Improving Generalizability of PET DL Algorithms: List-Mode Reconstructions Improve DOTATATE PET Hepatic Lesion Detection Performance.

Deep learning (DL) algorithms used for DOTATATE PET lesion detection typically require large, well-annotated training datasets. These are difficult to obtain due to low incidence of gastroenteropancreatic neuroendocrine tumors (GEP-NETs) and the high cost of manual annotation. Furthermore, networks trained and tested with data acquired from site specific PET/CT instrumentation, acquisition and processing protocols have reduced performance when tested with offsite data. This lack of generalizability requires even larger, more diverse training datasets. The objective of this study is to investigate the feasibility of improving DL algorithm performance by better matching the background noise in training datasets to higher noise, out-of-domain testing datasets. 68Ga-DOTATATE PET/CT datasets were obtained from two scanners: Scanner1, a state-of-the-art digital PET/CT (GE DMI PET/CT; n = 83 subjects), and Scanner2, an older-generation analog PET/CT (GE STE; n = 123 subjects). Set1, the data set from Scanner1, was reconstructed with standard clinical parameters (5 min; Q.Clear) and list-mode reconstructions (VPFXS 2, 3, 4, and 5-min). Set2, data from Scanner2 representing out-of-domain clinical scans, used standard iterative reconstruction (5 min; OSEM). A deep neural network was trained with each dataset: Network1 for Scanner1 and Network2 for Scanner2. DL performance (Network1) was tested with out-of-domain test data (Set2). To evaluate the effect of training sample size, we tested DL model performance using a fraction (25%, 50% and 75%) of Set1 for training. Scanner1, list-mode 2-min reconstructed data demonstrated the most similar noise level compared that of Set2, resulting in the best performance (F1 = 0.713). This was not significantly different compared to the highest performance, upper-bound limit using in-domain training for Network2 (F1 = 0.755; p-value = 0.103). Regarding sample size, the F1 score significantly increased from 25% training data (F1 = 0.478) to 100% training data (F1 = 0.713; p < 0.001). List-mode data from modern PET scanners can be reconstructed to better match the noise properties of older scanners. Using existing data and their associated annotations dramatically reduces the cost and effort in generating these datasets and significantly improves the performance of existing DL algorithms. List-mode reconstructions can provide an efficient, low-cost method to improve DL algorithm generalizability.

Model-agnostic explanations for survival prediction models.

Advanced machine learning methods capable of capturing complex and nonlinear relationships can be used in biomedical research to accurately predict time-to-event outcomes. However, these methods have been criticized as "black boxes" that are not interpretable and thus are difficult to trust in making important clinical decisions. Explainable machine learning proposes the use of model-agnostic explainers that can be applied to predictions from any complex model. These explainers describe how a patient's characteristics are contributing to their prediction, and thus provide insight into how the model is arriving at that prediction. The specific application of these explainers to survival prediction models can be used to obtain explanations for (i) survival predictions at particular follow-up times, and (ii) a patient's overall predicted survival curve. Here, we present a model-agnostic approach for obtaining these explanations from any survival prediction model. We extend the local interpretable model-agnostic explainer framework for classification outcomes to survival prediction models. Using simulated data, we assess the performance of the proposed approaches under various settings. We illustrate application of the new methodology using prostate cancer data.

Feasibility and limitations of using commercial databases to evaluate residential mobility in registry-based research on childhood cancer.

BACKGROUND: Researchers have used commercial databases containing residential addresses to reduce exposure misclassification in case-control studies. Our objective is to evaluate the potential systematic bias regarding case status when reconstructing residential locations from commercial databases. METHODS: Our study population of 3640 Colorado-born children includes 520 children diagnosed with acute lymphocytic leukemia between 2002 and 2019. We aligned addresses and date ranges obtained from LexisNexis with registry dates to determine three dichotomous outcomes: Found in LexisNexis, conception date found in LexisNexis, and reference date/diagnosis date found in LexisNexis. We applied logistic regression to determine whether outcomes differed by case status. RESULTS: Mothers of cases were 39% more likely to be found in LexisNexis than mothers of controls (OR = 1.39, 95% CI: 0.97, 2). Of the mothers found in LexisNexis, a conception address was 33% more likely (OR= 1.33, 95% CI: 1.06, 1.66) and a reference/diagnosis address was 60% more likely (OR= 1.60, 95% CI: 1.21, 2.12) to be found for mothers of cases than mothers of controls. CONCLUSION: This study indicates that use of commercial databases to reconstruct residential locations may systematically bias results in case-control studies of childhood cancers.

SpaceANOVA: Spatial Co-occurrence Analysis of Cell Types in Multiplex Imaging Data Using Point Process and Functional ANOVA.

Multiplex imaging platforms have enabled the identification of the spatial organization of different types of cells in complex tissue or the tumor microenvironment. Exploring the potential variations in the spatial co-occurrence or colocalization of different cell types across distinct tissue or disease classes can provide significant pathological insights, paving the way for intervention strategies. However, the existing methods in this context either rely on stringent statistical assumptions or suffer from a lack of generalizability. We present a highly powerful method to study differential spatial co-occurrence of cell types across multiple tissue or disease groups, based on the theories of the Poisson point process and functional analysis of variance. Notably, the method accommodates multiple images per subject and addresses the problem of missing tissue regions, commonly encountered due to data-collection complexities. We demonstrate the superior statistical power and robustness of the method in comparison with existing approaches through realistic simulation studies. Furthermore, we apply the method to three real data sets on different diseases collected using different imaging platforms. In particular, one of these data sets reveals novel insights into the spatial characteristics of various types of colorectal adenoma.

Location-Aware Encoding for Lesion Detection in 68Ga-DOTATATE Positron Emission Tomography Images.

OBJECTIVE: Lesion detection with positron emission tomography (PET) imaging is critical for tumor staging, treatment planning, and advancing novel therapies to improve patient outcomes, especially for neuroendocrine tumors (NETs). Current lesion detection methods often require manual cropping of regions/volumes of interest (ROIs/VOIs) a priori, or rely on multi-stage, cascaded models, or use multi-modality imaging to detect lesions in PET images. This leads to significant inefficiency, high variability and/or potential accumulative errors in lesion quantification. To tackle this issue, we propose a novel single-stage lesion detection method using only PET images. METHODS: We design and incorporate a new, plug-and-play codebook learning module into a U-Net-like neural network and promote lesion location-specific feature learning at multiple scales. We explicitly regularize the codebook learning with direct supervision at the network's multi-level hidden layers and enforce the network to learn multi-scale discriminative features with respect to predicting lesion positions. The network automatically combines the predictions from the codebook learning module and other layers via a learnable fusion layer. RESULTS: We evaluate the proposed method on a real-world clinical 68Ga-DOTATATE PET image dataset, and our method produces significantly better lesion detection performance than recent state-of-the-art approaches. CONCLUSION: We present a novel deep learning method for single-stage lesion detection in PET imaging data, with no ROI/VOI cropping in advance, no multi-stage modeling and no multi-modality data. SIGNIFICANCE: This study provides a new perspective for effective and efficient lesion identification in PET, potentially accelerating novel therapeutic regimen development for NETs and ultimately improving patient outcomes including survival.

Learning Without Real Data Annotations to Detect Hepatic Lesions in PET Images.

OBJECTIVE: Deep neural networks have been recently applied to lesion identification in fluorodeoxyglucose (FDG) positron emission tomography (PET) images, but they typically rely on a large amount of well-annotated data for model training. This is extremely difficult to achieve for neuroendocrine tumors (NETs), because of low incidence of NETs and expensive lesion annotation in PET images. The objective of this study is to design a novel, adaptable deep learning method, which uses no real lesion annotations but instead low-cost, list mode-simulated data, for hepatic lesion detection in real-world clinical NET PET images. METHODS: We first propose a region-guided generative adversarial network (RG-GAN) for lesion-preserved image-to-image translation. Then, we design a specific data augmentation module for our list-mode simulated data and incorporate this module into the RG-GAN to improve model training. Finally, we combine the RG-GAN, the data augmentation module and a lesion detection neural network into a unified framework for joint-task learning to adaptatively identify lesions in real-world PET data. RESULTS: The proposed method outperforms recent state-of-the-art lesion detection methods in real clinical 68Ga-DOTATATE PET images, and produces very competitive performance with the target model that is trained with real lesion annotations. CONCLUSION: With RG-GAN modeling and specific data augmentation, we can obtain good lesion detection performance without using any real data annotations. SIGNIFICANCE: This study introduces an adaptable deep learning method for hepatic lesion identification in NETs, which can significantly reduce human effort for data annotation and improve model generalizability for lesion detection with PET imaging.

Aromatase and steroid sulfatase from human placenta.

Cytochrome P450 aromatase (AROM) and steroid (estrone (E1)/dehydroepiandrosterone (DHEA)) sulfatase (STS) are the two key enzymes responsible for the biosynthesis of estrogens in human, and maintenance of the critical balance between androgens and estrogens. Human AROM, an integral membrane protein of the endoplasmic reticulum, is a member of the Fe-heme containing cytochrome P450 superfamily having a cysteine thiolate as the fifth Fe-coordinating ligand. It is the only enzyme known to catalyze the conversion of androgens with non-aromatic A-rings to estrogens characterized by the aromatic A-ring. Human STS, also an integral membrane protein of the endoplasmic reticulum, is a Ca2+-dependent enzyme that catalyzes the hydrolysis of sulfate esters of E1 and DHEA to yield the respective unconjugated steroids, the precursors of the most potent forms of estrogens and androgens, namely, 17ß-estradiol (E2), 16α,17ß-estriol (E3), testosterone (TST) and dihydrotestosterone (DHT). Expression of these steroidogenic enzymes locally within various organs and tissues of the endocrine, reproductive, and central nervous systems is the key for maintaining high levels of the reproductive steroids. Thus, the enzymes have been drug targets for the prevention and treatment of diseases associated with steroid hormone excesses, especially in breast and prostate malignancies and endometriosis. Both AROM and STS have been the subjects of vigorous research for the past six decades. In this article, we review the procedures of their extraction and purification from human term placenta are described in detail, along with the activity assays.

TreeKernel: interpretable kernel machine tests for interactions between -omics and clinical predictors with applications to metabolomics and COPD phenotypes.

BACKGROUND: In this paper, we are interested in interactions between a high-dimensional -omics dataset and clinical covariates. The goal is to evaluate the relationship between a phenotype of interest and a high-dimensional omics pathway, where the effect of the omics data depends on subjects' clinical covariates (age, sex, smoking status, etc.). For instance, metabolic pathways can vary greatly between sexes which may also change the relationship between certain metabolic pathways and a clinical phenotype of interest. We propose partitioning the clinical covariate space and performing a kernel association test within those partitions. To illustrate this idea, we focus on hierarchical partitions of the clinical covariate space and kernel tests on metabolic pathways. RESULTS: We see that our proposed method outperforms competing methods in most simulation scenarios. It can identify different relationships among clinical groups with higher power in most scenarios while maintaining a proper Type I error rate. The simulation studies also show a robustness to the grouping structure within the clinical space. We also apply the method to the COPDGene study and find several clinically meaningful interactions between metabolic pathways, the clinical space, and lung function. CONCLUSION: TreeKernel provides a simple and interpretable process for testing for relationships between high-dimensional omics data and clinical outcomes in the presence of interactions within clinical cohorts. The method is broadly applicable to many studies.

FunSpace: A functional and spatial analytic approach to cell imaging data using entropy measures.

Spatial heterogeneity in the tumor microenvironment (TME) plays a critical role in gaining insights into tumor development and progression. Conventional metrics typically capture the spatial differential between TME cellular patterns by either exploring the cell distributions in a pairwise fashion or aggregating the heterogeneity across multiple cell distributions without considering the spatial contribution. As such, none of the existing approaches has fully accounted for the simultaneous heterogeneity caused by both cellular diversity and spatial configurations of multiple cell categories. In this article, we propose an approach to leverage spatial entropy measures at multiple distance ranges to account for the spatial heterogeneity across different cellular organizations. Functional principal component analysis (FPCA) is applied to estimate FPC scores which are then served as predictors in a Cox regression model to investigate the impact of spatial heterogeneity in the TME on survival outcome, potentially adjusting for other confounders. Using a non-small cell lung cancer dataset (n = 153) as a case study, we found that the spatial heterogeneity in the TME cellular composition of CD14+ cells, CD19+ B cells, CD4+ and CD8+ T cells, and CK+ tumor cells, had a significant non-zero effect on the overall survival (p = 0.027). Furthermore, using a publicly available multiplexed ion beam imaging (MIBI) triple-negative breast cancer dataset (n = 33), our proposed method identified a significant impact of cellular interactions between tumor and immune cells on the overall survival (p = 0.046). In simulation studies under different spatial configurations, the proposed method demonstrated a high predictive power by accounting for both clinical effect and the impact of spatial heterogeneity.

SpaceANOVA: Spatial co-occurrence analysis of cell types in multiplex imaging data using point process and functional ANOVA.

Motivation: Multiplex imaging platforms have enabled the identification of the spatial organization of different types of cells in complex tissue or tumor microenvironment (TME). Exploring the potential variations in the spatial co-occurrence or co-localization of different cell types across distinct tissue or disease classes can provide significant pathological insights, paving the way for intervention strategies. However, the existing methods in this context either rely on stringent statistical assumptions or suffer from a lack of generalizability. Results: We present a highly powerful method to study differential spatial co-occurrence of cell types across multiple tissue or disease groups, based on the theories of the Poisson point process (PPP) and functional analysis of variance (FANOVA). Notably, the method accommodates multiple images per subject and addresses the problem of missing tissue regions, commonly encountered in such a context due to the complex nature of the data-collection procedure. We demonstrate the superior statistical power and robustness of the method in comparison to existing approaches through realistic simulation studies. Furthermore, we apply the method to three real datasets on different diseases collected using different imaging platforms. In particular, one of these datasets reveals novel insights into the spatial characteristics of various types of precursor lesions associated with colorectal cancer. Availability: The associated R package can be found here, https://github.com/sealx017/SpaceANOVA.

Structures and functions of human placental aromatase and steroid sulfatase, two key enzymes in estrogen biosynthesis.

Cytochrome P450 aromatase (AROM) and steroid sulfatase (STS) are the two key enzymes for the biosynthesis of estrogens in human, and maintenance of the critical balance between androgens and estrogens. Human AROM, an integral membrane protein of the endoplasmic reticulum, is a member of the cytochrome P450 superfamily. It is the only enzyme to catalyze the conversion of androgens with non-aromatic A-rings to estrogens characterized by the aromatic A-ring. Human STS, also an integral membrane protein of the endoplasmic reticulum, is a Ca2+-dependent enzyme that catalyzes the hydrolysis of sulfate esters of estrone and dehydroepiandrosterone to the unconjugated steroids, the precursors of the most potent forms of estrogens and androgens, namely, 17ß-estradiol, 16α,17ß-estriol, testosterone and dihydrotestosterone. Expression of these steroidogenic enzymes locally within organs and tissues of the endocrine, reproductive, and central nervous systems is the key for maintaining high levels of the reproductive steroids. The enzymes have been drug targets for the prevention and treatment of diseases associated with steroid hormone excesses, especially in breast, endometrial and prostate malignancies. Both enzymes have been the subjects of vigorous research for the past six decades. In this article, we review the important findings on their structure-function relationships, specifically, the work that began with unravelling of the closely guarded secrets, namely, the 3-D structures, active sites, mechanisms of action, origins of substrate specificity and the basis of membrane integration. Remarkably, these studies were conducted on the enzymes purified in their pristine forms from human placenta, the discarded and their most abundant source. The purification, assay, crystallization, and structure determination methodologies are described. Also reviewed are their functional quaternary organizations, post-translational modifications and the advancements made in the structure-guided inhibitor design efforts. Outstanding questions that still remain open are summarized in closing.

Dysregulated Lymphocyte Antigen Receptor Signaling in Common Variable Immunodeficiency with Granulomatous Lymphocytic Interstitial Lung Disease.

PURPOSE: A subset of common variable immunodeficiency (CVID) patients either presents with or develops autoimmune and lymphoproliferative complications, such as granulomatous lymphocytic interstitial lung disease (GLILD), a major cause of morbidity and mortality in CVID. While a myriad of phenotypic lymphocyte derangements has been associated with and described in GLILD, defects in T and B cell antigen receptor (TCR/BCR) signaling in CVID and CVID with GLILD (CVID/GLILD) remain undefined, hindering discovery of biomarkers for disease monitoring, prognostic prediction, and personalized medicine approaches. METHODS: To identify perturbations of immune cell subsets and TCR/BCR signal transduction, we applied mass cytometry analysis to peripheral blood mononuclear cells (PBMCs) from healthy control participants (HC), CVID, and CVID/GLILD patients. RESULTS: Patients with CVID, regardless of GLILD status, had increased frequency of HLADR+CD4+ T cells, CD57+CD8+ T cells, and CD21lo B cells when compared to healthy controls. Within these cellular populations in CVID/GLILD patients only, engagement of T or B cell antigen receptors resulted in discordant downstream signaling responses compared to CVID. In CVID/GLILD patients, CD21lo B cells showed perturbed BCR-mediated phospholipase C gamma and extracellular signal-regulated kinase activation, while HLADR+CD4+ T cells and CD57+CD8+ T cells displayed disrupted TCR-mediated activation of kinases most proximal to the receptor. CONCLUSION: Both CVID and CVID/GLILD patients demonstrate an activated T and B cell phenotype compared to HC. However, only CVID/GLILD patients exhibit altered TCR/BCR signaling in the activated lymphocyte subsets. These findings contribute to our understanding of the mechanisms of immune dysregulation in CVID with GLILD.

Uncertainty-Aware Deep Learning Classification of Adamantinomatous Craniopharyngioma from Preoperative MRI.

Diagnosis of adamantinomatous craniopharyngioma (ACP) is predominantly determined through invasive pathological examination of a neurosurgical biopsy specimen. Clinical experts can distinguish ACP from Magnetic Resonance Imaging (MRI) with an accuracy of 86%, and 9% of ACP cases are diagnosed this way. Classification using deep learning (DL) provides a solution to support a non-invasive diagnosis of ACP through neuroimaging, but it is still limited in implementation, a major reason being the lack of predictive uncertainty representation. We trained and tested a DL classifier on preoperative MRI from 86 suprasellar tumor patients across multiple institutions. We then applied a Bayesian DL approach to calibrate our previously published ACP classifier, extending beyond point-estimate predictions to predictive distributions. Our original classifier outperforms random forest and XGBoost models in classifying ACP. The calibrated classifier underperformed our previously published results, indicating that the original model was overfit. Mean values of the predictive distributions were not informative regarding model uncertainty. However, the variance of predictive distributions was indicative of predictive uncertainty. We developed an algorithm to incorporate predicted values and the associated uncertainty to create a classification abstention mechanism. Our model accuracy improved from 80.8% to 95.5%, with a 34.2% abstention rate. We demonstrated that calibration of DL models can be used to estimate predictive uncertainty, which may enable clinical translation of artificial intelligence to support non-invasive diagnosis of brain tumors in the future.

Structure of human placental steroid sulfatase at 2.0 angstrom resolution: Catalysis, quaternary association, and a secondary ligand site.

Human placental estrone (E1)/dehydroepiandrosterone (DHEA) sulfatase (human placental steroid sulfatase; hSTS) is an integral membrane protein of the endoplasmic reticulum. This Ca2+-dependent enzyme catalyzes the hydrolysis of sulfate esters of E1 and DHEA to yield the respective unconjugated steroids, which then act as precursors for the biosynthesis of 17ß-estradiol (E2) and dihydrotestosterone (DHT), respectively, the most potent forms of estrogens and androgens. hSTS is a key enzyme for the local production of E2 and DHT in the breast and the prostate. The enzyme is known to be responsible for maintaining high levels of estrogens in the breast tumor cells. The crystal structure of hSTS purified from human placenta has previously been reported at 2.6 Å resolution. Here we present the structure of hSTS determined at the superior 2.0 Å resolution bringing new clarity to the atomic architecture of the active site. The molecular basis of catalysis and steroid-protein interaction are revisited in light of the new data. We also reexamine the enzyme's quaternary association and its implication on the membrane integration. A secondary ligand binding pocket at the intermolecular interface and adjacent to the active site access channel, buried into the gill of the mushroom-shaped molecule, has been identified. Its role as well as that of a phosphate ion bound to an exposed histidine side chain are examined from the structure-function perspective. Higher resolution data also aids in the tracing of an important loop missing in the previous structure.

Spike-Associated Networks Predict Postsurgical Outcomes in Children With Refractory Epilepsy.

PURPOSE: Up to half of the children undergoing epilepsy surgery will continue to have seizures (szs) despite a cortical resection or ablation. Functional connectivity has shown promise in better identifying the epileptogenic zone. We hypothesized that cortical areas showing high information outflow during interictal epileptiform discharges are part of the epileptogenic zone. METHODS: We identified 22 children with focal epilepsy who had undergone stereo electroencephalography, surgical resection or ablation, and had ≥1 year of postsurgical follow-up. The mean phase slope index, a directed measure of functional connectivity, was calculated for each electrode contact during interictal epileptiform discharges. The positive predictive value and negative predictive value for a sz-free outcome were calculated based on whether high information outflow brain regions were resected. RESULTS: Resection of high outflow (z-score ≥ 1) and very high outflow (z-score ≥ 2) electrode contacts was associated with higher sz freedom (high outflow: χ 2 statistic = 59.1; P < 0.001; very high outflow: χ 2 statistic = 31.3; P < 0.001). The positive predictive value and negative predictive value for sz freedom based on resection at the electrode level increased at higher z-score thresholds with a peak positive predictive value of 0.86 and a peak negative predictive value of 0.9. CONCLUSIONS: Better identification of the epileptogenic zone has the potential to improve epilepsy surgery outcomes. If the surgical plan can be modified to include these very high outflow areas, more children might achieve sz freedom. Conversely, if deficits from resecting these areas are unacceptable, ineffective surgeries could be avoided and alternative therapies offered.

Transportability Methods for Time-to-Event Outcomes: Application in Adjuvant Colon Cancer Trials.

PURPOSE: Differences in the benefits of treatment on 5-year overall survival have been observed in 12 randomized phase III colon cancer adjuvant clinical trials from the ACCENT group. We investigated the reasons for these differences by incorporating the distribution of the observed covariates from each trial. MATERIALS AND METHODS: We applied state-of-the-art transportability methods on the basis of causal inference, and compared them with a conventional meta-analysis approach to predict the treatment effect for the target population. Prediction errors were defined to evaluate whether the identifiability conditions necessary for causal inference were satisfied among the 12 trials, and to measure the performance of each method. RESULTS: In the one-trial-at-a-time transportability analysis, the ranks of prediction errors for the target population were mostly consistent with the discrepancy in treatment effects among the 12 trials across the three models. The overall prediction errors between the leave-one-trial-out transportability method and the conventional individual participant data meta-analysis approach were very similar, and more than 40% lower than the overall prediction errors from the one-trial-at-a-time transportability method. CONCLUSION: The discrepancy in treatment effects among the 12 trials is unlikely to arise from the choice of model specification or distribution of observed covariates but from the distribution of unobserved covariates or study-level features. The ability to quantify heterogeneity among the 12 trials was greatly reduced in both the leave-one-trial-out transportability method and the conventional meta-analysis approach compared with the one-trial-at-a-time transportability method.

Population-level and individual-level explainers for propensity score matching in observational studies.

Precis: The exclusion of unmatched observations in propensity score matching has implications for the generalizability of causal effects. Machine learning methods can help to identify how the study population differs from the unmatched subpopulation. Background: There has been widespread use of propensity scores in evaluating the effect of cancer treatments on survival, particularly in administrative databases and cancer registries. A byproduct of certain matching schemes is the exclusion of observations. Borrowing an analogy from clinical trials, one can view these exclusions as subjects that do not satisfy eligibility criteria. Methods: Developing identification rules for these "data-driven eligibility criteria" in observational studies on both population and individual levels helps to ascertain the population on which causal effects are being made. This article presents a machine learning method to determine the representativeness of causal effects in two different datasets from the National Cancer Database. Results: Decision trees reveal that groups with certain features have a higher probability of inclusion in the study population than older patients. In the first dataset, younger age categories had an inclusion probability of at least 0.90 in all models, while the probability for the older category ranged from 0.47 to 0.65. Most trees split once more on an even higher age at a lower node, suggesting that the oldest patients are the least likely to be matched. In the second set of data, both age and surgery status were associated with inclusion. Conclusion: The methodology presented in this paper underscores the need to consider exclusions in propensity score matching procedures as well as complementing matching with other propensity score adjustments.

Small cell carcinoma of the ovary hypercalcemic type (SCCOHT): A review and novel case with dual germline SMARCA4 and BRCA2 mutations.

Small cell carcinoma of the ovary hypercalcemic type (SCCOHT) is a rare and aggressive disease. While classically linked to mutations in SMARCA4, we describe a case in a patient with both SMARCA4 and BRCA2 germline mutations. We describe her disease presentation, histopathology and treatment with adjuvant systemic chemotherapy, interval hyperthermic intraperitoneal chemotherapy, high dose chemotherapy with stem cell rescue, and maintenance with a poly-ADP-ribose polymerase inhibitor (PARPi). Additionally, we share spatial transcriptomics completed on original tumor.

Causal Inference in Radiomics: Framework, Mechanisms, and Algorithms.

The widespread use of machine learning algorithms in radiomics has led to a proliferation of flexible prognostic models for clinical outcomes. However, a limitation of these techniques is their black-box nature, which prevents the ability for increased mechanistic phenomenological understanding. In this article, we develop an inferential framework for estimating causal effects with radiomics data. A new challenge is that the exposure of interest is latent so that new estimation procedures are needed. We leverage a multivariate version of partial least squares for causal effect estimation. The methodology is illustrated with applications to two radiomics datasets, one in osteosarcoma and one in glioblastoma.

A Pipeline for the Implementation and Visualization of Explainable Machine Learning for Medical Imaging Using Radiomics Features.

Machine learning (ML) models have been shown to predict the presence of clinical factors from medical imaging with remarkable accuracy. However, these complex models can be difficult to interpret and are often criticized as "black boxes". Prediction models that provide no insight into how their predictions are obtained are difficult to trust for making important clinical decisions, such as medical diagnoses or treatment. Explainable machine learning (XML) methods, such as Shapley values, have made it possible to explain the behavior of ML algorithms and to identify which predictors contribute most to a prediction. Incorporating XML methods into medical software tools has the potential to increase trust in ML-powered predictions and aid physicians in making medical decisions. Specifically, in the field of medical imaging analysis the most used methods for explaining deep learning-based model predictions are saliency maps that highlight important areas of an image. However, they do not provide a straightforward interpretation of which qualities of an image area are important. Here, we describe a novel pipeline for XML imaging that uses radiomics data and Shapley values as tools to explain outcome predictions from complex prediction models built with medical imaging with well-defined predictors. We present a visualization of XML imaging results in a clinician-focused dashboard that can be generalized to various settings. We demonstrate the use of this workflow for developing and explaining a prediction model using MRI data from glioma patients to predict a genetic mutation.