An Interpretable Deep Hierarchical Semantic Convolutional Neural Network for Lung Nodule Malignancy Classification.

While deep learning methods have demonstrated performance comparable to human readers in tasks such as computer-aided diagnosis, these models are difficult to interpret, do not incorporate prior domain knowledge, and are often considered as a "black-box." The lack of model interpretability hinders them from being fully understood by end users such as radiologists. In this paper, we present a novel interpretable deep hierarchical semantic convolutional neural network (HSCNN) to predict whether a given pulmonary nodule observed on a computed tomography (CT) scan is malignant. Our network provides two levels of output: 1) low-level semantic features; and 2) a high-level prediction of nodule malignancy. The low-level outputs reflect diagnostic features often reported by radiologists and serve to explain how the model interprets the images in an expert-interpretable manner. The information from these low-level outputs, along with the representations learned by the convolutional layers, are then combined and used to infer the high-level output. This unified architecture is trained by optimizing a global loss function including both low- and high-level tasks, thereby learning all the parameters within a joint framework. Our experimental results using the Lung Image Database Consortium (LIDC) show that the proposed method not only produces interpretable lung cancer predictions but also achieves significantly better results compared to using a 3D CNN alone.

Percutaneous Cryoablation for the Treatment of Recurrent Malignant Pleural Mesothelioma: Safety, Early-Term Efficacy, and Predictors of Local Recurrence.

PURPOSE: To determine safety and early-term efficacy of CT-guided cryoablation for treatment of recurrent mesothelioma and assess risk factors for local recurrence. MATERIALS AND METHODS: During the period 2008-2012, 24 patients underwent 110 cryoablations for recurrent mesothelioma tumors in 89 sessions. Median patient age was 69 years (range, 48-82 y). Median tumor size was 30 mm (range, 9-113 mm). Complications were graded using Common Terminology Criteria for Adverse Events version 4.0 (CTCAE v4.0). Recurrence was diagnosed on CT or positron emission tomography/CT by increasing size, nodular enhancement, or hypermetabolic activity and analyzed using the Kaplan-Meier method. Cox proportional hazards model was used to determine covariates associated with local tumor recurrence. RESULTS: Median duration of follow-up was 14.5 months. Complications occurred in 8 of 110 cryoablations (7.3%). All but 1 complication were graded CTCAE v4.0 1 or 2. No procedure-related deaths occurred. Freedom from local recurrence was observed in 100% of cases at 30 days, 92.5% at 6 months, 90.8% at 1 year, 87.3% at 2 years, and 73.7% at 3 years. Tumor recurrence was diagnosed 4.5-24.5 months after cryoablation (mean 5.7 months). Risk of tumor recurrence was associated with a smaller ablative margin from the edge of tumor to iceball ablation margin (multivariate hazard ratio 0.68, CI 0.48-0.95, P = .024). CONCLUSIONS: CT-guided cryoablation is safe for local control of recurrent mesothelioma, with a low rate of complications and promising early-term efficacy. A smaller ablative margin may predispose to tumor recurrence.

Percutaneous cryoablation for the treatment of recurrent thymoma: preliminary safety and efficacy.

Thymoma is the most common primary tumor of the anterior mediastinum and often recurs after initial surgical resection. In this case series, percutaneous cryoablation, a locally ablative technique, was used to treat 25 mediastinal and pleural recurrent thymoma lesions in five patients. Safety and short-term efficacy data were collected. In 23 percutaneous cryoablations (92%), there were no or minimal complications. One serious complication, myasthenia gravis flare, occurred. Over the duration of follow-up (median, 331 d), 18 of 20 ablated lesions (90%) showed no evidence of local recurrence. Percutaneous cryoablation shows promise as a safe and effective treatment modality for recurrent thymoma.

EDICNet: An end-to-end detection and interpretable malignancy classification network for pulmonary nodules in computed tomography.

We present an interpretable end-to-end computer-aided detection and diagnosis tool for pulmonary nodules on computed tomography (CT) using deep learning-based methods. The proposed network consists of a nodule detector and a nodule malignancy classifier. We used RetinaNet to train a nodule detector using 7,607 slices containing 4,234 nodule annotations and validated it using 2,323 slices containing 1,454 nodule annotations drawn from the LIDC-IDRI dataset. The average precision for the nodule class in the validation set reached 0.24 at an intersection over union (IoU) of 0.5. The trained nodule detector was externally validated using a UCLA dataset. We then used a hierarchical semantic convolutional neural network (HSCNN) to classify whether a nodule was benign or malignant and generate semantic (radiologist-interpretable) features (e.g., mean diameter, consistency, margin), training the model on 149 cases with diagnostic CTs collected from the same UCLA dataset. A total of 149 nodule-centered patches from the UCLA dataset were used to train the HSCNN. Using 5-fold cross validation and data augmentation, the mean AUC and mean accuracy in the validation set for predicting nodule malignancy achieved 0.89 and 0.74, respectively. Meanwhile, the mean accuracy for predicting nodule mean diameter, consistency, and margin were 0.59, 0.74, and 0.75, respectively. We have developed an initial end-to-end pipeline that automatically detects nodules ≥ 5 mm on CT studies and labels identified nodules with radiologist-interpreted features automatically.

Generating Reward Functions Using IRL Towards Individualized Cancer Screening.

Cancer screening can benefit from individualized decision-making tools that decrease overdiagnosis. The heterogeneity of cancer screening participants advocates the need for more personalized methods. Partially observable Markov decision processes (POMDPs), when defined with an appropriate reward function, can be used to suggest optimal, individualized screening policies. However, determining an appropriate reward function can be challenging. Here, we propose the use of inverse reinforcement learning (IRL) to form rewards functions for lung and breast cancer screening POMDPs. Using experts (physicians) retrospective screening decisions for lung and breast cancer screening, we developed two POMDP models with corresponding reward functions. Specifically, the maximum entropy (MaxEnt) IRL algorithm with an adaptive step size was employed to learn rewards more efficiently; and combined with a multiplicative model to learn state-action pair rewards for a POMDP. The POMDP screening models were evaluated based on their ability to recommend appropriate screening decisions before the diagnosis of cancer. The reward functions learned with the MaxEnt IRL algorithm, when combined with POMDP models in lung and breast cancer screening, demonstrate performance comparable to experts. The Cohen's Kappa score of agreement between the POMDPs and physicians' predictions was high in breast cancer and had a decreasing trend in lung cancer.

Clinical phenomapping and outcomes after heart transplantation.

BACKGROUND: Survival after heart transplantation (HTx) is limited by complications related to alloreactivity, immune suppression, and adverse effects of pharmacologic therapies. We hypothesize that time-dependent phenomapping of clinical and molecular data sets is a valuable approach to clinical assessments and guiding medical management to improve outcomes. METHODS: We analyzed clinical, therapeutic, biomarker, and outcome data from 94 adult HTx patients and 1,557 clinical encounters performed between January 2010 and April 2013. Multivariate analyses were used to evaluate the association between immunosuppression therapy, biomarkers, and the combined clinical end point of death, allograft loss, retransplantation, and rejection. Data were analyzed by K-means clustering (K = 2) to identify patterns of similar combined immunosuppression management, and percentile slopes were computed to examine the changes in dosages over time. Findings were correlated with clinical parameters, human leucocyte antigen antibody titers, and peripheral blood mononuclear cell gene expression of the AlloMap (CareDx, Inc., Brisbane, CA) test genes. An intragraft, heart tissue gene coexpression network analysis was performed. RESULTS: Unsupervised cluster analysis of immunosuppressive therapies identified 2 groups, 1 characterized by a steeper immunosuppression minimization, associated with a higher likelihood for the combined end point, and the other by a less pronounced change. A time-dependent phenomap suggested that patients in the group with higher event rates had increased human leukocyte antigen class I and II antibody titers, higher expression of the FLT3 AlloMap gene, and lower expression of the MARCH8 and WDR40A AlloMap genes. Intramyocardial biomarker-related coexpression network analysis of the FLT3 gene showed an immune system-related network underlying this biomarker. CONCLUSIONS: Time-dependent precision phenotyping is a mechanistically insightful, data-driven approach to characterize patterns of clinical care and identify ways to improve clinical management and outcomes.

A Bayesian model for estimating multi-state disease progression.

A growing number of individuals who are considered at high risk of cancer are now routinely undergoing population screening. However, noted harms such as radiation exposure, overdiagnosis, and overtreatment underscore the need for better temporal models that predict who should be screened and at what frequency. The mean sojourn time (MST), an average duration period when a tumor can be detected by imaging but with no observable clinical symptoms, is a critical variable for formulating screening policy. Estimation of MST has been long studied using continuous Markov model (CMM) with Maximum likelihood estimation (MLE). However, a lot of traditional methods assume no observation error of the imaging data, which is unlikely and can bias the estimation of the MST. In addition, the MLE may not be stably estimated when data is sparse. Addressing these shortcomings, we present a probabilistic modeling approach for periodic cancer screening data. We first model the cancer state transition using a three state CMM model, while simultaneously considering observation error. We then jointly estimate the MST and observation error within a Bayesian framework. We also consider the inclusion of covariates to estimate individualized rates of disease progression. Our approach is demonstrated on participants who underwent chest x-ray screening in the National Lung Screening Trial (NLST) and validated using posterior predictive p-values and Pearson's chi-square test. Our model demonstrates more accurate and sensible estimates of MST in comparison to MLE.

Prediction of lung cancer incidence on the low-dose computed tomography arm of the National Lung Screening Trial: A dynamic Bayesian network.

INTRODUCTION: Identifying high-risk lung cancer individuals at an early disease stage is the most effective way of improving survival. The landmark National Lung Screening Trial (NLST) demonstrated the utility of low-dose computed tomography (LDCT) imaging to reduce mortality (relative to X-ray screening). As a result of the NLST and other studies, imaging-based lung cancer screening programs are now being implemented. However, LDCT interpretation results in a high number of false positives. A set of dynamic Bayesian networks (DBN) were designed and evaluated to provide insight into how longitudinal data can be used to help inform lung cancer screening decisions. METHODS: The LDCT arm of the NLST dataset was used to build and explore five DBNs for high-risk individuals. Three of these DBNs were built using a backward construction process, and two using structure learning methods. All models employ demographics, smoking status, cancer history, family lung cancer history, exposure risk factors, comorbidities related to lung cancer, and LDCT screening outcome information. Given the uncertainty arising from lung cancer screening, a cancer state-space model based on lung cancer staging was utilized to characterize the cancer status of an individual over time. The models were evaluated on balanced training and test sets of cancer and non-cancer cases to deal with data imbalance and overfitting. RESULTS: Results were comparable to expert decisions. The average area under the curve (AUC) of the receiver operating characteristic (ROC) for the three intervention points of the NLST trial was higher than 0.75 for all models. Evaluation of the models on the complete LDCT arm of the NLST dataset (N=25,486) demonstrated satisfactory generalization. Consensus of predictions over similar cases is reported in concordance statistics between the models' and the physicians' predictions. The models' predictive ability with respect to missing data was also evaluated with the sample of cases that missed the second screening exam of the trial (N=417). The DBNs outperformed comparison models such as logistic regression and naïve Bayes. CONCLUSION: The lung cancer screening DBNs demonstrated high discrimination and predictive power with the majority of cancer and non-cancer cases.

THYROSIM App for Education and Research Predicts Potential Health Risks of Over-the-Counter Thyroid Supplements.

BACKGROUND: Computer simulation tools for education and research are making increasingly effective use of the Internet and personal devices. To facilitate these activities in endocrinology and metabolism, a mechanistically based simulator of human thyroid hormone and thyrotropin (TSH) regulation dynamics was developed and further validated, and it was implemented as a facile and freely accessible web-based and personal device application: the THYROSIM app. This study elucidates and demonstrates its utility in a research context by exploring key physiological effects of over-the-counter thyroid supplements. METHODS: THYROSIM has a simple and intuitive user interface for teaching and conducting simulated "what-if" experiments. User-selectable "experimental" test-input dosages (oral, intravenous pulses, intravenous infusions) are represented by animated graphical icons integrated with a cartoon of the hypothalamic-pituitary-thyroid axis. Simulations of familiar triiodothyronine (T3), thyroxine (T4), and TSH temporal dynamic responses to these exogenous stimuli are reported graphically, along with normal ranges on the same single interface page; and multiple sets of simulated experimental results are superimposable to facilitate comparative analyses. RESULTS AND CONCLUSIONS: This study shows that THYROSIM accurately reproduces a wide range of published clinical study data reporting hormonal kinetic responses to large and small oral hormone challenges. Simulation examples of partial thyroidectomies and malabsorption illustrate typical usage by optionally changing thyroid gland secretion and/or gut absorption rates--expressed as percentages of normal--as well as additions of oral hormone dosing, all directly on the interface, and visualizing the kinetic responses to these challenges. Classroom and patient education usage--with public health implications--is illustrated by predictive simulated responses to nonprescription thyroid health supplements analyzed previously for T3 and T4 content. Notably, it was found that T3 in supplements has potentially more serious pathophysiological effects than does T4--concomitant with low-normal TSH levels. Some preparations contain enough T3 to generate thyrotoxic conditions, with supernormal serum T3-spiking and subnormal serum T4 and TSH levels and, in some cases, with normal or low-normal range TSH levels due to thyroidal axis negative feedback. These results suggest that appropriate regulation of these products is needed.

A data-driven approach for quality assessment of radiologic interpretations.

Given the increasing emphasis on delivering high-quality, cost-efficient healthcare, improved methodologies are needed to measure the accuracy and utility of ordered diagnostic examinations in achieving the appropriate diagnosis. Here, we present a data-driven approach for performing automated quality assessment of radiologic interpretations using other clinical information (e.g., pathology) as a reference standard for individual radiologists, subspecialty sections, imaging modalities, and entire departments. Downstream diagnostic conclusions from the electronic medical record are utilized as "truth" to which upstream diagnoses generated by radiology are compared. The described system automatically extracts and compares patient medical data to characterize concordance between clinical sources. Initial results are presented in the context of breast imaging, matching 18 101 radiologic interpretations with 301 pathology diagnoses and achieving a precision and recall of 84% and 92%, respectively. The presented data-driven method highlights the challenges of integrating multiple data sources and the application of information extraction tools to facilitate healthcare quality improvement.