TotalSegmentator: Robust Segmentation of 104 Anatomic Structures in CT Images

Instance-level medical image classification for text-based retrieval in a medical data integration center

A medical data integration center integrates a large volume of medical images from clinical departments, including X-rays, CT scans, and MRI scans. Ideally, all images should be indexed appropriately with standard clinical terms. However, some images have incorrect or missing annotations, which creates challenges in searching and integrating data centrally. To address this issue, accurate and meaningful descriptors are needed for indexing fields, enabling users to efficiently search for desired images and integrate them with international standards. This paper aims to provide concise annotation for missing or incorrectly indexed fields, incorporating essential instance-level information such as radiology modalities (e.g., X-rays), anatomical regions (e.g., chest), and body orientations (e.g., lateral) using a Deep Learning classification model - ResNet50. To demonstrate the capabilities of our algorithm in generating annotations for indexing fields, we conducted three experiments using two open-source datasets, the ROCO dataset, and the IRMA dataset, along with a custom dataset featuring SNOMED CT labels. While the outcomes of these experiments are satisfactory (Precision of >75%) for less critical tasks and serve as a valuable testing ground for image retrieval, they also underscore the need for further exploration of potential challenges. This essay elaborates on the identified issues and presents well-founded recommendations for refining and advancing our proposed approach.

SUVmean ratios of liver/muscle and lung/muscle from 13N-NH3 PET perfusion outperformed traditional myocardial viability parameters in predicting survival after CABG

PURPOSE

Myocardial viability evaluation in predicting survival after coronary artery bypass graft (CABG) remains debatable. Thus, this study aimed to investigate the role of 13N-NH3/18F-FDG PET myocardial viability scan in predicting treatment outcomes and survival.

METHODS

90 patients with CABG and pre-surgical PET-based myocardial viability scan were retrospectively reviewed. Perfusion-metabolism features, myocardium motion parameters, and patient characteristics were recorded. Additionally, the SUVmean of blood pool, lung, liver, spleen, and muscle were measured and the SUVmean ratios were calculated. Factors associated with treatment outcomes and survival were analyzed by Logistic and Cox regressions. Nomogram models were subsequently established to predict ejection fraction (EF) improvement and survival outcomes.

RESULTS

The mean EF of these 90 patients was 38.1 ± 9.5% and 46.0 ± 9.2% before and after CABG surgery, and 35 patients (38.9%) achieved EF improvement ≥ 10%. EF measurements by PET and echocardiogram showed a reasonable linear correlation (R = 0.752). Sex, pre-surgical EF, mismatch of the left ventricle, total perfusion deficit (TPD), and peak ejection rate (PER) were independent predictive factors of EF improvements. Surgery waiting time, valve damage, and SUVmean ratio of Liver/Muscle were independently predictive of event-free survival (EFS), while valve damage, together with SUVmean ratio of either Liver/Muscle or Lung/Muscle, were independently predictive of overall survival (OS).

CONCLUSION

Although traditional cardiac parameters from PET-based myocardial viability can effectively predict EF improvements after CABG, SUVmean ratios of liver/muscle and lung/muscle from 13N-NH3 PET perfusion outperformed these parameters in predicting survival.

Cumulative rib fracture risk after stereotactic body radiotherapy in patients with localized non-small cell lung cancer

INTRODUCTION

Rib fracture is a known complication after stereotactic body radiotherapy (SBRT). Patient-related parameters are essential to provide patient-tailored risk estimation, however, their impact on rib fracture is less documented compared to dosimetric parameters. This study aimed to predict the risk of rib fractures in patients with localized non-small cell lung cancer (NSCLC) post-SBRT based on both patient-related and dosimetric parameters with death as a competing risk.

MATERIALS AND METHODS

In total, 602 patients with localized NSCLC treated with SBRT between 2010-2020 at Odense University Hospital, Denmark were included. All patients received SBRT with 45-66 Gray (Gy)/3 fractions. Rib fractures were identified in CT-scans using a word embedding model. The cumulative incidence function was based on cause-specific Cox hazard models with variable selection based on cross-validation model likelihood performed using 50 bootstraps.

RESULTS

In total, 19 % of patients experienced a rib fracture. The cumulative risk of rib fracture increased rapidly from 6-54 months post-SBRT. Female gender, bone density, near max dose to the rib, V30 and V40 to the rib, gross tumor volume, and mean lung dose were significantly associated with rib fracture risk in univariable analysis. The final multi-variable model consisted of V20 and V30 to the rib and mean lung dose.

CONCLUSION

Female gender and low bone density in male patients are significant predictors of rib fracture risk. The final model predicting cumulative rib fracture risk of 19 % in patients with localized NSCLC treated with SBRT contained no patient-related parameters, suggesting that dosimetric parameters are the primary drivers.

Trauma and 'Whole' Body Computed Tomography: Role, Protocols, Appropriateness, and Evidence to Support its Use and When

Imaging plays a crucial role in the immediate evaluation of the trauma patient, particularly using multi-detector computed tomography (CT), and especially in moderately to severely injured trauma patients. There are specific areas of relative consensus, while other aspects of whole-body computed tomography (WB-CT) use remain controversial and are subject to opinion/debate based on the current literature. Even a few hours of a delayed diagnosis may result in a detrimental outcome for the patient. One must utilize all the tools available to enhance the interpretation of images. It is also important to recognize imaging pitfalls and artifacts to avoid unnecessary intervention.

SPINEPS-automatic whole spine segmentation of T2-weighted MR images using a two-phase approach to multi-class semantic and instance segmentation

OBJECTIVES

Introducing SPINEPS, a deep learning method for semantic and instance segmentation of 14 spinal structures (ten vertebra substructures, intervertebral discs, spinal cord, spinal canal, and sacrum) in whole-body sagittal T2-weighted turbo spin echo images.

MATERIAL AND METHODS

This local ethics committee-approved study utilized a public dataset (train/test 179/39 subjects, 137 female), a German National Cohort (NAKO) subset (train/test 1412/65 subjects, mean age 53, 694 female), and an in-house dataset (test 10 subjects, mean age 70, 5 female). SPINEPS is a semantic segmentation model, followed by a sliding window approach utilizing a second model to create instance masks from the semantic ones. Segmentation evaluation metrics included the Dice score and average symmetrical surface distance (ASSD). Statistical significance was assessed using the Wilcoxon signed-rank test.

RESULTS

On the public dataset, SPINEPS outperformed a nnUNet baseline on every structure and metric (e.g., an average over vertebra instances: dice 0.933 vs 0.911, p < 0.001, ASSD 0.21 vs 0.435, p < 0.001). SPINEPS trained on automated annotations of the NAKO achieves an average global Dice score of 0.918 on the combined NAKO and in-house test split. Adding the training data from the public dataset outperforms this (average instance-wise Dice score over the vertebra substructures 0.803 vs 0.778, average global Dice score 0.931 vs 0.918).

CONCLUSION

SPINEPS offers segmentation of 14 spinal structures in T2w sagittal images. It provides a semantic mask and an instance mask separating the vertebrae and intervertebral discs. This is the first publicly available algorithm to enable this segmentation.

KEY POINTS

Question No publicly available automatic approach can yield semantic and instance segmentation masks for the whole spine (including posterior elements) in T2-weighted sagittal TSE images. Findings Segmenting semantically first and then instance-wise outperforms a baseline trained directly on instance segmentation. The developed model produces high-resolution MRI segmentations for the whole spine. Clinical relevance This study introduces an automatic approach to whole spine segmentation, including posterior elements, in arbitrary fields of view T2w sagittal MR images, enabling easy biomarker extraction, automatic localization of pathologies and degenerative diseases, and quantifying analyses as downstream research.

Correlation of dynamic blood dose with clinical outcomes in radiotherapy for head-and-neck cancer

BACKGROUND AND PURPOSE

Radiation-induced lymphopenia (RIL) during cancer radiotherapy is receiving growing attention due to its association with adverse clinical outcomes. Correlations between RIL and poorer locoregional control (LRC), distant-metastasis-free survival (DMFS), and overall survival (OS) have been demonstrated across multiple treatment sites. Estimates of radiation delivered to circulating blood or lymphocytes have been shown to be correlated with severe RIL. This study aims to evaluate whether blood dose estimates are equally correlated with patient outcomes directly.

MATERIALS AND METHODS

For 298 head-and-neck cancer patients, blood dose was estimated via the total body dose (Dbody), a static blood dose model considering the mean dose to relevant organs and tissues (Dstatic), and a dynamic model which further included temporal aspects such as blood flow and treatment delivery time (Ddynamic). The latter utilized hematological dose (HEDOS), an open-source computational tool for blood dose simulations. Survival analysis was performed to evaluate potential correlations between blood dose and LRC, DMFS, and OS.

RESULTS

Multivariable Cox regression analysis found a statistically significant (p < 0.05) correlation between various dynamic blood dose metrics and clinical outcomes. Dbody and Dstatic did not correlate with any of the outcomes considered.

CONCLUSION

A statistically significant correlation between the dynamic blood dose model and adverse clinical outcomes was observed. During multivariable regression analysis, neither static blood dose model exhibited a statistically significant correlation with any of the outcomes studied.

A student trained convolutional neural network competing with a commercial AI software and experts in organ at risk segmentation

This retrospective, multi-centered study aimed to improve high-quality radiation treatment (RT) planning workflows by training and testing a Convolutional Neural Network (CNN) to perform auto segmentations of organs at risk (OAR) for prostate cancer (PCa) patients, specifically the bladder and rectum. The objective of this project was to develop a clinically applicable and robust artificial intelligence (AI) system to assist radiation oncologists in OAR segmentation. The CNN was trained using manual contours in CT-datasets from diagnostic 68Ga-PSMA-PET/CTs by a student, then validated (n = 30, PET/CTs) and tested (n = 16, planning CTs). Further segmentations were generated by a commercial artificial intelligence (cAI) software. The ground truth were manual contours from expert radiation oncologists. The performance was evaluated using the Dice-Sørensen Coefficient (DSC), visual analysis and a Turing test. The CNN yielded excellent results in both cohorts and OARs with a DSCmedian > 0.87, the cAI resulted in a DSC > 0.78. In the visual assessment, 67% (bladder) and 75% (rectum) of the segmentations were rated as acceptable for treatment planning. With a misclassification rate of 45.5% (bladder) and 51.1% (rectum), the CNN passed the Turing test. The metrics, visual assessment and the Turing test confirmed the clinical applicability and therefore the support in clinical routine.

Adversarial robustness improvement for X-ray bone segmentation using synthetic data created from computed tomography scans

Deep learning-based image analysis offers great potential in clinical practice. However, it faces mainly two challenges: scarcity of large-scale annotated clinical data for training and susceptibility to adversarial data in inference. As an example, an artificial intelligence (AI) system could check patient positioning, by segmenting and evaluating relative positions of anatomical structures in medical images. Nevertheless, data to train such AI system might be highly imbalanced with mostly well-positioned images being available. Thus, we propose the use of synthetic X-ray images and annotation masks forward projected from 3D photon-counting CT volumes to create realistic non-optimally positioned X-ray images for training. An open-source model (TotalSegmentator) was used to annotate the clavicles in 3D CT volumes. We evaluated model robustness with respect to the internal (simulated) patient rotation α on real-data-trained models and real&synthetic-data-trained models. Our results showed that real&synthetic- data-trained models have Dice score percentage improvements of 3% to 15% across different α groups compared to the real-data-trained model. Therefore, we demonstrated that synthetic data could be supplementary used to train and enrich heavily underrepresented conditions to increase model robustness.

Influence of X-ray spectrum and bowtie filter characterisation on the accuracy of Monte Carlo simulated organ doses: Validation in a whole-body CT scanning mode

PURPOSE

For patient-specific CT dosimetry, Monte Carlo dose simulations require an accurate description of the CT scanner. However, quantitative spectral information and information on the bowtie filter material and shape from the manufacturer is often not available. In this study, the influence of different X-ray spectra and bowtie filter characterisation methods on simulated CT organ doses is studied.

METHODS

Using ImpactMC, organ doses of whole-body CTs were simulated in twenty adult whole-body voxel models, generated from PET/CT examinations previously conducted in these patients. Simulated CT organ doses based on the manufacturer X-ray spectra and bowtie filter data were compared with those obtained using alternative characterisation models, including spectrum generators and experimentally measured dose data. A total of four different X-ray spectra and one bowtie filter model were defined based on these data.

RESULTS

For all X-ray spectra and bowtie filter combinations, estimated CT organ doses are within 6% from those resulting from simulations with the CT characterisation models provided by the manufacturer. While varying the bowtie filter model results in CT organ dose differences smaller than 1%, dose differences up to 6% are observed when X-ray spectra are not based on the quantitative data from the manufacturer.

CONCLUSIONS

Estimated organ doses slightly depend on the applied CT characterisation model. When manufacturer's data are not available, half-value layer and dose measurements provide sufficient input to obtain equivalent X-ray spectra and bowtie filter profiles, respectively.

AI-Generated Annotations Dataset for Diverse Cancer Radiology Collections in NCI Image Data Commons

The National Cancer Institute (NCI) Image Data Commons (IDC) offers publicly available cancer radiology collections for cloud computing, crucial for developing advanced imaging tools and algorithms. Despite their potential, these collections are minimally annotated; only 4% of DICOM studies in collections considered in the project had existing segmentation annotations. This project increases the quantity of segmentations in various IDC collections. We produced high-quality, AI-generated imaging annotations dataset of tissues, organs, and/or cancers for 11 distinct IDC image collections. These collections contain images from a variety of modalities, including computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography (PET). The collections cover various body parts, such as the chest, breast, kidneys, prostate, and liver. A portion of the AI annotations were reviewed and corrected by a radiologist to assess the performance of the AI models. Both the AI's and the radiologist's annotations were encoded in conformance to the Digital Imaging and Communications in Medicine (DICOM) standard, allowing for seamless integration into the IDC collections as third-party analysis collections. All the models, images and annotations are publicly accessible.

Fully Automated Region-Specific Human-Perceptive-Equivalent Image Quality Assessment: Application to 18F-FDG PET Scans

INTRODUCTION

We propose a fully automated framework to conduct a region-wise image quality assessment (IQA) on whole-body 18F-FDG PET scans. This framework (1) can be valuable in daily clinical image acquisition procedures to instantly recognize low-quality scans for potential rescanning and/or image reconstruction, and (2) can make a significant impact in dataset collection for the development of artificial intelligence-driven 18F-FDG PET analysis models by rejecting low-quality images and those presenting with artifacts, toward building clean datasets.

PATIENTS AND METHODS

Two experienced nuclear medicine physicians separately evaluated the quality of 174 18F-FDG PET images from 87 patients, for each body region, based on a 5-point Likert scale. The body regisons included the following: (1) the head and neck, including the brain, (2) the chest, (3) the chest-abdomen interval (diaphragmatic region), (4) the abdomen, and (5) the pelvis. Intrareader and interreader reproducibility of the quality scores were calculated using 39 randomly selected scans from the dataset. Utilizing a binarized classification, images were dichotomized into low-quality versus high-quality for physician quality scores ≤3 versus >3, respectively. Inputting the 18F-FDG PET/CT scans, our proposed fully automated framework applies 2 deep learning (DL) models on CT images to perform region identification and whole-body contour extraction (excluding extremities), then classifies PET regions as low and high quality. For classification, 2 mainstream artificial intelligence-driven approaches, including machine learning (ML) from radiomic features and DL, were investigated. All models were trained and evaluated on scores attributed by each physician, and the average of the scores reported. DL and radiomics-ML models were evaluated on the same test dataset. The performance evaluation was carried out on the same test dataset for radiomics-ML and DL models using the area under the curve, accuracy, sensitivity, and specificity and compared using the Delong test with P values <0.05 regarded as statistically significant.

RESULTS

In the head and neck, chest, chest-abdomen interval, abdomen, and pelvis regions, the best models achieved area under the curve, accuracy, sensitivity, and specificity of [0.97, 0.95, 0.96, and 0.95], [0.85, 0.82, 0.87, and 0.76], [0.83, 0.76, 0.68, and 0.80], [0.73, 0.72, 0.64, and 0.77], and [0.72, 0.68, 0.70, and 0.67], respectively. In all regions, models revealed highest performance, when developed on the quality scores with higher intrareader reproducibility. Comparison of DL and radiomics-ML models did not show any statistically significant differences, though DL models showed overall improved trends.

CONCLUSIONS

We developed a fully automated and human-perceptive equivalent model to conduct region-wise IQA over 18F-FDG PET images. Our analysis emphasizes the necessity of developing separate models for body regions and performing data annotation based on multiple experts' consensus in IQA studies.

Fully automated measurement of noise, signal-to-noise ratio, and contrast-to-noise ratio on chest CT images: feasibility and efficiency

BACKGROUND

Rapid and accurate measurement of computed tomography (CT) image noise, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) is a clinical challenge.

PURPOSE

To explore the feasibility of intelligent measurement of chest CT image noise, SNR, and CNR.

MATERIAL AND METHODS

A total of 300 chest CT scans were included in the study, which was divided into research dataset, internal test dataset, and external test dataset. Based on the research dataset, automatically segment and measure the average CT values and standard deviation (SD) of CT values for background air and lung field under different thresholds to obtain noise, SNR, and CNR results. Using the results of manual measurements as the reference standard, we determine the optimal threshold with the highest consistency. Using internal and external test datasets, validate the consistency of automated measurements of noise, SNR, and CNR at the optimal CT threshold with reference standards.

RESULTS

With background air set at -900 HU and lung field at -800 HU as thresholds, the automated measurements of noise, SNR, and CNR demonstrate the highest consistency with the reference standards. At the optimal threshold, the noise, SNR, and CNR measured automatically on both the internal (intraclass correlation coefficient [ICC] = 0.85-0.96) and external (ICC = 0.75-0.85) test datasets exhibit high consistency with their respective reference standards.

CONCLUSION

The method we explored can intelligently measure the noise, SNR, and CNR of chest CT images, exhibits high consistency with radiologists, and offers a novel tool for image quality evaluation and analysis.

MedLSAM: Localize and segment anything model for 3D CT images

Recent advancements in foundation models have shown significant potential in medical image analysis. However, there is still a gap in models specifically designed for medical image localization. To address this, we introduce MedLAM, a 3D medical foundation localization model that accurately identifies any anatomical part within the body using only a few template scans. MedLAM employs two self-supervision tasks: unified anatomical mapping (UAM) and multi-scale similarity (MSS) across a comprehensive dataset of 14,012 CT scans. Furthermore, we developed MedLSAM by integrating MedLAM with the Segment Anything Model (SAM). This innovative framework requires extreme point annotations across three directions on several templates to enable MedLAM to locate the target anatomical structure in the image, with SAM performing the segmentation. It significantly reduces the amount of manual annotation required by SAM in 3D medical imaging scenarios. We conducted extensive experiments on two 3D datasets covering 38 distinct organs. Our findings are twofold: (1) MedLAM can directly localize anatomical structures using just a few template scans, achieving performance comparable to fully supervised models; (2) MedLSAM closely matches the performance of SAM and its specialized medical adaptations with manual prompts, while minimizing the need for extensive point annotations across the entire dataset. Moreover, MedLAM has the potential to be seamlessly integrated with future 3D SAM models, paving the way for enhanced segmentation performance. Our code is public at https://github.com/openmedlab/MedLSAM.

CT-based body composition analysis and pulmonary fat attenuation volume as biomarkers to predict overall survival in patients with non-specific interstitial pneumonia

BACKGROUND

Non-specific interstitial pneumonia (NSIP) is an interstitial lung disease that can result in end-stage fibrosis. We investigated the influence of body composition and pulmonary fat attenuation volume (CTpfav) on overall survival (OS) in NSIP patients.

METHODS

In this retrospective single-center study, 71 NSIP patients with a median age of 65 years (interquartile range 21.5), 39 females (55%), who had a computed tomography from August 2009 to February 2018, were included, of whom 38 (54%) died during follow-up. Body composition analysis was performed using an open-source nnU-Net-based framework. Features were combined into: Sarcopenia (muscle/bone); Fat (total adipose tissue/bone); Myosteatosis (inter-/intra-muscular adipose tissue/total adipose tissue); Mediastinal (mediastinal adipose tissue/bone); and Pulmonary fat index (CTpfav/lung volume). Kaplan-Meier analysis with a log-rank test and multivariate Cox regression were used for survival analyses.

RESULTS

Patients with a higher (> median) Sarcopenia and lower (< median) Mediastinal Fat index had a significantly better survival probability (2-year survival rate: 83% versus 71% for high versus low Sarcopenia index, p = 0.023; 83% versus 72% for low versus high Mediastinal fat index, p = 0.006). In univariate analysis, individuals with a higher Pulmonary fat index exhibited significantly worse survival probability (2-year survival rate: 61% versus 94% for high versus low, p = 0.003). Additionally, it was an independent risk predictor for death (hazard ratio 2.37, 95% confidence interval 1.03-5.48, p = 0.043).

CONCLUSION

Fully automated body composition analysis offers interesting perspectives in patients with NSIP. Pulmonary fat index was an independent predictor of OS.

RELEVANCE STATEMENT

The Pulmonary fat index is an independent predictor of OS in patients with NSIP and demonstrates the potential of fully automated, deep-learning-driven body composition analysis as a biomarker for prognosis estimation.

KEY POINTS

This is the first study assessing the potential of CT-based body composition analysis in patients with non-specific interstitial pneumonia (NSIP). A single-center analysis of 71 patients with board-certified diagnosis of NSIP is presented Indices related to muscle, mediastinal fat, and pulmonary fat attenuation volume were significantly associated with survival at univariate analysis. CT pulmonary fat attenuation volume, normalized by lung volume, resulted as an independent predictor for death.

Deep learning-based defacing tool for CT angiography: CTA-DEFACE

The growing use of artificial neural network (ANN) tools for computed tomography angiography (CTA) data analysis underscores the necessity for elevated data protection measures. We aimed to establish an automated defacing pipeline for CTA data. In this retrospective study, CTA data from multi-institutional cohorts were utilized to annotate facemasks (n = 100) and train an ANN model, subsequently tested on an external institution's dataset (n = 50) and compared to a publicly available defacing algorithm. Face detection (MTCNN) and verification (FaceNet) networks were applied to measure the similarity between the original and defaced CTA images. Dice similarity coefficient (DSC), face detection probability, and face similarity measures were calculated to evaluate model performance. The CTA-DEFACE model effectively segmented soft face tissue in CTA data achieving a DSC of 0.94 ± 0.02 (mean ± standard deviation) on the test set. Our model was benchmarked against a publicly available defacing algorithm. After applying face detection and verification networks, our model showed substantially reduced face detection probability (p < 0.001) and similarity to the original CTA image (p < 0.001). The CTA-DEFACE model enabled robust and precise defacing of CTA data. The trained network is publicly accessible at www.github.com/neuroAI-HD/CTA-DEFACE . RELEVANCE STATEMENT: The ANN model CTA-DEFACE, developed for automatic defacing of CT angiography images, achieves significantly lower face detection probabilities and greater dissimilarity from the original images compared to a publicly available model. The algorithm has been externally validated and is publicly accessible. KEY POINTS: The developed ANN model (CTA-DEFACE) automatically generates facemasks for CT angiography images. CTA-DEFACE offers superior deidentification capabilities compared to a publicly available model. By means of graphics processing unit optimization, our model ensures rapid processing of medical images. Our model underwent external validation, underscoring its reliability for real-world application.

Protein-truncating variant in APOL3 increases chronic kidney disease risk in epistasis with APOL1 risk alleles

BACKGROUNDTwo coding alleles within the APOL1 gene, G1 and G2, found almost exclusively in individuals genetically similar to West African populations, contribute substantially to the pathogenesis of chronic kidney disease (CKD). The APOL gene cluster on chromosome 22 contains a total of 6 APOL genes that have arisen as a result of gene duplication.METHODSUsing a genome-first approach in the Penn Medicine BioBank, we identified 62 protein-altering variants in the 6 APOL genes with a minor allele frequency of >0.1% in a population of participants genetically similar to African reference populations and performed population-specific phenome-wide association studies.RESULTSWe identified rs1108978, a stop-gain variant in APOL3 (p.Q58*), to be significantly associated with increased CKD risk, even after conditioning on APOL1 G1/G2 carrier status. These findings were replicated in the Veterans Affairs Million Veteran Program and the All of Us Research Program. APOL3 p.Q58* was also significantly associated with a number of quantitative traits linked to CKD, including decreased kidney volume. This truncating variant contributed the most risk for CKD in patients monoallelic for APOL1 G1/G2, suggesting an epistatic interaction and a potential protective effect of wild-type APOL3 against APOL1-induced kidney disease.CONCLUSIONThis study demonstrates the utility of targeting population-specific variants in a genome-first approach, even in the context of well-studied gene-disease relationships.FUNDINGNational Heart, Lung, and Blood Institute (F30HL172382, R01HL169378, R01HL169458), Doris Duke Foundation (grant 2023-2024), National Institute of Biomedical Imaging and Bioengineering (P41EB029460), and National Center for Advancing Translational Sciences (UL1-TR-001878).

Opportunistic Screening for Low Bone Mineral Density in Adults with Cystic Fibrosis Using Low-Dose Computed Tomography of the Chest with Artificial Intelligence

Background: Cystic fibrosis bone disease (CFBD) is a common comorbidity in adult people with cystic fibrosis (pwCF), resulting in an increased risk of bone fractures. This study evaluated the capacity of artificial intelligence (AI)-assisted low-dose chest CT (LDCT) opportunistic screening for detecting low bone mineral density (BMD) in adult pwCF. Methods: In this retrospective single-center study, 65 adult pwCF (mean age 30.1 ± 7.5 years) underwent dual-energy X-ray absorptiometry (DXA) of the lumbar vertebrae L1 to L4 to determine BMD and corresponding z-scores and completed LDCTs of the chest within three months as part of routine clinical care. A fully automated CT-based AI algorithm measured the attenuation values (Hounsfield units [HU]) of the thoracic vertebrae Th9-Th12 and first lumbar vertebra L1. The ability of the algorithm to diagnose CFBD was assessed using receiver operating characteristic (ROC) curves. Results: HU values of Th9 to L1 and DXA-derived BMD and the corresponding z-scores of L1 to L4 showed a strong correlation (all p < 0.05). The area under the curve (AUC) for diagnosing low BMD was highest for L1 (0.796; p = 0.001) and Th11 (0.835; p < 0.001), resulting in a specificity of 84.9% at a sensitivity level of 75%. The HU threshold values for distinguishing normal from low BMD were <197 (L1) and <212 (Th11), respectively. Conclusions: Routine LDCT of the chest with the fully automated AI-guided determination of thoracic and lumbar vertebral attenuation values is a valuable tool for predicting low BMD in adult pwCF, with the best results for Th11 and L1. However, further studies are required to define clear threshold values.

Automated deep learning segmentation of cardiac inflammatory FDG PET

BACKGROUND

Fluorodeoxyglucose positron emission tomography (FDG PET) with suppression of myocardial glucose utilization plays a pivotal role in diagnosing cardiac sarcoidosis. Reorientation of images to match perfusion datasets and myocardial segmentation enables consistent image scaling and quantification. However, such manual tasks are cumbersome. We developed a 3D U-Net deep-learning (DL) algorithm for automated myocardial segmentation in cardiac sarcoidosis FDG PET.

METHODS

The DL model was trained on FDG PET scans from 316 patients with left ventricular contours derived from paired perfusion datasets. Qualitative analysis of clinical readability was performed to compare DL segmentation with the current automated method on a 50-patient test subset. Additionally, left ventricle displacement and angulation, as well as SUVmax sampling were compared with inter-user reproducibility results. A hybrid workflow was also investigated to accelerate study processing time.

RESULTS

DL segmentation enhanced readability scores in over 90% of cases compared with the standard segmentation currently used in the software. DL segmentation performed similar to a trained technologist, surpassing standard segmentation for left ventricle displacement and angulation, as well as correlation of SUVmax. Using the DL segmentation as initial placement for manual segmentation significantly decreased the processing time.

CONCLUSION

A novel DL-based automated segmentation tool markedly improves processing of cardiac sarcoidosis FDG PET. This tool yields optimized splash display of sarcoidosis FDG PET datasets with no user input and offers significant processing time improvement for manual segmentation of such datasets.

Organomics: A Concept Reflecting the Importance of PET/CT Healthy Organ Radiomics in Non-Small Cell Lung Cancer Prognosis Prediction Using Machine Learning

PURPOSE

Non-small cell lung cancer is the most common subtype of lung cancer. Patient survival prediction using machine learning (ML) and radiomics analysis proved to provide promising outcomes. However, most studies reported in the literature focused on information extracted from malignant lesions. This study aims to explore the relevance and additional value of information extracted from healthy organs in addition to tumoral tissue using ML algorithms.

PATIENTS AND METHODS

This study included PET/CT images of 154 patients collected from available online databases. The gross tumor volume and 33 volumes of interest defined on healthy organs were segmented using nnU-Net deep learning-based segmentation. Subsequently, 107 radiomic features were extracted from PET and CT images (Organomics). Clinical information was combined with PET and CT radiomics from organs and gross tumor volumes considering 19 different combinations of inputs. Finally, different feature selection (FS; 5 methods) and ML (6 algorithms) algorithms were tested in a 3-fold data split cross-validation scheme. The performance of the models was quantified in terms of the concordance index (C-index) metric.

RESULTS

For an input combination of all radiomics information, most of the selected features belonged to PET Organomics and CT Organomics. The highest C-index (0.68) was achieved using univariate C-index FS method and random survival forest ML model using CT Organomics + PET Organomics as input as well as minimum depth FS method and CoxPH ML model using PET Organomics as input. Considering all 17 combinations with C-index higher than 0.65, Organomics from PET or CT images were used as input in 16 of them.

CONCLUSIONS

The selected features and C-indices demonstrated that the additional information extracted from healthy organs of both PET and CT imaging modalities improved the ML performance. Organomics could be a step toward exploiting the whole information available from multimodality medical images, contributing to the emerging field of digital twins in health care.

Monte Carlo-based simulation of virtual 3 and 4-dimensional cone-beam computed tomography from computed tomography images: An end-to-end framework and a deep learning-based speedup strategy

Background and purpose

In radiotherapy, precise comparison of fan-beam computed tomography (CT) and cone-beam CT (CBCT) arises as a commonplace, yet intricate task. This paper proposes a publicly available end-to-end pipeline featuring an intrinsic deep-learning-based speedup technique for generating virtual 3D and 4D CBCT from CT images.

Materials and methods

Physical properties, derived from CT intensity information, are obtained through automated whole-body segmentation of organs and tissues. Subsequently, Monte Carlo (MC) simulations generate CBCT X-ray projections for a full circular arc around the patient employing acquisition settings matched with a clinical CBCT scanner (modeled according to Varian TrueBeam specifications). In addition to 3D CBCT reconstruction, a 4D CBCT can be simulated with a fully time-resolved MC simulation by incorporating respiratory correspondence modeling. To address the computational complexity of MC simulations, a deep-learning-based speedup technique is developed and integrated that uses projection data simulated with a reduced number of photon histories to predict a projection that matches the image characteristics and signal-to-noise ratio of the reference simulation.

Results

MC simulations with default parameter setting yield CBCT images with high agreement to ground truth data acquired by a clinical CBCT scanner. Furthermore, the proposed speedup technique achieves up to 20-fold speedup while preserving image features and resolution compared to the reference simulation.

Conclusion

The presented MC pipeline and speedup approach provide an openly accessible end-to-end framework for researchers and clinicians to investigate limitations of image-guided radiation therapy workflows built on both (4D) CT and CBCT images.

Robust Ensemble of Two Different Multimodal Approaches to Segment 3D Ischemic Stroke Segmentation Using Brain Tumor Representation Among Multiple Center Datasets

Ischemic stroke segmentation at an acute stage is vital in assessing the severity of patients' impairment and guiding therapeutic decision-making for reperfusion. Although many deep learning studies have shown attractive performance in medical segmentation, it is difficult to use these models trained on public data with private hospitals' datasets. Here, we demonstrate an ensemble model that employs two different multimodal approaches for generalization, a more effective way to perform on external datasets. First, after we jointly train a segmentation model on diffusion-weighted imaging (DWI) and apparent diffusion coefficient (ADC) MR modalities, the model is inferred on the DWI images. Second, a channel-wise segmentation model is trained by concatenating the DWI and ADC images as input, and then is inferred using both MR modalities. Before training with ischemic stroke data, we utilized BraTS 2021, a public brain tumor dataset, for transfer learning. An extensive ablation study evaluates which strategy learns better representations for ischemic stroke segmentation. In our study, nnU-Net well-known for robustness is selected as our baseline model. Our proposed method is evaluated on three different datasets: the Asan Medical Center (AMC) I and II, and the 2022 Ischemic Stroke Lesion Segmentation (ISLES). Our experiments are widely validated over a large, multi-center, and multi-scanner dataset with a huge amount of 846 scans. Not only stroke lesion models can benefit from transfer learning using brain tumor data, but combining the MR modalities using different training schemes also highly improves segmentation performance. The method achieved a top-1 ranking in the ongoing ISLES'22 challenge and performed particularly well on lesion-wise metrics of interest to neuroradiologists, achieving a Dice coefficient of 78.69% and a lesion-wise F1 score of 82.46%. Also, the method was relatively robust on the AMC I (Dice, 60.35%; lesion-wise F1, 68.30%) and II (Dice; 74.12%; lesion-wise F1, 67.53%) datasets in different settings. The high segmentation accuracy of our proposed method could improve radiologists' ability to detect ischemic stroke lesions in MRI images. Our model weights and inference code are available on https://github.com/MDOpx/ISLES22-model-inference .

An Automated Vertebrae Localization, Segmentation, and Osteoporotic Compression Fracture Detection Pipeline for Computed Tomographic Imaging

Osteoporosis is the most common chronic metabolic bone disease worldwide. Vertebral compression fracture (VCF) is the most common type of osteoporotic fracture. Approximately 700,000 osteoporotic VCFs are diagnosed annually in the USA alone, resulting in an annual economic burden of ~$13.8B. With an aging population, the rate of osteoporotic VCFs and their associated burdens are expected to rise. Those burdens include pain, functional impairment, and increased medical expenditure. Therefore, it is of utmost importance to develop an analytical tool to aid in the identification of VCFs. Computed Tomography (CT) imaging is commonly used to detect occult injuries. Unlike the existing VCF detection approaches based on CT, the standard clinical criteria for determining VCF relies on the shape of vertebrae, such as loss of vertebral body height. We developed a novel automated vertebrae localization, segmentation, and osteoporotic VCF detection pipeline for CT scans using state-of-the-art deep learning models to bridge this gap. To do so, we employed a publicly available dataset of spine CT scans with 325 scans annotated for segmentation, 126 of which also graded for VCF (81 with VCFs and 45 without VCFs). Our approach attained 96% sensitivity and 81% specificity in detecting VCF at the vertebral-level, and 100% accuracy at the subject-level, outperforming deep learning counterparts tested for VCF detection without segmentation. Crucially, we showed that adding predicted vertebrae segments as inputs significantly improved VCF detection at both vertebral and subject levels by up to 14% Sensitivity and 20% Specificity (p-value = 0.028).

A Guideline for Open-Source Tools to Make Medical Imaging Data Ready for Artificial Intelligence Applications: A Society of Imaging Informatics in Medicine (SIIM) Survey

In recent years, the role of Artificial Intelligence (AI) in medical imaging has become increasingly prominent, with the majority of AI applications approved by the FDA being in imaging and radiology in 2023. The surge in AI model development to tackle clinical challenges underscores the necessity for preparing high-quality medical imaging data. Proper data preparation is crucial as it fosters the creation of standardized and reproducible AI models while minimizing biases. Data curation transforms raw data into a valuable, organized, and dependable resource and is a fundamental process to the success of machine learning and analytical projects. Considering the plethora of available tools for data curation in different stages, it is crucial to stay informed about the most relevant tools within specific research areas. In the current work, we propose a descriptive outline for different steps of data curation while we furnish compilations of tools collected from a survey applied among members of the Society of Imaging Informatics (SIIM) for each of these stages. This collection has the potential to enhance the decision-making process for researchers as they select the most appropriate tool for their specific tasks.

Impact of breathing signal-guided dose modulation on step-and-shoot 4D CT image reconstruction

BACKGROUND

Breathing signal-guided 4D CT sequence scanning such as the intelligent 4D CT (i4DCT) approach reduces imaging artifacts compared to conventional 4D CT. By design, i4DCT captures entire breathing cycles during beam-on periods, leading to redundant projection data and increased radiation exposure to patients exhibiting prolonged exhalation phases. A recently proposed breathing-guided dose modulation (DM) algorithm promises to lower the imaging dose by temporarily reducing the CT tube current, but the impact on image reconstruction and the resulting images have not been investigated.

PURPOSE

We evaluate the impact of breathing signal-guided DM on 4D CT image reconstruction and corresponding images.

METHODS

This study is designed as a comparative and retrospective analysis based on 104 4D CT datasets. Each dataset underwent retrospective reconstruction twice: (a) utilizing the acquired clinical projection data for reconstruction, which yields reference image data, and (b) excluding projections acquired during potential DM phases from image reconstruction, resulting in DM-affected image data. Resulting images underwent automatic organ segmentation (lung/liver). (Dis)Similarity of reference and DM-affected images were quantified by the Dice coefficient of the entire organ masks and the organ overlaps within the DM-affected slices. Further, for lung cases, (a) and (b) were deformably registered and median magnitudes of the obtained displacement field were computed. Eventually, for 17 lung cases, gross tumor volumes (GTV) were recontoured on both (a) and (b). Target volume similarity was quantified by the Hausdorff distance.

RESULTS

DM resulted in a median imaging dose reduction of 15.4% (interquartile range [IQR]: 11.3%-19.9%) for the present patient cohort. Dice coefficients for lung (n = 73 $n=73$ ) and liver (n = 31 $n=31$ ) patients were consistently high for both the entire organs and the DM-affected slices (IQR lung:0.985 / 0.982 $0.985/0.982$ [entire lung/DM-affected slices only] to0.992 / 0.989 $0.992/0.989$ ; IQR liver:0.977 / 0.972 $0.977/0.972$ to0.986 / 0.986 $0.986/0.986$ ), demonstrating that DM did not cause organ distortions or alterations. Median displacements for DM-affected to reference image registration varied; however, only two out of 73 cases exhibited a median displacement larger than one isotropic 1mm 3 ${\rm mm}^3$ voxel size. The impact on GTV definition for the end-exhalation phase was also minor (median Hausdorff distance: 0.38 mm, IQR: 0.15-0.46 mm).

CONCLUSION

This study demonstrates that breathing signal-guided DM has a minimal impact on image reconstruction and image appearance while improving patient safety by reducing dose exposure.

AI-Derived Left Ventricular Mass From Noncontrast Cardiac CT: Correlation With Contrast CT Angiography and CMR

Background

Noncontrast computed tomography (CT) scans are not used for evaluating left ventricle myocardial mass (LV mass), which is typically evaluated with contrast CT or cardiovascular magnetic resonance imaging (CMR).

Objectives

The purpose of the study was to assess the feasibility of LV mass estimation from standard, ECG-gated, noncontrast CT using an artificial intelligence (AI) approach and compare it with coronary CT angiography (CTA) and CMR.

Methods

We enrolled consecutive patients who underwent coronary CTA, which included noncontrast CT calcium scanning and contrast CTA, and CMR. The median interval between coronary CTA and CMR was 22 days (interquartile range: 3-76). We utilized a no new UNet AI model that automatically segmented noncontrast CT structures. AI measurement of LV mass was compared to contrast CTA and CMR.

Results

A total of 316 patients (age: 57.1 ± 16.7 years, 56% male) were included. The AI segmentation took on average 22 seconds per case. An excellent correlation was observed between AI and contrast CTA LV mass measures (r = 0.84, P < 0.001), with no significant differences (136.5 ± 55.3 g vs 139.6 ± 56.9 g, P = 0.133). Bland-Altman analysis showed minimal bias of 2.9. When compared to CMR, measured LV mass was higher with AI (136.5 ± 55.3 g vs 127.1 ± 53.1 g, P < 0.001). There was an excellent correlation between AI and CMR (r = 0.85, P < 0.001), with a small bias (-9.4). There were no statistical differences between the correlations of LV mass between contrast CTA and CMR or AI and CMR.

Conclusions

The AI-based automated estimation of LV mass from noncontrast CT demonstrated excellent correlations and minimal biases when compared to contrast CTA and CMR.

Automated abdominal CT contrast phase detection using an interpretable and open-source artificial intelligence algorithm

OBJECTIVES

To develop and validate an open-source artificial intelligence (AI) algorithm to accurately detect contrast phases in abdominal CT scans.

MATERIALS AND METHODS

Retrospective study aimed to develop an AI algorithm trained on 739 abdominal CT exams from 2016 to 2021, from 200 unique patients, covering 1545 axial series. We performed segmentation of five key anatomic structures-aorta, portal vein, inferior vena cava, renal parenchyma, and renal pelvis-using TotalSegmentator, a deep learning-based tool for multi-organ segmentation, and a rule-based approach to extract the renal pelvis. Radiomics features were extracted from the anatomical structures for use in a gradient-boosting classifier to identify four contrast phases: non-contrast, arterial, venous, and delayed. Internal and external validation was performed using the F1 score and other classification metrics, on the external dataset "VinDr-Multiphase CT".

RESULTS

The training dataset consisted of 172 patients (mean age, 70 years ± 8, 22% women), and the internal test set included 28 patients (mean age, 68 years ± 8, 14% women). In internal validation, the classifier achieved an accuracy of 92.3%, with an average F1 score of 90.7%. During external validation, the algorithm maintained an accuracy of 90.1%, with an average F1 score of 82.6%. Shapley feature attribution analysis indicated that renal and vascular radiodensity values were the most important for phase classification.

CONCLUSION

An open-source and interpretable AI algorithm accurately detects contrast phases in abdominal CT scans, with high accuracy and F1 scores in internal and external validation, confirming its generalization capability.

CLINICAL RELEVANCE STATEMENT

Contrast phase detection in abdominal CT scans is a critical step for downstream AI applications, deploying algorithms in the clinical setting, and for quantifying imaging biomarkers, ultimately allowing for better diagnostics and increased access to diagnostic imaging.

KEY POINTS

Digital Imaging and Communications in Medicine labels are inaccurate for determining the abdominal CT scan phase. AI provides great help in accurately discriminating the contrast phase. Accurate contrast phase determination aids downstream AI applications and biomarker quantification.

Metaverse in surgery - origins and future potential

The metaverse refers to a collective virtual space that combines physical and digital realities to create immersive, interactive environments. This space is powered by technologies such as augmented reality (AR), virtual reality (VR), artificial intelligence (AI) and blockchain. In healthcare, the metaverse can offer many applications. Specifically in surgery, potential uses of the metaverse include the possibility of conducting immersive surgical training in a VR or AR setting, and enhancing surgical planning through the adoption of three-dimensional virtual models and simulated procedures. At the intraoperative level, AR-guided surgery can assist the surgeon in real time to increase surgical precision in tumour identification and selective management of vessels. In post-operative care, potential uses of the metaverse include recovery monitoring and patient education. In urology, AR and VR have been widely explored in the past decade, mainly for surgical navigation in prostate and kidney cancer surgery, whereas only anecdotal metaverse experiences have been reported to date, specifically in partial nephrectomy. In the future, further integration of AI will improve the metaverse experience, potentially increasing the possibility of carrying out surgical navigation, data collection and virtual trials within the metaverse. However, challenges concerning data security and regulatory compliance must be addressed before the metaverse can be used to improve patient care.

Small metal artifact detection and inpainting in cardiac CT images

Background

Quantification of cardiac motion on pre-treatment CT imaging for stereotactic arrhythmia radiotherapy patients is difficult due to the presence of image artifacts caused by metal leads of implantable cardioverter-defibrillators (ICDs). The CT scanners' onboard metal artifact reduction tool does not sufficiently reduce these artifacts. More advanced artifact reduction techniques require the raw CT projection data and thus are not applicable to already reconstructed CT images. New methods are needed to accurately reduce the metal artifacts in already reconstructed CTs to recover the otherwise lost anatomical information.

Purpose

To develop a methodology to automatically detect metal artifacts in cardiac CT scans and inpaint the affected volume with anatomically consistent structures and values.

Methods

Breath-hold ECG-gated 4DCT scans of 12 patients who underwent cardiac radiation therapy for treating ventricular tachycardia were collected. The metal artifacts in the images caused by the ICD leads were manually contoured. A 2D U-Net deep learning (DL) model was developed to segment the metal artifacts automatically using eight patients for training, two for validation, and two for testing. A dataset of 592 synthetic CTs was prepared by adding segmented metal artifacts from the patient 4DCT images to artifact-free cardiac CTs of 148 patients. A 3D image inpainting DL model was trained to refill the metal artifact portion in the synthetic images with realistic image contents that approached the ground truth artifact-free images. The trained inpainting model was evaluated by analyzing the automated segmentation results of the four heart chambers with and without artifacts on the synthetic dataset. Additionally, the raw cardiac patient images with metal artifacts were processed using the inpainting model and the results of metal artifact reduction were qualitatively inspected.

Results

The artifact detection model worked well and produced a Dice score of 0.958 ± 0.008. The inpainting model for synthesized cases was able to recreate images that were nearly identical to the ground truth with a structural similarity index of 0.988 ± 0.012. With the chamber segmentations on the artifact-free images as the reference, the average surface Dice scores improved from 0.684 ± 0.247 to 0.964 ± 0.067 and the Hausdorff distance reduced from 3.4 ± 3.9 mm to 0.7 ± 0.7 mm. The inpainting model's use on cardiac patient CTs was visually inspected and the artifact-inpainted images were visually plausible.

Conclusion

We successfully developed two deep models to detect and inpaint metal artifacts in cardiac CT images. These deep models are useful to improve the heart chamber segmentation and cardiac motion analysis in CT images corrupted by mental artifacts. The trained models and example data are available to the public through GitHub.

AI-enabled CT-guided end-to-end quantification of total cardiac activity in 18FDG cardiac PET/CT for detection of cardiac sarcoidosis

Purpose

[18F]-fluorodeoxyglucose ([18F]FDG) positron emission tomography (PET) plays a central role in diagnosing and managing cardiac sarcoidosis. We propose a fully automated pipeline for quantification of [18F]FDG PET activity using deep learning (DL) segmentation of cardiac chambers on computed tomography (CT) attenuation maps and evaluate several quantitative approaches based on this framework.

Methods

We included consecutive patients undergoing [18F]FDG PET/CT for suspected cardiac sarcoidosis. DL segmented left atrium, left ventricular(LV), right atrium, right ventricle, aorta, LV myocardium, and lungs from CT attenuation scans. CT-defined anatomical regions were applied to [18F]FDG PET images automatically to target to background ratio (TBR), volume of inflammation (VOI) and cardiometabolic activity (CMA) using full sized and shrunk segmentations.

Results

A total of 69 patients were included, with mean age of 56.1 ± 13.4 and cardiac sarcoidosis present in 29 (42%). CMA had the highest prediction performance (area under the receiver operating characteristic curve [AUC] 0.919, 95% confidence interval [CI] 0.858 - 0.980) followed by VOI (AUC 0.903, 95% CI 0.834 - 0.971), TBR (AUC 0.891, 95% CI 0.819 - 0.964), and maximum standardized uptake value (AUC 0.812, 95% CI 0.701 - 0.923). Abnormal CMA (≥1) had a sensitivity of 100% and specificity 65% for cardiac sarcoidosis. Lung quantification was able to identify patients with pulmonary abnormalities.

Conclusion

We demonstrate that fully automated volumetric quantification of [18F]FDG PET for cardiac sarcoidosis based on CT attenuation map-derived volumetry is feasible, rapid, and has high prediction performance. This approach provides objective measurements of cardiac inflammation with consistent definition of myocardium and background region.

Three-dimensional numerical schemes for the segmentation of the psoas muscle in X-ray computed tomography images

The analysis of the psoas muscle in morphological and functional imaging has proved to be an accurate approach to assess sarcopenia, i.e. a systemic loss of skeletal muscle mass and function that may be correlated to multifactorial etiological aspects. The inclusion of sarcopenia assessment into a radiological workflow would need the implementation of computational pipelines for image processing that guarantee segmentation reliability and a significant degree of automation. The present study utilizes three-dimensional numerical schemes for psoas segmentation in low-dose X-ray computed tomography images. Specifically, here we focused on the level set methodology and compared the performances of two standard approaches, a classical evolution model and a three-dimension geodesic model, with the performances of an original first-order modification of this latter one. The results of this analysis show that these gradient-based schemes guarantee reliability with respect to manual segmentation and that the first-order scheme requires a computational burden that is significantly smaller than the one needed by the second-order approach.

Children Are Not Small Adults: Addressing Limited Generalizability of an Adult Deep Learning CT Organ Segmentation Model to the Pediatric Population

Deep learning (DL) tools developed on adult data sets may not generalize well to pediatric patients, posing potential safety risks. We evaluated the performance of TotalSegmentator, a state-of-the-art adult-trained CT organ segmentation model, on a subset of organs in a pediatric CT dataset and explored optimization strategies to improve pediatric segmentation performance. TotalSegmentator was retrospectively evaluated on abdominal CT scans from an external adult dataset (n = 300) and an external pediatric data set (n = 359). Generalizability was quantified by comparing Dice scores between adult and pediatric external data sets using Mann-Whitney U tests. Two DL optimization approaches were then evaluated: (1) 3D nnU-Net model trained on only pediatric data, and (2) an adult nnU-Net model fine-tuned on the pediatric cases. Our results show TotalSegmentator had significantly lower overall mean Dice scores on pediatric vs. adult CT scans (0.73 vs. 0.81, P < .001) demonstrating limited generalizability to pediatric CT scans. Stratified by organ, there was lower mean pediatric Dice score for four organs (P < .001, all): right and left adrenal glands (right adrenal, 0.41 [0.39-0.43] vs. 0.69 [0.66-0.71]; left adrenal, 0.35 [0.32-0.37] vs. 0.68 [0.65-0.71]); duodenum (0.47 [0.45-0.49] vs. 0.67 [0.64-0.69]); and pancreas (0.73 [0.72-0.74] vs. 0.79 [0.77-0.81]). Performance on pediatric CT scans improved by developing pediatric-specific models and fine-tuning an adult-trained model on pediatric images where both methods significantly improved segmentation accuracy over TotalSegmentator for all organs, especially for smaller anatomical structures (e.g., > 0.2 higher mean Dice for adrenal glands; P < .001).

Intra-Individual Reproducibility of Automated Abdominal Organ Segmentation-Performance of TotalSegmentator Compared to Human Readers and an Independent nnU-Net Model

The purpose of this study is to assess segmentation reproducibility of artificial intelligence-based algorithm, TotalSegmentator, across 34 anatomical structures using multiphasic abdominal CT scans comparing unenhanced, arterial, and portal venous phases in the same patients. A total of 1252 multiphasic abdominal CT scans acquired at our institution between January 1, 2012, and December 31, 2022, were retrospectively included. TotalSegmentator was used to derive volumetric measurements of 34 abdominal organs and structures from the total of 3756 CT series. Reproducibility was evaluated across three contrast phases per CT and compared to two human readers and an independent nnU-Net trained on the BTCV dataset. Relative deviation in segmented volumes and absolute volume deviations (AVD) were reported. Volume deviation within 5% was considered reproducible. Thus, non-inferiority testing was conducted using a 5% margin. Twenty-nine out of 34 structures had volume deviations within 5% and were considered reproducible. Volume deviations for the adrenal glands, gallbladder, spleen, and duodenum were above 5%. Highest reproducibility was observed for bones (- 0.58% [95% CI: - 0.58, - 0.57]) and muscles (- 0.33% [- 0.35, - 0.32]). Among abdominal organs, volume deviation was 1.67% (1.60, 1.74). TotalSegmentator outperformed the reproducibility of the nnU-Net trained on the BTCV dataset with an AVD of 6.50% (6.41, 6.59) vs. 10.03% (9.86, 10.20; p < 0.0001), most notably in cases with pathologic findings. Similarly, TotalSegmentator's AVD between different contrast phases was superior compared to the interreader AVD for the same contrast phase (p = 0.036). TotalSegmentator demonstrated high intra-individual reproducibility for most abdominal structures in multiphasic abdominal CT scans. Although reproducibility was lower in pathologic cases, it outperforms both human readers and a nnU-Net trained on the BTCV dataset.

Holistic shape variation of the rib cage in an adult population

Traumatic injuries to the thorax are a common occurrence, and given the disparity in outcomes, injury risk is non-uniformly distributed within the population. Rib cage geometry, in conjunction with well-established biomechanical characteristics, is thought to influence injury tolerance, but quantifiable descriptions of adult rib cage shape as a whole are lacking. Here, we develop an automated pipeline to extract whole rib cage measurements from a large population and produce distributions of these measurements to assess variability in rib cage shape. Ten measurements of whole rib cage shape were collected from 1,719 individuals aged 25-45 years old including angular, linear, areal, and volumetric measures. The resulting pipeline produced measurements with a mean percent difference to manually collected measurements of 1.7% ± 1.6%, and the whole process takes 30 s per scan. Each measurement followed a normal distribution with a maximum absolute skew value of 0.43 and a maximum absolute excess kurtosis value of 0.6. Significant differences were found between the sexes (p < 0.001) in all except angular measures. Multivariate regression revealed that demographic predictors explain 29%-68% of the variance in the data. The angular measurements had the three lowest R2 values and were also the only three to have little correlation with subject stature. Unlike other measures, rib cage height had a negative correlation with BMI. Stature was the dominant demographic factor in predicting rib cage height, coronal area, sagittal area, and volume. Subject weight was the dominant demographic factor for rib cage width, depth, axial area, and angular measurements. Age was minimally important in this cohort of adults from a narrow age range. Individuals of similar height and weight had average rib cage measurements near the regression predictions, but the range of values across all subjects encompassed a large portion of their respective distributions. Our findings characterize the variability in adult rib cage geometry, including the variation within narrow demographic criteria. In future work, these can be integrated into computer aided engineering workflows to assess the influence of whole rib cage shape on the biomechanics of the adult human thorax.

Benchmarking deep learning-based low-dose CT image denoising algorithms

BACKGROUND

Long-lasting efforts have been made to reduce radiation dose and thus the potential radiation risk to the patient for computed tomography (CT) acquisitions without severe deterioration of image quality. To this end, various techniques have been employed over the years including iterative reconstruction methods and noise reduction algorithms.

PURPOSE

Recently, deep learning-based methods for noise reduction became increasingly popular and a multitude of papers claim ever improving performance both quantitatively and qualitatively. However, the lack of a standardized benchmark setup and inconsistencies in experimental design across studies hinder the verifiability and reproducibility of reported results.

METHODS

In this study, we propose a benchmark setup to overcome those flaws and improve reproducibility and verifiability of experimental results in the field. We perform a comprehensive and fair evaluation of several state-of-the-art methods using this standardized setup.

RESULTS

Our evaluation reveals that most deep learning-based methods show statistically similar performance, and improvements over the past years have been marginal at best.

CONCLUSIONS

This study highlights the need for a more rigorous and fair evaluation of novel deep learning-based methods for low-dose CT image denoising. Our benchmark setup is a first and important step towards this direction and can be used by future researchers to evaluate their algorithms.

Device-Less Data-Driven Cardiac and Respiratory Gating Using LAFOV PET Histo Images

Background: The outstanding capabilities of modern Positron Emission Tomography (PET) to highlight small tumor lesions and provide pathological function assessment are at peril from image quality degradation caused by respiratory and cardiac motion. However, the advent of the long axial field-of-view (LAFOV) scanners with increased sensitivity, alongside the precise time-of-flight (TOF) of modern PET systems, enables the acquisition of ultrafast time resolution images, which can be used for estimating and correcting the cyclic motion. Methods: 0.25 s so-called [18F]FDG PET histo image series were generated in the scope of for detecting respiratory and cardiac frequency estimates applicable for performing device-less data-driven gated image reconstructions. The frequencies of the cardiac and respiratory motion were estimated for 18 patients using Short Time Fourier Transform (STFT) with 20 s and 30 s window segments, respectively. Results: The Fourier analysis provided time points usable as input to the gated reconstruction based on eight equally spaced time gates. The cardiac investigations showed estimates in accordance with the measured pulse oximeter references (p = 0.97) and a mean absolute difference of 0.4 ± 0.3 beats per minute (bpm). The respiratory frequencies were within the expected range of 10-20 respirations per minute (rpm) in 16 out of 18 patients. Using this setup, the analysis of three patients with visible lung tumors showed an increase in tumor SUVmax and a decrease in tumor volume compared to the non-gated reconstructed image. Conclusions: The method can provide signals that were applicable for gated reconstruction of both cardiac and respiratory motion, providing a potential increased diagnostic accuracy.

Radiomics machine learning algorithm facilitates detection of small pancreatic neuroendocrine tumors on CT

PURPOSE

The purpose of this study was to develop a radiomics-based algorithm to identify small pancreatic neuroendocrine tumors (PanNETs) on CT and evaluate its robustness across manual and automated segmentations, exploring the feasibility of automated screening.

MATERIALS AND METHODS

Patients with pathologically confirmed T1 stage PanNETs and healthy controls undergoing dual-phase CT imaging were retrospectively identified. Manual segmentation of pancreas and tumors was performed, then automated pancreatic segmentations were generated using a pretrained neural network. A total of 1223 radiomics features were independently extracted from both segmentation volumes, in the arterial and venous phases separately. Ten final features were selected to train classifiers to identify PanNETs and controls. The cohort was divided into training and testing sets, and performance of classifiers was assessed using area under the receiver operator characteristic curve (AUC), specificity and sensitivity, and compared against two radiologists blinded to the diagnoses.

RESULTS

A total of 135 patients with 142 PanNETs, and 135 healthy controls were included. There were 168 women and 102 men, with a mean age of 55.4 ± 11.6 (standard deviation) years (range: 20-85 years). Median PanNET size was 1.3 cm (Q1, 1.0; Q3, 1.5; range: 0.5-1.9). The arterial phase LightGBM model achieved the best performance in the test set, with 90 % sensitivity (95 % confidence interval [CI]: 80-98), 76 % specificity (95 % CI: 62-88) and an AUC of 0.87 (95 % CI: 0.79-0.94). Using features from the automated segmentations, this model achieved an AUC of 0.86 (95 % CI: 0.79-0.93). In comparison, the two radiologists achieved a mean 50 % sensitivity and 100 % specificity using arterial phase CT images.

CONCLUSION

Radiomics features identify small PanNETs, with stable performance when extracted using automated segmentations. These models demonstrate high sensitivity, complementing the high specificity of radiologists, and could serve as opportunistic screeners.

Subcutaneous edema segmentation on abdominal CT using multi-class labels and iterative annotation

PURPOSE

Anasarca is a condition that results from organ dysfunctions, such as heart, kidney, or liver failure, characterized by the presence of edema throughout the body. The quantification of accumulated edema may have potential clinical benefits. This work focuses on accurately estimating the amount of edema non-invasively using abdominal CT scans, with minimal false positives. However, edema segmentation is challenging due to the complex appearance of edema and the lack of manually annotated volumes.

METHODS

We propose a weakly supervised approach for edema segmentation using initial edema labels from the current state-of-the-art method for edema segmentation (Intensity Prior), along with labels of surrounding tissues as anatomical priors. A multi-class 3D nnU-Net was employed as the segmentation network, and training was performed using an iterative annotation workflow.

RESULTS

We evaluated segmentation accuracy on a test set of 25 patients with edema. The average Dice Similarity Coefficient of the proposed method was similar to Intensity Prior (61.5% vs. 61.7%; p = 0.83 ). However, the proposed method reduced the average False Positive Rate significantly, from 1.8% to 1.1% ( p < 0.001 ). Edema volumes computed using automated segmentation had a strong correlation with manual annotation (R 2 = 0.87 ).

CONCLUSION

Weakly supervised learning using 3D multi-class labels and iterative annotation is an efficient way to perform high-quality edema segmentation with minimal false positives. Automated edema segmentation can produce edema volume estimates that are highly correlated with manual annotation. The proposed approach is promising for clinical applications to monitor anasarca using estimated edema volumes.

Training robust T1-weighted magnetic resonance imaging liver segmentation models using ensembles of datasets with different contrast protocols and liver disease etiologies

Image segmentation of the liver is an important step in treatment planning for liver cancer. However, manual segmentation at a large scale is not practical, leading to increasing reliance on deep learning models to automatically segment the liver. This manuscript develops a generalizable deep learning model to segment the liver on T1-weighted MR images. In particular, three distinct deep learning architectures (nnUNet, PocketNet, Swin UNETR) were considered using data gathered from six geographically different institutions. A total of 819 T1-weighted MR images were gathered from both public and internal sources. Our experiments compared each architecture's testing performance when trained both intra-institutionally and inter-institutionally. Models trained using nnUNet and its PocketNet variant achieved mean Dice-Sorensen similarity coefficients>0.9 on both intra- and inter-institutional test set data. The performance of these models suggests that nnUNet and PocketNet liver segmentation models trained on a large and diverse collection of T1-weighted MR images would on average achieve good intra-institutional segmentation performance.

The reliability of virtual non-contrast reconstructions of photon-counting detector CT scans in assessing abdominal organs

BACKGROUND

Spectral imaging of photon-counting detector CT (PCD-CT) scanners allows for generating virtual non-contrast (VNC) reconstruction. By analyzing 12 abdominal organs, we aimed to test the reliability of VNC reconstructions in preserving HU values compared to real unenhanced CT images.

METHODS

Our study included 34 patients with pancreatic cystic neoplasm (PCN). The VNC reconstructions were generated from unenhanced, arterial, portal, and venous phase PCD-CT scans using the Liver-VNC algorithm. The observed 11 abdominal organs were segmented by the TotalSegmentator algorithm, the PCNs were segmented manually. Average densities were extracted from unenhanced scans (HUunenhanced), postcontrast (HUpostcontrast) scans, and VNC reconstructions (HUVNC). The error was calculated as HUerror=HUVNC-HUunenhanced. Pearson's or Spearman's correlation was used to assess the association. Reproducibility was evaluated by intraclass correlation coefficients (ICC).

RESULTS

Significant differences between HUunenhanced and HUVNC[unenhanced] were found in vertebrae, paraspinal muscles, liver, and spleen. HUVNC[unenhanced] showed a strong correlation with HUunenhanced in all organs except spleen (r = 0.45) and kidneys (r = 0.78 and 0.73). In all postcontrast phases, the HUVNC had strong correlations with HUunenhanced in all organs except the spleen and kidneys. The HUerror had significant correlations with HUunenhanced in the muscles and vertebrae; and with HUpostcontrast in the spleen, vertebrae, and paraspinal muscles in all postcontrast phases. All organs had at least one postcontrast VNC reconstruction that showed good-to-excellent agreement with HUunenhanced during ICC analysis except the vertebrae (ICC: 0.17), paraspinal muscles (ICC: 0.64-0.79), spleen (ICC: 0.21-0.47), and kidneys (ICC: 0.10-0.31).

CONCLUSIONS

VNC reconstructions are reliable in at least one postcontrast phase for most organs, but further improvement is needed before VNC can be utilized to examine the spleen, kidneys, and vertebrae.

An automated and time-efficient framework for simulation of coronary blood flow under steady and pulsatile conditions

BACKGROUND AND OBJECTIVE

Invasive fractional flow reserve (FFR) measurement is the gold standard method for coronary artery disease (CAD) diagnosis. FFR-CT exploits computational fluid dynamics (CFD) for non-invasive evaluation of FFR, simulating coronary flow in virtual geometries reconstructed from computed tomography (CT), but suffers from cost-intensive computing process and uncertainties in the definition of patient specific boundary conditions (BCs). In this work, we investigated the use of time-averaged steady BCs, compared to pulsatile to reduce the computational time and deployed a self-adjusting method for the tuning of BCs to patient-specific clinical data.

METHODS

133 coronary arteries were reconstructed form CT images of patients suffering from CAD. For each vessel, invasive FFR was measured. After segmentation, the geometries were prepared for CFD simulation by clipping the outlets and discretizing into tetrahedral mesh. Steady BCs were defined in two steps: (i) rest BCs were extrapolated from clinical and image-derived data; (ii) hyperemic BCs were computed from resting conditions. Flow rate was iteratively adjusted during the simulation, until patient's aortic pressure was matched. Pulsatile BCs were defined exploiting the convergence values of steady BCs. After CFD simulation, lesion-specific hemodynamic indexes were computed and compared between group of patients for which surgery was indicated and not. The whole pipeline was implemented as a straightforward process, in which each single step is performed automatically.

RESULTS

Steady and pulsatile FFR-CT yielded a strong correlation (r = 0.988, p < 0.001) and correlated with invasive FFR (r = 0.797, p < 0.001). The per-point difference between the pressure and FFR-CT field predicted by the two methods was below 1 % and 2 %, respectively. Both approaches exhibited a good diagnostic performance: accuracy was 0.860 and 0.864, the AUC was 0.923 and 0.912, for steady and pulsatile case, respectively. The computational time required by steady BCs CFD was approximatively 30-folds lower than pulsatile case.

CONCLUSIONS

This work shows the feasibility of using steady BCs CFD for computing the FFR-CT in coronary arteries, as well as its computational and diagnostic performance within a fully automated pipeline.

Joint segmentation of sternocleidomastoid and skeletal muscles in computed tomography images using a multiclass learning approach

Deep-learning-based methods can improve robustness against individual variations in computed tomography (CT) images of the sternocleidomastoid muscle, which is a challenge when using conventional methods based on probabilistic atlases are used for automatic segmentation. Thus, this study proposes a novel multiclass learning approach for the joint segmentation of the sternocleidomastoid and skeletal muscles in CT images, and it employs a two-dimensional U-Net architecture. The proposed method concurrently learns and segmented segments the sternocleidomastoid muscle and the entire skeletal musculature. Consequently, three-dimensional segmentation results are generated for both muscle groups. Experiments conducted on a dataset of 30 body CT images demonstrated segmentation accuracies of 82.94% and 92.73% for the sternocleidomastoid muscle and entire skeletal muscle compartment, respectively. These results outperformed those of conventional methods, such as the single-region learning of a target muscle and multiclass learning of specific muscle pairs. Moreover, the multiclass learning paradigm facilitated a robust segmentation performance regardless of the input image range. This highlights the method's potential for cases that present muscle atrophy or reduced muscle strength. The proposed method exhibits promising capabilities for the high-accuracy joint segmentation of the sternocleidomastoid and skeletal muscles and is effective in recognizing skeletal muscles, thus, it holds promise for integration into computer-aided diagnostic systems for comprehensive musculoskeletal analysis. These findings are expected to enhance medical image analysis techniques and their applications in clinical decision support systems.

18F-FDG-PET/CT Uptake by Noncancerous Lung as a Predictor of Interstitial Lung Disease Induced by Immune Checkpoint Inhibitors

RATIONALE AND OBJECTIVES

Immune checkpoint inhibitors (ICIs) have improved lung cancer prognosis; however, ICI-related interstitial lung disease (ILD) is fatal and difficult to predict. Herein, we hypothesized that pre-existing lung inflammation on radiological imaging can be a potential risk factor for ILD onset. Therefore, we investigated the association between high uptake in noncancerous lung (NCL) on 18F- FDG-PET/CT and ICI-ILD in lung cancer.

METHODS

Patients with primary lung cancer who underwent FDG-PET/CT within three months prior to ICI therapy were retrospectively included. Artificial intelligence was utilized for extracting the NCL regions (background lung) from the lung contralateral to the primary tumor. FDG uptake by the NCL was assessed via the SUVmax (NCL-SUVmax), SUVmean (NCL-SUVmean), and total glycolytic activity (NCL-TGA)defined as NCL-SUVmean×NCL volume [mL]. NCL-SUVmean and NCL-TGA were calculated using the following four SUV thresholds: 0.5, 1.0, 1.5, and 2.0.

RESULTS

Of the 165 patients, 28 (17.0%) developed ILD. Univariate analysis showed that high values of NCL-SUVmax, NCL-SUVmean2.0 (SUV threshold=2.0), and NCL-TGA1.0 (SUV threshold=1.0) were significantly associated with ILD onset (all p = 0.003). Multivariate analysis adjusted for age, tumor FDG uptake, and pre-existing interstitial lung abnormalities revealed that a high NCL-TGA1.0 (≥149.45) was independently associated with ILD onset (odds ratio, 6.588; p = 0.002). Two-year cumulative incidence of ILD was significantly higher in the high NCL-TGA1.0 group than in the low group (58.4% vs. 14.4%; p < 0.001).

CONCLUSION

High uptake of NCL on FDG-PET/CT is correlated with ICI-ILD development, which could serve as a risk stratification tool before ICI therapy in primary lung cancer.

Addressing the Contrast Media Recognition Challenge: A Fully Automated Machine Learning Approach for Predicting Contrast Phases in CT Imaging

OBJECTIVES

Accurately acquiring and assigning different contrast-enhanced phases in computed tomography (CT) is relevant for clinicians and for artificial intelligence orchestration to select the most appropriate series for analysis. However, this information is commonly extracted from the CT metadata, which is often wrong. This study aimed at developing an automatic pipeline for classifying intravenous (IV) contrast phases and additionally for identifying contrast media in the gastrointestinal tract (GIT).

MATERIALS AND METHODS

This retrospective study used 1200 CT scans collected at the investigating institution between January 4, 2016 and September 12, 2022, and 240 CT scans from multiple centers from The Cancer Imaging Archive for external validation. The open-source segmentation algorithm TotalSegmentator was used to identify regions of interest (pulmonary artery, aorta, stomach, portal/splenic vein, liver, portal vein/hepatic veins, inferior vena cava, duodenum, small bowel, colon, left/right kidney, urinary bladder), and machine learning classifiers were trained with 5-fold cross-validation to classify IV contrast phases (noncontrast, pulmonary arterial, arterial, venous, and urographic) and GIT contrast enhancement. The performance of the ensembles was evaluated using the receiver operating characteristic area under the curve (AUC) and 95% confidence intervals (CIs).

RESULTS

For the IV phase classification task, the following AUC scores were obtained for the internal test set: 99.59% [95% CI, 99.58-99.63] for the noncontrast phase, 99.50% [95% CI, 99.49-99.52] for the pulmonary-arterial phase, 99.13% [95% CI, 99.10-99.15] for the arterial phase, 99.8% [95% CI, 99.79-99.81] for the venous phase, and 99.7% [95% CI, 99.68-99.7] for the urographic phase. For the external dataset, a mean AUC of 97.33% [95% CI, 97.27-97.35] and 97.38% [95% CI, 97.34-97.41] was achieved for all contrast phases for the first and second annotators, respectively. Contrast media in the GIT could be identified with an AUC of 99.90% [95% CI, 99.89-99.9] in the internal dataset, whereas in the external dataset, an AUC of 99.73% [95% CI, 99.71-99.73] and 99.31% [95% CI, 99.27-99.33] was achieved with the first and second annotator, respectively.

CONCLUSIONS

The integration of open-source segmentation networks and classifiers effectively classified contrast phases and identified GIT contrast enhancement using anatomical landmarks.

Deep learning-based fully automatic Risser stage assessment model using abdominal radiographs

BACKGROUND

Artificial intelligence has been increasingly used in medical imaging and has demonstrated expert level performance in image classification tasks.

OBJECTIVE

To develop a fully automatic approach for determining the Risser stage using deep learning on abdominal radiographs.

MATERIALS AND METHODS

In this multicenter study, 1,681 supine abdominal radiographs (age range, 9-18 years, 50% female) obtained between January 2019 and April 2022 were collected retrospectively from three medical institutions and graded manually using the United States Risser staging system. A total of 1,577 images from Hospitals 1 and 2 were used for development, and 104 images from Hospital 3 for external validation. From each radiograph, right and left iliac crest patch images were extracted using the pelvic bone segmentation model DeepLabv3 + with the EfficientNet-B0 encoder trained with 90 digitally reconstructed radiographs from pelvic computed tomography scans with a pelvic bone mask. Using these patch images, ConvNeXt-B was trained to grade according to the Risser classification. The model's performance was evaluated using accuracy, area under the receiver operating characteristic curve (AUROC), and mean absolute error.

RESULTS

The fully automatic Risser stage assessment model showed an accuracy of 0.87 and 0.75, mean absolute error of 0.13 and 0.26, and AUROC of 0.99 and 0.95 on internal and external test sets, respectively.

CONCLUSION

We developed a deep learning-based, fully automatic segmentation and classification model for Risser stage assessment using abdominal radiographs.

Pre-therapy PET-based voxel-wise dosimetry prediction by characterizing intra-organ heterogeneity in PSMA-directed radiopharmaceutical theranostics

BACKGROUND AND OBJECTIVE

Treatment planning through the diagnostic dimension of theranostics provides insights into predicting the absorbed dose of RPT, with the potential to individualize radiation doses for enhancing treatment efficacy. However, existing studies focusing on dose prediction from diagnostic data often rely on organ-level estimations, overlooking intra-organ variations. This study aims to characterize the intra-organ theranostic heterogeneity and utilize artificial intelligence techniques to localize them, i.e. to predict voxel-wise absorbed dose map based on pre-therapy PET.

METHODS

23 patients with metastatic castration-resistant prostate cancer treated with [177Lu]Lu-PSMA I&T RPT were retrospectively included. 48 treatment cycles with pre-treatment PET imaging and at least 3 post-therapeutic SPECT/CT imaging were selected. The distribution of PET tracer and RPT dose was compared for kidney, liver and spleen, characterizing intra-organ heterogeneity differences. Pharmacokinetic simulations were performed to enhance the understanding of the correlation. Two strategies were explored for pre-therapy voxel-wise dosimetry prediction: (1) organ-dose guided direct projection; (2) deep learning (DL)-based distribution prediction. Physical metrics, dose volume histogram (DVH) analysis, and identity plots were applied to investigate the predicted absorbed dose map.

RESULTS

Inconsistent intra-organ patterns emerged between PET imaging and dose map, with moderate correlations existing in the kidney (r = 0.77), liver (r = 0.5), and spleen (r = 0.58) (P < 0.025). Simulation results indicated the intra-organ pharmacokinetic heterogeneity might explain this inconsistency. The DL-based method achieved a lower average voxel-wise normalized root mean squared error of 0.79 ± 0.27%, regarding to ground-truth dose map, outperforming the organ-dose guided projection (1.11 ± 0.57%) (P < 0.05). DVH analysis demonstrated good prediction accuracy (R2 = 0.92 for kidney). The DL model improved the mean slope of fitting lines in identity plots (199% for liver), when compared to the theoretical optimal results of the organ-dose approach.

CONCLUSION

Our results demonstrated the intra-organ heterogeneity of pharmacokinetics may complicate pre-therapy dosimetry prediction. DL has the potential to bridge this gap for pre-therapy prediction of voxel-wise heterogeneous dose map.

Diagnostic efficacy of [68Ga]Ga-DOTA-GPFAPI-04 in patients with solid tumors in a head-to-head comparison with [18F]F-FDG: results from a prospective clinical study

PURPOSE

To identify the biodistribution and diagnostic performance of a novel fibroblast activation protein (FAP) targeted positron emission tomography (PET) tracer, [68Ga]Ga-DOTA-GPFAPI-04, in patients with solid tumors in a head-to-head comparison with [18F]F-FDG.

METHODS

Twenty-six patients histologically proven with cancers of nasopharyngeal (n = 5), esophagus (n = 5), gastro-esophagus (n = 1), stomach (n = 7), liver (n = 3), and colorectum (n = 5) were recruited for [68Ga]Ga-DOTA-GPFAPI-04 and [18F]F-FDG PET/CT scans on consecutive days. The primary endpoint was the diagnostic efficacy, with the histological diagnosis and the follow-up results selected as the gold standard. The secondary endpoint was the background uptake pattern. Two experienced nuclear medicine physicians who were blinded to the gold standard results while having essential awareness of the clinical context reviewed the images and labeled lesions by consensus for subsequent software-assisted lesion segmentation. Additionally, background organs were automatically segmented, assisted by artificial intelligence. Volume, mean, and maximum standard uptake values (SUVmean and SUVmax) of all segmentations were recorded. P < 0.05 was deemed as statistically significant.

RESULTS

Significant glandular uptake of [68Ga]Ga-DOTA-GPFAPI-04 was detected in the thyroid, pancreas, and submandibular glands, while moderate uptake was observed in the parotid glands. The myocardium and myometrium exhibited 2-3 times higher uptake of the radiotracer than that of the background levels in blood and liver. A total of 349 targeted lesions, consisting of 324 malignancies and 25 benign lesions, were segmented. [68Ga]Ga-DOTA-GPFAPI-04 is more sensitive than [18F]F-FDG, especially for abdominopelvic dissemination (1.000 vs. 0.475, P < 0.001). Interestingly, [18F]F-FDG demonstrated higher sensitivity for lung metastasis compared to [68Ga]Ga-DOTA-GPFAPI-04 (0.845 vs. 0.682, P = 0.003). The high glandular uptake made it difficult to delineate lesions in close proximity and masked two metastatic lesions in these organs.

CONCLUSION

Despite prominent glandular uptake, [68Ga]Ga-DOTA-GPFAPI-04 demonstrates favorable diagnostic performance. It is a promising probe scaffold for further development of FAP-targeted tumor theranostic agents.

Deep Learning Segmentation of Ascites on Abdominal CT Scans for Automatic Volume Quantification

Purpose To evaluate the performance of an automated deep learning method in detecting ascites and subsequently quantifying its volume in patients with liver cirrhosis and patients with ovarian cancer. Materials and Methods This retrospective study included contrast-enhanced and noncontrast abdominal-pelvic CT scans of patients with cirrhotic ascites and patients with ovarian cancer from two institutions, National Institutes of Health (NIH) and University of Wisconsin (UofW). The model, trained on The Cancer Genome Atlas Ovarian Cancer dataset (mean age [±SD], 60 years ± 11; 143 female), was tested on two internal datasets (NIH-LC and NIH-OV) and one external dataset (UofW-LC). Its performance was measured by the F1/Dice coefficient, SDs, and 95% CIs, focusing on ascites volume in the peritoneal cavity. Results On NIH-LC (25 patients; mean age, 59 years ± 14; 14 male) and NIH-OV (166 patients; mean age, 65 years ± 9; all female), the model achieved F1/Dice scores of 85.5% ± 6.1 (95% CI: 83.1, 87.8) and 82.6% ± 15.3 (95% CI: 76.4, 88.7), with median volume estimation errors of 19.6% (IQR, 13.2%-29.0%) and 5.3% (IQR: 2.4%-9.7%), respectively. On UofW-LC (124 patients; mean age, 46 years ± 12; 73 female), the model had a F1/Dice score of 83.0% ± 10.7 (95% CI: 79.8, 86.3) and median volume estimation error of 9.7% (IQR, 4.5%-15.1%). The model showed strong agreement with expert assessments, with r2 values of 0.79, 0.98, and 0.97 across the test sets. Conclusion The proposed deep learning method performed well in segmenting and quantifying the volume of ascites in patients with cirrhosis and those with ovarian cancer, in concordance with expert radiologist assessments. Keywords: Abdomen/GI, Cirrhosis, Deep Learning, Segmentation Supplemental material is available for this article. © RSNA, 2024 See also commentary by Aisen and Rodrigues in this issue.

The association of body composition phenotypes before chemotherapy with epithelial ovarian cancer mortality

BACKGROUND

The association of body composition with epithelial ovarian carcinoma (EOC) mortality is poorly understood. To date, evidence suggests that high adiposity is associated with decreased mortality (an obesity paradox), but the impact of muscle on this association has not been investigated. Herein, we define associations of muscle and adiposity joint-exposure body composition phenotypes with EOC mortality.

METHODS

Body composition from 500 women in the Body Composition and Epithelial Ovarian Cancer Survival Study was dichotomized as normal or low skeletal muscle index (SMI), a proxy for sarcopenia, and high or low adiposity. Four phenotypes were classified as fit (normal SMI and low adiposity; reference; 16.2%), overweight or obese (normal SMI and high adiposity; 51.2%), sarcopenia and overweight or obese (low SMI and high adiposity; 15.6%), and sarcopenia or cachexia (low SMI and low adiposity; 17%). We used multivariable Cox models to estimate associations of each phenotype with mortality for EOC overall and high-grade serous ovarian carcinoma (HGSOC).

RESULTS

Overweight or obesity was associated with up to 51% and 104% increased mortality in EOC and HGSOC [Hazard Ratio (HR)] = 1.51, 95% CI = 1.05 to 2.19 and HR = 2.04, 95% CI = 1.29 to 3.21). Sarcopenia and overweight or obesity was associated with up to 66% and 67% increased mortality in EOC and HGSOC (HR = 1.66, 95% CI = 1.13 to 2.45 and HR = 1.67, 95% CI = 1.05 to 2.68). Sarcopenia or cachexia was associated with up to 73% and 109% increased mortality in EOC and HGSOC (HR = 1.73, 95% CI = 1.14 to 2.63 and HR = 2.09, 95% CI = 1.25 to 3.50).

CONCLUSIONS

Overweight or obesity, sarcopenia and overweight or obesity, and sarcopenia or cachexia phenotypes were each associated with increased mortality in EOC and HGSOC. Exercise and dietary interventions could be leveraged as ancillary treatment strategies for improving outcomes in the most fatal gynecological malignancy with no previously established modifiable prognostic factors.

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.