Comparing Commercial and Open-Source Large Language Models for Labeling Chest Radiograph Reports.

Background Rapid advances in large language models (LLMs) have led to the development of numerous commercial and open-source models. While recent publications have explored OpenAI's GPT-4 to extract information of interest from radiology reports, there has not been a real-world comparison of GPT-4 to leading open-source models. Purpose To compare different leading open-source LLMs to GPT-4 on the task of extracting relevant findings from chest radiograph reports. Materials and Methods Two independent datasets of free-text radiology reports from chest radiograph examinations were used in this retrospective study performed between February 2, 2024, and February 14, 2024. The first dataset consisted of reports from the ImaGenome dataset, providing reference standard annotations from the MIMIC-CXR database acquired between 2011 and 2016. The second dataset consisted of randomly selected reports created at the Massachusetts General Hospital between July 2019 and July 2021. In both datasets, the commercial models GPT-3.5 Turbo and GPT-4 were compared with open-source models that included Mistral-7B and Mixtral-8 × 7B (Mistral AI), Llama 2-13B and Llama 2-70B (Meta), and Qwen1.5-72B (Alibaba Group), as well as CheXbert and CheXpert-labeler (Stanford ML Group), in their ability to accurately label the presence of multiple findings in radiograph text reports using zero-shot and few-shot prompting. The McNemar test was used to compare F1 scores between models. Results On the ImaGenome dataset (n = 450), the open-source model with the highest score, Llama 2-70B, achieved micro F1 scores of 0.97 and 0.97 for zero-shot and few-shot prompting, respectively, compared with the GPT-4 F1 scores of 0.98 and 0.98 (P > .99 and < .001 for superiority of GPT-4). On the institutional dataset (n = 500), the open-source model with the highest score, an ensemble model, achieved micro F1 scores of 0.96 and 0.97 for zero-shot and few-shot prompting, respectively, compared with the GPT-4 F1 scores of 0.98 and 0.97 (P < .001 and > .99 for superiority of GPT-4). Conclusion Although GPT-4 was superior to open-source models in zero-shot report labeling, few-shot prompting with a small number of example reports closely matched the performance of GPT-4. The benefit of few-shot prompting varied across datasets and models. © RSNA, 2024 Supplemental material is available for this article.

Diffusion tensor transformation for personalizing target volumes in radiation therapy.

Diffusion tensor imaging (DTI) is used in tumor growth models to provide information on the infiltration pathways of tumor cells into the surrounding brain tissue. When a patient-specific DTI is not available, a template image such as a DTI atlas can be transformed to the patient anatomy using image registration. This study investigates a model, the invariance under coordinate transform (ICT), that transforms diffusion tensors from a template image to the patient image, based on the principle that the tumor growth process can be mapped, at any point in time, between the images using the same transformation function that we use to map the anatomy. The ICT model allows the mapping of tumor cell densities and tumor fronts (as iso-levels of tumor cell density) from the template image to the patient image for inclusion in radiotherapy treatment planning. The proposed approach transforms the diffusion tensors to simulate tumor growth in locally deformed anatomy and outputs the tumor cell density distribution over time. The ICT model is validated in a cohort of ten brain tumor patients. Comparative analysis with the tumor cell density in the original template image shows that the ICT model accurately simulates tumor cell densities in the deformed image space. By creating radiotherapy target volumes as tumor fronts, this study provides a framework for more personalized radiotherapy treatment planning, without the use of additional imaging.

CASCADE: Context-Aware Data-Driven AI for Streamlined Multidisciplinary Tumor Board Recommendations in Oncology.

This study addresses the potential of machine learning in predicting treatment recommendations for patients with hepatocellular carcinoma (HCC). Using an IRB-approved retrospective study of patients discussed at a multidisciplinary tumor board, clinical and imaging variables were extracted and used in a gradient-boosting machine learning algorithm, XGBoost. The algorithm's performance was assessed using confusion matrix metrics and the area under the Receiver Operating Characteristics (ROC) curve. The study included 140 patients (mean age 67.7 ± 8.9 years), and the algorithm was found to be predictive of all eight treatment recommendations made by the board. The model's predictions were more accurate than those based on published therapeutic guidelines by ESMO and NCCN. The study concludes that a machine learning model incorporating clinical and imaging variables can predict treatment recommendations made by an expert multidisciplinary tumor board, potentially aiding clinical decision-making in settings lacking subspecialty expertise.

No code machine learning: validating the approach on use-case for classifying clavicle fractures.

PURPOSE: We created an infrastructure for no code machine learning (NML) platform for non-programming physicians to create NML model. We tested the platform by creating an NML model for classifying radiographs for the presence and absence of clavicle fractures. METHODS: Our IRB-approved retrospective study included 4135 clavicle radiographs from 2039 patients (mean age 52 ± 20 years, F:M 1022:1017) from 13 hospitals. Each patient had two-view clavicle radiographs with axial and anterior-posterior projections. The positive radiographs had either displaced or non-displaced clavicle fractures. We configured the NML platform to automatically retrieve the eligible exams using the series' unique identification from the hospital virtual network archive via web access to DICOM Objects. The platform trained a model until the validation loss plateaus. Once the testing was complete, the platform provided the receiver operating characteristics curve and confusion matrix for estimating sensitivity, specificity, and accuracy. RESULTS: The NML platform successfully retrieved 3917 radiographs (3917/4135, 94.7 %) and parsed them for creating a ML classifier with 2151 radiographs in the training, 100 radiographs for validation, and 1666 radiographs in testing datasets (772 radiographs with clavicle fracture, 894 without clavicle fracture). The network identified clavicle fracture with 90 % sensitivity, 87 % specificity, and 88 % accuracy with AUC of 0.95 (confidence interval 0.94-0.96). CONCLUSION: A NML platform can help physicians create and test machine learning models from multicenter imaging datasets such as the one in our study for classifying radiographs based on the presence of clavicle fracture.

Integrating muscle fiber orientation from visible human data into radiotherapy target volumes.

Objective.A major challenge in treatment of tumors near skeletal muscle is defining the target volume for suspected tumor invasion into the muscle. This study develops a framework that generates radiation target volumes with muscle fiber orientation directly integrated into their definition. The framework is applied to nineteen sacral tumor patients with suspected infiltration into surrounding muscles.Approach.To compensate for the poor soft-tissue contrast of CT images, muscle fiber orientation is derived from cryo-images of two cadavers from the human visible project (VHP). The approach consists of (a) detecting image gradients in the cadaver images representative of muscle fibers, (b) mapping this information onto the patient image, and (c) embedding the muscle fiber orientation into an expansion method to generate patient-specific clinical target volumes (CTV). The validation tested the consistency of image gradient orientation across VHP subjects for the piriformis, gluteus maximus, paraspinal, gluteus medius, and gluteus minimus muscles. The model robustness was analyzed by comparing CTVs generated using different VHP subjects. The difference in shape between the new CTVs and standard CTV was analyzed for clinical impact.Main results.Good agreement was found between the image gradient orientation across VHP subjects, as the voxel-wise median cosine similarity was at least 0.86 (for the gluteus minimus) and up to 0.98 for the piriformis. The volume and surface similarity between the CTVs generating from different VHP subjects was on average at least 0.95 and 5.13 mm for the Dice Similarity Coefficient and the Hausdorff 95% Percentile Index, showing excellent robustness. Finally, compared to the standard CTV with different margins in muscle and non-muscle tissue, the new CTV margins are reduced in muscle tissue depending on the chosen clinical margins.Significance.This study implements a method to integrate muscle fiber orientation into the target volume without the need for additional imaging.

Expert-centered Evaluation of Deep Learning Algorithms for Brain Tumor Segmentation.

Purpose To present results from a literature survey on practices in deep learning segmentation algorithm evaluation and perform a study on expert quality perception of brain tumor segmentation. Materials and Methods A total of 180 articles reporting on brain tumor segmentation algorithms were surveyed for the reported quality evaluation. Additionally, ratings of segmentation quality on a four-point scale were collected from medical professionals for 60 brain tumor segmentation cases. Results Of the surveyed articles, Dice score, sensitivity, and Hausdorff distance were the most popular metrics to report segmentation performance. Notably, only 2.8% of the articles included clinical experts' evaluation of segmentation quality. The experimental results revealed a low interrater agreement (Krippendorff α, 0.34) in experts' segmentation quality perception. Furthermore, the correlations between the ratings and commonly used quantitative quality metrics were low (Kendall tau between Dice score and mean rating, 0.23; Kendall tau between Hausdorff distance and mean rating, 0.51), with large variability among the experts. Conclusion The results demonstrate that quality ratings are prone to variability due to the ambiguity of tumor boundaries and individual perceptual differences, and existing metrics do not capture the clinical perception of segmentation quality. Keywords: Brain Tumor Segmentation, Deep Learning Algorithms, Glioblastoma, Cancer, Machine Learning Clinical trial registration nos. NCT00756106 and NCT00662506 Supplemental material is available for this article. © RSNA, 2023.

From Machine Learning to Patient Outcomes: A Comprehensive Review of AI in Pancreatic Cancer.

Pancreatic cancer is a highly aggressive and difficult-to-detect cancer with a poor prognosis. Late diagnosis is common due to a lack of early symptoms, specific markers, and the challenging location of the pancreas. Imaging technologies have improved diagnosis, but there is still room for improvement in standardizing guidelines. Biopsies and histopathological analysis are challenging due to tumor heterogeneity. Artificial Intelligence (AI) revolutionizes healthcare by improving diagnosis, treatment, and patient care. AI algorithms can analyze medical images with precision, aiding in early disease detection. AI also plays a role in personalized medicine by analyzing patient data to tailor treatment plans. It streamlines administrative tasks, such as medical coding and documentation, and provides patient assistance through AI chatbots. However, challenges include data privacy, security, and ethical considerations. This review article focuses on the potential of AI in transforming pancreatic cancer care, offering improved diagnostics, personalized treatments, and operational efficiency, leading to better patient outcomes.

Enrichment of lung cancer computed tomography collections with AI-derived annotations.

Public imaging datasets are critical for the development and evaluation of automated tools in cancer imaging. Unfortunately, many do not include annotations or image-derived features, complicating downstream analysis. Artificial intelligence-based annotation tools have been shown to achieve acceptable performance and can be used to automatically annotate large datasets. As part of the effort to enrich public data available within NCI Imaging Data Commons (IDC), here we introduce AI-generated annotations for two collections containing computed tomography images of the chest, NSCLC-Radiomics, and a subset of the National Lung Screening Trial. Using publicly available AI algorithms, we derived volumetric annotations of thoracic organs-at-risk, their corresponding radiomics features, and slice-level annotations of anatomical landmarks and regions. The resulting annotations are publicly available within IDC, where the DICOM format is used to harmonize the data and achieve FAIR (Findable, Accessible, Interoperable, Reusable) data principles. The annotations are accompanied by cloud-enabled notebooks demonstrating their use. This study reinforces the need for large, publicly accessible curated datasets and demonstrates how AI can aid in cancer imaging.

Machine learning-based detection of sarcopenic obesity and association with adverse outcomes in patients undergoing surgical treatment for spinal metastases.

OBJECTIVE: The distributions and proportions of lean and fat tissues may help better assess the prognosis and outcomes of patients with spinal metastases. Specifically, in obese patients, sarcopenia may be easily overlooked as a poor prognostic indicator. The role of this body phenotype, sarcopenic obesity (SO), has not been adequately studied among patients undergoing surgical treatment for spinal metastases. To this end, here the authors investigated the role of SO as a potential prognostic factor in patients undergoing surgical treatment for spinal metastases. METHODS: The authors identified patients who underwent surgical treatment for spinal metastases between 2010 and 2020. A validated deep learning approach evaluated sarcopenia and adiposity on routine preoperative CT images. Based on composition analyses, patients were classified with SO or nonsarcopenic obesity. After nearest-neighbor propensity matching that accounted for confounders, the authors compared the rates and odds of postoperative complications, length of stay, 30-day readmission, and all-cause mortality at 90 days and 1 year between the SO and nonsarcopenic obesity groups. RESULTS: A total of 62 patients with obesity underwent surgical treatment for spinal metastases during the study period. Of these, 37 patients had nonsarcopenic obesity and 25 had SO. After propensity matching, 50 records were evaluated that were equally composed of patients with nonsarcopenic obesity and SO (25 patients each). Patients with SO were noted to have increased odds of nonhome discharge (OR 6.0, 95% CI 1.69-21.26), 30-day readmission (OR 3.27, 95% CI 1.01-10.62), and 90-day (OR 4.85, 95% CI 1.29-18.26) and 1-year (OR 3.78, 95% CI 1.17-12.19) mortality, as well as increased time to mortality after surgery (12.60 ± 19.84 months vs 37.16 ± 35.19 months, p = 0.002; standardized mean difference 0.86). No significant differences were noted in terms of length of stay or postoperative complications when comparing the two groups (p > 0.05). CONCLUSIONS: The SO phenotype was associated with increased odds of nonhome discharge, readmission, and postoperative mortality. This study suggests that SO may be an important prognostic factor to consider when developing care plans for patients with spinal metastases.

Not without Context-A Multiple Methods Study on Evaluation and Correction of Automated Brain Tumor Segmentations by Experts.

RATIONALE AND OBJECTIVES: Brain tumor segmentations are integral to the clinical management of patients with glioblastoma, the deadliest primary brain tumor in adults. The manual delineation of tumors is time-consuming and highly provider-dependent. These two problems must be addressed by introducing automated, deep-learning-based segmentation tools. This study aimed to identify criteria experts use to evaluate the quality of automatically generated segmentations and their thought processes as they correct them. MATERIALS AND METHODS: Multiple methods were used to develop a detailed understanding of the complex factors that shape experts' perception of segmentation quality and their thought processes in correcting proposed segmentations. Data from a questionnaire and semistructured interview with neuro-oncologists and neuroradiologists were collected between August and December 2021 and analyzed using a combined deductive and inductive approach. RESULTS: Brain tumors are highly complex and ambiguous segmentation targets. Therefore, physicians rely heavily on the given context related to the patient and clinical context in evaluating the quality and need to correct brain tumor segmentation. Most importantly, the intended clinical application determines the segmentation quality criteria and editing decisions. Physicians' personal beliefs and preferences about the capabilities of AI algorithms and whether questionable areas should not be included are additional criteria influencing the perception of segmentation quality and appearance of an edited segmentation. CONCLUSION: Our findings on experts' perceptions of segmentation quality will allow the design of improved frameworks for expert-centered evaluation of brain tumor segmentation models. In particular, the knowledge presented here can inspire the development of brain tumor-specific metrics for segmentation model training and evaluation.

The influence of anisotropy on the clinical target volume of brain tumor patients.

Objective.Current radiotherapy guidelines for glioma target volume definition recommend a uniform margin expansion from the gross tumor volume (GTV) to the clinical target volume (CTV), assuming uniform infiltration in the invaded brain tissue. However, glioma cells migrate preferentially along white matter tracts, suggesting that white matter directionality should be considered in an anisotropic CTV expansion. We investigate two models of anisotropic CTV expansion and evaluate their clinical feasibility.Approach.To incorporate white matter directionality into the CTV, a diffusion tensor imaging (DTI) atlas is used. The DTI atlas consists of water diffusion tensors that are first spatially transformed into local tumor resistance tensors, also known as metric tensors, and secondly fed to a CTV expansion algorithm to generate anisotropic CTVs. Two models of spatial transformation are considered in the first step. The first model assumes that tumor cells experience reduced resistance parallel to the white matter fibers. The second model assumes that the anisotropy of tumor cell resistance is proportional to the anisotropy observed in DTI, with an 'anisotropy weighting parameter' controlling the proportionality. The models are evaluated in a cohort of ten brain tumor patients.Main results.To evaluate the sensitivity of the model, a library of model-generated CTVs was computed by varying the resistance and anisotropy parameters. Our results indicate that the resistance coefficient had the most significant effect on the global shape of the CTV expansion by redistributing the target volume from potentially less involved gray matter to white matter tissue. In addition, the anisotropy weighting parameter proved useful in locally increasing CTV expansion in regions characterized by strong tissue directionality, such as near the corpus callosum.Significance.By incorporating anisotropy into the CTV expansion, this study is a step toward an interactive CTV definition that can assist physicians in incorporating neuroanatomy into a clinically optimized CTV.

Abnormal vascular structure and function within brain metastases is linked to pembrolizumab resistance.

BACKGROUND: We recently conducted a phase 2 trial (NCT028865685) evaluating intracranial efficacy of pembrolizumab for brain metastases (BM) of diverse histologies. Our study met its primary efficacy endpoint and illustrates that pembrolizumab exerts promising activity in a select group of patients with BM. Given the importance of aberrant vasculature in mediating immunosuppression, we explored the relationship between checkpoint inhibitor (ICI) efficacy and vascular architecture in the hopes of identifying potential mechanisms of intracranial ICI response or resistance for BM. METHODS: Using Vessel Architectural Imaging (VAI), a histologically validated quantitative metric for in vivo tumor vascular physiology, we analyzed dual echo DSC/DCE MRI for 44 patients on trial. Tumor and peri-tumor cerebral blood volume/flow, vessel size, arterial- and venous-dominance, and vascular permeability were measured before and after treatment with pembrolizumab. RESULTS: BM that progressed on ICI were characterized by a highly aberrant vasculature dominated by large-caliber vessels. In contrast, ICI-responsive BM possessed a more structurally balanced vasculature consisting of both small and large vessels, and there was a trend towards a decrease in under-perfused tissue, suggesting a reversal of the negative effects of hypoxia. In the peri-tumor region, development of smaller blood vessels, consistent with neo-angiogenesis, was associated with tumor growth before radiographic evidence of contrast enhancement on anatomical MRI. CONCLUSIONS: This study, one of the largest functional imaging studies for BM, suggests that vascular architecture is linked with ICI efficacy. Studies identifying modulators of vascular architecture, and effects on immune activity, are warranted and may inform future combination treatments.

Structural and functional vascular dysfunction within brain metastases is linked to pembrolizumab inefficacy.

Structurally and functionally aberrant vasculature is a hallmark of tumor angiogenesis and treatment resistance. Given the synergistic link between aberrant tumor vasculature and immunosuppression, we analyzed perfusion MRI for 44 patients with brain metastases (BM) undergoing treatment with pembrolizumab. To date, vascular-immune communication, or the relationship between immune checkpoint inhibitor (ICI) efficacy and vascular architecture, has not been well-characterized in human imaging studies. We found that ICI-responsive BM possessed a structurally balanced vascular makeup, which was linked to improved vascular efficiency and an immune-stimulatory microenvironment. In contrast, ICI-resistant BM were characterized by a lack of immune cell infiltration and a highly aberrant vasculature dominated by large-caliber vessels. Peri-tumor region analysis revealed early functional changes predictive of ICI resistance before radiographic evidence on conventional MRI. This study was one of the largest functional imaging studies for BM and establishes a foundation for functional studies that illuminate the mechanisms linking patterns of vascular architecture with immunosuppression, as targeting these aspects of cancer biology may serve as the basis for future combination treatments.

Body composition and lung cancer-associated cachexia in TRACERx.

Cancer-associated cachexia (CAC) is a major contributor to morbidity and mortality in individuals with non-small cell lung cancer. Key features of CAC include alterations in body composition and body weight. Here, we explore the association between body composition and body weight with survival and delineate potential biological processes and mediators that contribute to the development of CAC. Computed tomography-based body composition analysis of 651 individuals in the TRACERx (TRAcking non-small cell lung Cancer Evolution through therapy (Rx)) study suggested that individuals in the bottom 20th percentile of the distribution of skeletal muscle or adipose tissue area at the time of lung cancer diagnosis, had significantly shorter lung cancer-specific survival and overall survival. This finding was validated in 420 individuals in the independent Boston Lung Cancer Study. Individuals classified as having developed CAC according to one or more features at relapse encompassing loss of adipose or muscle tissue, or body mass index-adjusted weight loss were found to have distinct tumor genomic and transcriptomic profiles compared with individuals who did not develop such features. Primary non-small cell lung cancers from individuals who developed CAC were characterized by enrichment of inflammatory signaling and epithelial-mesenchymal transitional pathways, and differentially expressed genes upregulated in these tumors included cancer-testis antigen MAGEA6 and matrix metalloproteinases, such as ADAMTS3. In an exploratory proteomic analysis of circulating putative mediators of cachexia performed in a subset of 110 individuals from TRACERx, a significant association between circulating GDF15 and loss of body weight, skeletal muscle and adipose tissue was identified at relapse, supporting the potential therapeutic relevance of targeting GDF15 in the management of CAC.

Head CT deep learning model is highly accurate for early infarct estimation.

Non-contrast head CT (NCCT) is extremely insensitive for early (< 3-6 h) acute infarct identification. We developed a deep learning model that detects and delineates suspected early acute infarcts on NCCT, using diffusion MRI as ground truth (3566 NCCT/MRI training patient pairs). The model substantially outperformed 3 expert neuroradiologists on a test set of 150 CT scans of patients who were potential candidates for thrombectomy (60 stroke-negative, 90 stroke-positive middle cerebral artery territory only infarcts), with sensitivity 96% (specificity 72%) for the model versus 61-66% (specificity 90-92%) for the experts; model infarct volume estimates also strongly correlated with those of diffusion MRI (r2 > 0.98). When this 150 CT test set was expanded to include a total of 364 CT scans with a more heterogeneous distribution of infarct locations (94 stroke-negative, 270 stroke-positive mixed territory infarcts), model sensitivity was 97%, specificity 99%, for detection of infarcts larger than the 70 mL volume threshold used for patient selection in several major randomized controlled trials of thrombectomy treatment.

Machine Learning for Adrenal Gland Segmentation and Classification of Normal and Adrenal Masses at CT.

Background Adrenal masses are common, but radiology reporting and recommendations for management can be variable. Purpose To create a machine learning algorithm to segment adrenal glands on contrast-enhanced CT images and classify glands as normal or mass-containing and to assess algorithm performance. Materials and Methods This retrospective study included two groups of contrast-enhanced abdominal CT examinations (development data set and secondary test set). Adrenal glands in the development data set were manually segmented by radiologists. Images in both the development data set and the secondary test set were manually classified as normal or mass-containing. Deep learning segmentation and classification models were trained on the development data set and evaluated on both data sets. Segmentation performance was evaluated with use of the Dice similarity coefficient (DSC), and classification performance with use of sensitivity and specificity. Results The development data set contained 274 CT examinations (251 patients; median age, 61 years; 133 women), and the secondary test set contained 991 CT examinations (991 patients; median age, 62 years; 578 women). The median model DSC on the development test set was 0.80 (IQR, 0.78-0.89) for normal glands and 0.84 (IQR, 0.79-0.90) for adrenal masses. On the development reader set, the median interreader DSC was 0.89 (IQR, 0.78-0.93) for normal glands and 0.89 (IQR, 0.85-0.97) for adrenal masses. Interreader DSC for radiologist manual segmentation did not differ from automated machine segmentation (P = .35). On the development test set, the model had a classification sensitivity of 83% (95% CI: 55, 95) and specificity of 89% (95% CI: 75, 96). On the secondary test set, the model had a classification sensitivity of 69% (95% CI: 58, 79) and specificity of 91% (95% CI: 90, 92). Conclusion A two-stage machine learning pipeline was able to segment the adrenal glands and differentiate normal adrenal glands from those containing masses. © RSNA, 2022 Online supplemental material is available for this article.

Utility of Normalized Body Composition Areas, Derived From Outpatient Abdominal CT Using a Fully Automated Deep Learning Method, for Predicting Subsequent Cardiovascular Events.

BACKGROUND. CT-based body composition (BC) measurements have historically been too resource intensive to analyze for widespread use and have lacked robust comparison with traditional weight metrics for predicting cardiovascular risk. OBJECTIVE. The aim of this study was to determine whether BC measurements obtained from routine CT scans by use of a fully automated deep learning algorithm could predict subsequent cardiovascular events independently from weight, BMI, and additional cardiovascular risk factors. METHODS. This retrospective study included 9752 outpatients (5519 women and 4233 men; mean age, 53.2 years; 890 patients self-reported their race as Black and 8862 self-reported their race as White) who underwent routine abdominal CT at a single health system from January 2012 through December 2012 and who were given no major cardiovascular or oncologic diagnosis within 3 months of undergoing CT. Using publicly available code, fully automated deep learning BC analysis was performed at the L3 vertebral body level to determine three BC areas (skeletal muscle area [SMA], visceral fat area [VFA], and subcutaneous fat area [SFA]). Age-, sex-, and race-normalized reference curves were used to generate z scores for the three BC areas. Subsequent myocardial infarction (MI) or stroke was determined from the electronic medical record. Multivariable-adjusted Cox proportional hazards models were used to determine hazard ratios (HRs) for MI or stroke within 5 years after CT for the three BC area z scores, with adjustment for normalized weight, normalized BMI, and additional cardiovascular risk factors (smoking status, diabetes diagnosis, and systolic blood pressure). RESULTS. In multivariable models, age-, race-, and sex-normalized VFA was associated with subsequent MI risk (HR of highest quartile compared with lowest quartile, 1.31 [95% CI, 1.03-1.67], p = .04 for overall effect) and stroke risk (HR of highest compared with lowest quartile, 1.46 [95% CI, 1.07-2.00], p = .04 for overall effect). In multivariable models, normalized SMA, SFA, weight, and BMI were not associated with subsequent MI or stroke risk. CONCLUSION. VFA derived from fully automated and normalized analysis of abdominal CT examinations predicts subsequent MI or stroke in Black and White patients, independent of traditional weight metrics, and should be considered an adjunct to BMI in risk models. CLINICAL IMPACT. Fully automated and normalized BC analysis of abdominal CT has promise to augment traditional cardiovascular risk prediction models.

Improving the repeatability of deep learning models with Monte Carlo dropout.

The integration of artificial intelligence into clinical workflows requires reliable and robust models. Repeatability is a key attribute of model robustness. Ideal repeatable models output predictions without variation during independent tests carried out under similar conditions. However, slight variations, though not ideal, may be unavoidable and acceptable in practice. During model development and evaluation, much attention is given to classification performance while model repeatability is rarely assessed, leading to the development of models that are unusable in clinical practice. In this work, we evaluate the repeatability of four model types (binary classification, multi-class classification, ordinal classification, and regression) on images that were acquired from the same patient during the same visit. We study the each model's performance on four medical image classification tasks from public and private datasets: knee osteoarthritis, cervical cancer screening, breast density estimation, and retinopathy of prematurity. Repeatability is measured and compared on ResNet and DenseNet architectures. Moreover, we assess the impact of sampling Monte Carlo dropout predictions at test time on classification performance and repeatability. Leveraging Monte Carlo predictions significantly increases repeatability, in particular at the class boundaries, for all tasks on the binary, multi-class, and ordinal models leading to an average reduction of the 95% limits of agreement by 16% points and of the class disagreement rate by 7% points. The classification accuracy improves in most settings along with the repeatability. Our results suggest that beyond about 20 Monte Carlo iterations, there is no further gain in repeatability. In addition to the higher test-retest agreement, Monte Carlo predictions are better calibrated which leads to output probabilities reflecting more accurately the true likelihood of being correctly classified.

Highdicom: a Python Library for Standardized Encoding of Image Annotations and Machine Learning Model Outputs in Pathology and Radiology.

Machine learning (ML) is revolutionizing image-based diagnostics in pathology and radiology. ML models have shown promising results in research settings, but the lack of interoperability between ML systems and enterprise medical imaging systems has been a major barrier for clinical integration and evaluation. The DICOM® standard specifies information object definitions (IODs) and services for the representation and communication of digital images and related information, including image-derived annotations and analysis results. However, the complexity of the standard represents an obstacle for its adoption in the ML community and creates a need for software libraries and tools that simplify working with datasets in DICOM format. Here we present the highdicom library, which provides a high-level application programming interface (API) for the Python programming language that abstracts low-level details of the standard and enables encoding and decoding of image-derived information in DICOM format in a few lines of Python code. The highdicom library leverages NumPy arrays for efficient data representation and ties into the extensive Python ecosystem for image processing and machine learning. Simultaneously, by simplifying creation and parsing of DICOM-compliant files, highdicom achieves interoperability with the medical imaging systems that hold the data used to train and run ML models, and ultimately communicate and store model outputs for clinical use. We demonstrate through experiments with slide microscopy and computed tomography imaging, that, by bridging these two ecosystems, highdicom enables developers and researchers to train and evaluate state-of-the-art ML models in pathology and radiology while remaining compliant with the DICOM standard and interoperable with clinical systems at all stages. To promote standardization of ML research and streamline the ML model development and deployment process, we made the library available free and open-source at https://github.com/herrmannlab/highdicom .

A Fully Automated Deep Learning Pipeline for Multi-Vertebral Level Quantification and Characterization of Muscle and Adipose Tissue on Chest CT Scans.

Body composition on chest CT scans encompasses a set of important imaging biomarkers. This study developed and validated a fully automated analysis pipeline for multi-vertebral level assessment of muscle and adipose tissue on routine chest CT scans. This study retrospectively trained two convolutional neural networks on 629 chest CT scans from 629 patients (55% women; mean age, 67 years ± 10 [standard deviation]) obtained between 2014 and 2017 prior to lobectomy for primary lung cancer at three institutions. A slice-selection network was developed to identify an axial image at the level of the fifth, eighth, and 10th thoracic vertebral bodies. A segmentation network (U-Net) was trained to segment muscle and adipose tissue on an axial image. Radiologist-guided manual-level selection and segmentation generated ground truth. The authors then assessed the predictive performance of their approach for cross-sectional area (CSA) (in centimeters squared) and attenuation (in Hounsfield units) on an independent test set. For the pipeline, median absolute error and intraclass correlation coefficients for both tissues were 3.6% (interquartile range, 1.3%-7.0%) and 0.959-0.998 for the CSA and 1.0 HU (interquartile range, 0.0-2.0 HU) and 0.95-0.99 for median attenuation. This study demonstrates accurate and reliable fully automated multi-vertebral level quantification and characterization of muscle and adipose tissue on routine chest CT scans. Keywords: Skeletal Muscle, Adipose Tissue, CT, Chest, Body Composition Analysis, Convolutional Neural Network (CNN), Supervised Learning Supplemental material is available for this article. © RSNA, 2022.