segmentation in Paediatric Patients

T1-weighted magnetic resonance images provide a comprehensive view of the morphology of the human brain at the macro scale. These images are usually the input of a segmentation process that aims detecting the anatomical structures labeling them according to a predefined set of target tissues. Automated methods for brain tissue segmentation rely on anatomical priors of the human brain structures. This is the reason why their performance is quite accurate on healthy individuals. Nevertheless model-based tools become less accurate in clinical practice, specifically in the cases of severe lesions or highly distorted cerebral anatomy. More recently there are empirical evidences that a data-driven approach can be more robust in presence of alterations of brain structures, even though the learning model is trained on healthy brains. Our contribution is a benchmark to support an open investigation on how the tissue segmentation of distorted brains can be improved by adopting a supervised learning approach. We formulate a precise definition of the task and propose an evaluation metric for a fair and quantitative comparison. The training sample is composed of almost one thousand healthy individuals. Data include both T1-weighted MR images and their labeling of brain tissues. The test sample is a collection of several tens of individuals with severe brain distortions. Data and code are openly published on BrainLife, an open science platform for reproducible neuroscience data analysis.

Brain Tumor Segmentation Using Deep Capsule Network and Latent-Dynamic Conditional Random Fields

Because of the large variabilities in brain tumors, automating segmentation remains a difficult task. We propose an automated method to segment brain tumors by integrating the deep capsule network (CapsNet) and the latent-dynamic condition random field (LDCRF). The method consists of three main processes to segment the brain tumor—pre-processing, segmentation, and post-processing. In pre-processing, the N4ITK process involves correcting each MR image’s bias field before normalizing the intensity. After that, image patches are used to train CapsNet during the segmentation process. Then, with the CapsNet parameters determined, we employ image slices from an axial view to learn the LDCRF-CapsNet. Finally, we use a simple thresholding method to correct the labels of some pixels and remove small 3D-connected regions from the segmentation outcomes. On the BRATS 2015 and BRATS 2021 datasets, we trained and evaluated our method and discovered that it outperforms and can compete with state-of-the-art methods in comparable conditions.

Enhancing the REMBRANDT MRI collection with expert segmentation labels and quantitative radiomic features

Malignancy of the brain and CNS is unfortunately a common diagnosis. A large subset of these lesions tends to be high grade tumors which portend poor prognoses and low survival rates, and are estimated to be the tenth leading cause of death worldwide. The complex nature of the brain tissue environment in which these lesions arise offers a rich opportunity for translational research. Magnetic Resonance Imaging (MRI) can provide a comprehensive view of the abnormal regions in the brain, therefore, its applications in the translational brain cancer research is considered essential for the diagnosis and monitoring of disease. Recent years has seen rapid growth in the field of radiogenomics, especially in cancer, and scientists have been able to successfully integrate the quantitative data extracted from medical images (also known as radiomics) with genomics to answer new and clinically relevant questions. In this paper, we took raw MRI scans from the REMBRANDT data collection from public domain, and performed volumetric segmentation to identify subregions of the brain. Radiomic features were then extracted to represent the MRIs in a quantitative yet summarized format. This resulting dataset now enables further biomedical and integrative data analysis, and is being made public via the NeuroImaging Tools & Resources Collaboratory (NITRC) repository ( https://www.nitrc.org/projects/rembrandt_brain/ ).

Scalable and accessible personalized photodynamic therapy optimization with FullMonte and PDT-SPACE

SIGNIFICANCE

Open-source software packages have been extensively used in the past three decades in medical imaging and diagnostics, aiming to study the feasibility of the application ex vivo. Unfortunately, most of the existing open-source tools require some software engineering background to install the prerequisite libraries, choose a suitable computational platform, and combine several software tools to address different applications.

AIM

To facilitate the use of open-source software in medical applications, enabling computational studies of treatment outcomes prior to the complex in-vivo setting.

APPROACH

FullMonteWeb, an open-source, user-friendly web-based software with a graphical user interface for interstitial photodynamic therapy (iPDT) modeling, visualization, and optimization, is introduced. The software can perform Monte Carlo simulations of light propagation in biological tissues, along with iPDT plan optimization. FullMonteWeb installs and runs the required software and libraries on Amazon Web Services (AWS), allowing scalable computing without complex set up.

RESULTS

FullMonteWeb allows simulation of large and small problems on the most appropriate compute hardware, enabling cost improvements of 10  ×   versus always running on a single platform. Case studies in optical property estimation and diffuser placement optimization highlight FullMonteWeb's versatility.

CONCLUSION

The FullMonte open source suite enables easier and more cost-effective in-silico studies for iPDT.

Multi-scanner and multi-modal lumbar vertebral body and intervertebral disc segmentation database

Magnetic resonance imaging (MRI) is widely utilized for diagnosing and monitoring of spinal disorders. For a number of applications, particularly those related to quantitative MRI, an essential step towards achieving reliable and objective measurements is the segmentation of the examined structures. Performed manually, such process is time-consuming and prone to errors, posing a bottleneck to its clinical applicability. A more efficient analysis would be achieved by automating a segmentation process. However, routine spine MRI acquisitions pose several challenges for achieving robust and accurate segmentations, due to varying MRI acquisition characteristics occurring in data acquired from different sites. Moreover, heterogeneous annotated datasets, collected from multiple scanners with different pulse sequence protocols, are limited. Thus, we present a manually segmented lumbar spine MRI database containing a wide range of data obtained from multiple scanners and pulse sequences, with segmentations of lumbar vertebral bodies and intervertebral discs. The database is intended for the use in developing and testing of automated lumbar spine segmentation algorithms in multi-domain scenarios.

Measurement(s)	Vertebral Body • Intervertebral Disc
Technology Type(s)	Magnetic Resonance Imaging

Deep Learning Model for Intracranial Hemangiopericytoma and Meningioma Classification

Background

Intracranial hemangiopericytoma/solitary fibrous tumor (SFT/HPC) is a rare type of neoplasm containing malignancies of infiltration, peritumoral edema, bleeding, or bone destruction. However, SFT/HPC has similar radiological characteristics as meningioma, which had different clinical managements and outcomes. This study aims to discriminate SFT/HPC and meningioma via deep learning approaches based on routine preoperative MRI.

Methods

We enrolled 236 patients with histopathological diagnosis of SFT/HPC (n = 144) and meningioma (n = 122) from 2010 to 2020 in Xiangya Hospital. Radiological features were extracted manually, and a radiological diagnostic model was applied for classification. And a deep learning pretrained model ResNet-50 was adapted to train T1-contrast images for predicting tumor class. Deep learning model attention mechanism was visualized by class activation maps.

Results

Our study reports that SFT/HPC was found to have more invasion to venous sinus (p = 0.001), more cystic components (p < 0.001), and more heterogeneous enhancement patterns (p < 0.001). Deep learning model achieved a high classification accuracy of 0.889 with receiver-operating characteristic curve area under the curve (AUC) of 0.91 in the validation set. Feature maps showed distinct clustering of SFT/HPC and meningioma in the training and test cohorts, respectively. And the attention of the deep learning model mainly focused on the tumor bulks that represented the solid texture features of both tumors for discrimination.

Analysis of Brain Structure and Neural Organization in Dystrophin-Deficient Model Mice with Magnetic Resonance Imaging at 7 T

Background:

Dystrophin strengthens muscle cells; however, in muscular dystrophy, dystrophin is deficient due to an abnormal sugar chain. This abnormality occurs in skeletal muscle and in brain tissue.

Objective:

This study aimed to non-invasively analyze the neural organization of the brain in muscular dystrophy. We used a mouse model of muscular dystrophy to study whether changes in brain structure and neurodegeneration following dystrophin deficiency can be assessed by 7T magnetic resonance imaging.

Methods:

C57BL/10-mdx (X chromosome-linked muscular dystrophy) mice were used as the dystrophic mouse model and healthy mice were used as controls. Ventricular enlargement is one of the most common brain malformations in dystrophin-deficient patients. Therefore, we examined whether ventricular enlargement was observed in C57BL/10-mdx using transverse-relaxation weighted images. Brain parenchyma analysis was performed using diffusion MRI with diffusion tensor images and neurite orientation dispersion and density imaging. Parenchymal degeneration was assessed in terms of directional diffusion, nerve fiber diffusion, and dendritic scattering density.

Results:

For the volume of brain ventricles analyzed by T2WI, the average size was 1.5 times larger in mdx mice compared to control mice. In the brain parenchyma, a significant difference (p < 0.05) was observed in parameters indicating disturbances in the direction of nerve fibers and dendritic scattering density in the white matter region.

Conclusion:

Our results show that changes in brain structure due to dystrophin deficiency can be assessed in detail without tissue destruction by combining diffusion tensor images and neurite orientation dispersion and density imaging analyses.

Improving Automated Glioma Segmentation in Routine Clinical Use Through Artificial Intelligence-Based Replacement of Missing Sequences With Synthetic Magnetic Resonance Imaging Scans

OBJECTIVES

Although automated glioma segmentation holds promise for objective assessment of tumor biology and response, its routine clinical use is impaired by missing sequences, for example, due to motion artifacts. The aim of our study was to develop and validate a generative adversarial network for synthesizing missing sequences to allow for a robust automated segmentation.

MATERIALS AND METHODS

Our model was trained on data from The Cancer Imaging Archive (n = 238 WHO II-IV gliomas) to synthesize either missing FLAIR, T2-weighted, T1-weighted (T1w), or contrast-enhanced T1w images from available sequences, using a novel tumor-targeting loss to improve synthesis of tumor areas. We validated performance in a test set from both the REMBRANDT repository and our local institution (n = 68 WHO II-IV gliomas), using qualitative image appearance metrics, but also segmentation performance with state-of-the-art segmentation models. Segmentation of synthetic images was compared with 2 commonly used strategies for handling missing input data, entering a blank mask or copying an existing sequence.

RESULTS

Across tumor areas and missing sequences, synthetic images generally outperformed both conventional approaches, in particular when FLAIR was missing. Here, for edema and whole tumor segmentation, we improved the Dice score, a common metric for evaluation of segmentation performance, by 12% and 11%, respectively, over the best conventional method. No method was able to reliably replace missing contrast-enhanced T1w images.

DISCUSSION

Replacing missing nonenhanced magnetic resonance sequences via synthetic images significantly improves segmentation quality over most conventional approaches. This model is freely available and facilitates more widespread use of automated segmentation in routine clinical use, where missing sequences are common.

Clinical Imaging-Derived Metrics of Corticospinal Tract Structural Integrity Are Associated With Post-stroke Motor Outcomes: A Retrospective Study

OBJECTIVE

The primary objective of this study was to retrospectively investigate associations between clinical magnetic resonance imaging-based (MRI) metrics of corticospinal tract (CST) status and paretic upper extremity (PUE) motor recovery in patients that completed acute inpatient rehabilitation (AR) post-stroke.

METHODS

We conducted a longitudinal chart review of patients post-stroke who received care in the Emory University Hospital system during acute hospitalization, AR, and outpatient therapy. We extracted demographic information, stroke characteristics, and longitudinal documentation of post-stroke motor function from institutional electronic medical records. Serial assessments of paretic shoulder abduction and finger extension were estimated (E-SAFE) and an estimated Action Research Arm Test (E-ARAT) score was used to quantify 3-month PUE motor function outcome. Clinically-diagnostic MRI were used to create lesion masks that were spatially normalized and overlaid onto a white matter tract atlas delineating CST contributions emanating from six cortical seed regions to obtain the percentage of CST lesion overlap. Metric associations were investigated with correlation and cluster analyses, Kruskal-Wallis tests, classification and regression tree analysis.

RESULTS

Thirty-four patients met study eligibility criteria. All CST overlap percentages were correlated with E-ARAT however, ventral premotor tract (PMv) overlap was the only tract that remained significantly correlated after multiple comparisons adjustment. Lesion overlap percentage in CST contributions from all seed regions was significantly different between outcome categories. Using MRI metrics alone, dorsal premotor (PMd) and PMv tracts classified recovery outcome category with 79.4% accuracy. When clinical and MRI metrics were combined, AR E-SAFE, patient age, and overall CST lesion overlap classified patients with 88.2% accuracy.

CONCLUSIONS

Study findings revealed clinical MRI-derived CST lesion overlap was associated with PUE motor outcome post-stroke and that cortical projections within the CST, particularly those emanating from non-M1 cortical areas, prominently ventral premotor (PMv) and dorsal premotor (PMd) cortices, distinguished between PUE outcome groups. Exploratory predictive models using clinical MRI metrics, either alone or in combination with clinical measures, were able to accurately identify recovery outcome category for the study cohort during both the acute and early subacute phases of post-stroke recovery. Prospective studies are recommended to determine the predictive utility of including clinical imaging-based biomarkers of white matter tract structural integrity in predictive models of post-stroke recovery.

Rectal MRI radiomics for predicting pathological complete response: Where we are

Radiomics using rectal MRI radiomics has emerged as a promising approach in predicting pathological complete response. In this study, we present a typical pipeline of a radiomics analysis and review recent studies, exploring applications, development of radiomics methodologies and model construction in pCR prediction. Finally, we will offer our opinion about the future and discuss the next steps of rectal MRI radiomics for predicting pCR.

The diagnostic value of contrast enhancement on MRI in diffuse and anaplastic gliomas

PURPOSE

We evaluated differentiations in gadolinium contrast enhancement (CE) between low-grade WHO °II and high-grade WHO °III gliomas in conventional MRI, which have been repeatedly questioned.

METHODS

Ninety-nine patients, who underwent first resection of WHO°II and °III gliomas, were retrospectively retrieved from a prospective database. The quantitative metric volume of Gd-CE in T1-weighted pre-operative MRI was measured using volumetric segmentation.

RESULTS

The OR to detect CE in anaplastic gliomas was seven times higher than that in diffuse gliomas (CI95% 2.8-17.2, p<0.0001). No CE was seen in 50% (8/16) of focal anaplastic and in 28% (10/36) of entirely anaplastic gliomas. CE was present in 21% (10/47) of diffuse gliomas. Anaplasia correlated with a larger CE volume (r=0.49, p<0.0001) and provided additional 4 cm³ of CE volume compared to entirely diffuse tumors. The OR to have CE was 3.6 times for IDH1 wild-type tumors (CI95% 1.3-10.2, p=0.05) and 4.8 for tumors with ATRX expression (CI95% 1.3-17.2, p=0.05). In all sub-groups, at least a quarter of cases showed no CE at all and there were cases with present CE.

CONCLUSION

CE is associated with higher odds of unfavorable prognostic features like anaplasia, wild-type IDH1 and retained ATRX. There was no CE in one-fourth of anaplastic gliomas and half of gliomas with focal anaplasia.

Development and Validation of a Plasma FAM19A5 and MRI-Based Radiomics Model for Prediction of Parkinson's Disease and Parkinson's Disease With Depression

Background: Prediction and early diagnosis of Parkinson’s disease (PD) and Parkinson’s disease with depression (PDD) are essential for the clinical management of PD.

Objectives: The present study aimed to develop a plasma Family with sequence similarity 19, member A5 (FAM19A5) and MRI-based radiomics nomogram to predict PD and PDD.

Methods: The study involved 176 PD patients and 181 healthy controls (HC). Sandwich enzyme-linked immunosorbent assay (ELISA) was used to measure FAM19A5 concentration in the plasma samples collected from all participants. For enrolled subjects, MRI data were collected from 164 individuals (82 in the PD group and 82 in the HC group). The bilateral amygdala, head of the caudate nucleus, putamen, and substantia nigra, and red nucleus were manually labeled on the MR images. Radiomics features of the labeled regions were extracted. Further, machine learning methods were applied to shrink the feature size and build a predictive radiomics signature. The resulting radiomics signature was combined with plasma FAM19A5 concentration and other risk factors to establish logistic regression models for the prediction of PD and PDD.

Results: The plasma FAM19A5 levels (2.456 ± 0.517) were recorded to be significantly higher in the PD group as compared to the HC group (2.23 ± 0.457) (P < 0.001). Importantly, the plasma FAM19A5 levels were also significantly higher in the PDD subgroup (2.577 ± 0.408) as compared to the non-depressive subgroup (2.406 ± 0.549) (P = 0.045 < 0.05). The model based on the combination of plasma FAM19A5 and radiomics signature showed excellent predictive validity for PD and PDD, with AUCs of 0.913 (95% CI: 0.861–0.955) and 0.937 (95% CI: 0.845–0.970), respectively.

Conclusion: Altogether, the present study reported the development of nomograms incorporating radiomics signature, plasma FAM19A5, and clinical risk factors, which might serve as potential tools for early prediction of PD and PDD in clinical settings.

Development and Practical Implementation of a Deep Learning-Based Pipeline for Automated Pre- and Postoperative Glioma Segmentation

BACKGROUND AND PURPOSE

Quantitative volumetric segmentation of gliomas has important implications for diagnosis, treatment, and prognosis. We present a deep-learning model that accommodates automated preoperative and postoperative glioma segmentation with a pipeline for clinical implementation. Developed and engineered in concert, the work seeks to accelerate clinical realization of such tools.

MATERIALS AND METHODS

A deep learning model, autoencoder regularization-cascaded anisotropic, was developed, trained, and tested fusing key elements of autoencoder regularization with a cascaded anisotropic convolutional neural network. We constructed a dataset consisting of 437 cases with 40 cases reserved as a held-out test and the remainder split 80:20 for training and validation. We performed data augmentation and hyperparameter optimization and used a mean Dice score to evaluate against baseline models. To facilitate clinical adoption, we developed the model with an end-to-end pipeline including routing, preprocessing, and end-user interaction.

RESULTS

The autoencoder regularization-cascaded anisotropic model achieved median and mean Dice scores of 0.88/0.83 (SD, 0.09), 0.89/0.84 (SD, 0.08), and 0.81/0.72 (SD, 0.1) for whole-tumor, tumor core/resection cavity, and enhancing tumor subregions, respectively, including both preoperative and postoperative follow-up cases. The overall total processing time per case was ∼10 minutes, including data routing (∼1 minute), preprocessing (∼6 minute), segmentation (∼1-2 minute), and postprocessing (∼1 minute). Implementation challenges were discussed.

CONCLUSIONS

We show the feasibility and advantages of building a coordinated model with a clinical pipeline for the rapid and accurate deep learning segmentation of both preoperative and postoperative gliomas. The ability of the model to accommodate cases of postoperative glioma is clinically important for follow-up. An end-to-end approach, such as used here, may lead us toward successful clinical translation of tools for quantitative volume measures for glioma.

Magnetic resonance image-based brain tumour segmentation methods: A systematic review

Background

Image segmentation is an essential step in the analysis and subsequent characterisation of brain tumours through magnetic resonance imaging. In the literature, segmentation methods are empowered by open-access magnetic resonance imaging datasets, such as the brain tumour segmentation dataset. Moreover, with the increased use of artificial intelligence methods in medical imaging, access to larger data repositories has become vital in method development.

Purpose

To determine what automated brain tumour segmentation techniques can medical imaging specialists and clinicians use to identify tumour components, compared to manual segmentation.

Methods

We conducted a systematic review of 572 brain tumour segmentation studies during 2015-2020. We reviewed segmentation techniques using T1-weighted, T2-weighted, gadolinium-enhanced T1-weighted, fluid-attenuated inversion recovery, diffusion-weighted and perfusion-weighted magnetic resonance imaging sequences. Moreover, we assessed physics or mathematics-based methods, deep learning methods, and software-based or semi-automatic methods, as applied to magnetic resonance imaging techniques. Particularly, we synthesised each method as per the utilised magnetic resonance imaging sequences, study population, technical approach (such as deep learning) and performance score measures (such as Dice score).

Statistical tests

We compared median Dice score in segmenting the whole tumour, tumour core and enhanced tumour.

Results

We found that T1-weighted, gadolinium-enhanced T1-weighted, T2-weighted and fluid-attenuated inversion recovery magnetic resonance imaging are used the most in various segmentation algorithms. However, there is limited use of perfusion-weighted and diffusion-weighted magnetic resonance imaging. Moreover, we found that the U-Net deep learning technology is cited the most, and has high accuracy (Dice score 0.9) for magnetic resonance imaging-based brain tumour segmentation.

Conclusion

U-Net is a promising deep learning technology for magnetic resonance imaging-based brain tumour segmentation. The community should be encouraged to contribute open-access datasets so training, testing and validation of deep learning algorithms can be improved, particularly for diffusion- and perfusion-weighted magnetic resonance imaging, where there are limited datasets available.