Comprehensive tissue deconvolution of cell-free DNA by deep learning for disease diagnosis and monitoring.

Plasma cell-free DNA (cfDNA) is a noninvasive biomarker for cell death of all organs. Deciphering the tissue origin of cfDNA can reveal abnormal cell death because of diseases, which has great clinical potential in disease detection and monitoring. Despite the great promise, the sensitive and accurate quantification of tissue-derived cfDNA remains challenging to existing methods due to the limited characterization of tissue methylation and the reliance on unsupervised methods. To fully exploit the clinical potential of tissue-derived cfDNA, here we present one of the largest comprehensive and high-resolution methylation atlas based on 521 noncancer tissue samples spanning 29 major types of human tissues. We systematically identified fragment-level tissue-specific methylation patterns and extensively validated them in orthogonal datasets. Based on the rich tissue methylation atlas, we develop the first supervised tissue deconvolution approach, a deep-learning-powered model, cfSort, for sensitive and accurate tissue deconvolution in cfDNA. On the benchmarking data, cfSort showed superior sensitivity and accuracy compared to the existing methods. We further demonstrated the clinical utilities of cfSort with two potential applications: aiding disease diagnosis and monitoring treatment side effects. The tissue-derived cfDNA fraction estimated from cfSort reflected the clinical outcomes of the patients. In summary, the tissue methylation atlas and cfSort enhanced the performance of tissue deconvolution in cfDNA, thus facilitating cfDNA-based disease detection and longitudinal treatment monitoring.

Investigating MRI-Associated Biological Aspects of Racial Disparities in Prostate Cancer for African American and White Men.

BACKGROUND: Understanding the characteristics of multiparametric MRI (mpMRI) in patients from different racial/ethnic backgrounds is important for reducing the observed gaps in clinical outcomes. PURPOSE: To investigate the diagnostic performance of mpMRI and quantitative MRI parameters of prostate cancer (PCa) in African American (AA) and matched White (W) men. STUDY TYPE: Retrospective. SUBJECTS: One hundred twenty-nine patients (43 AA, 86 W) with histologically proven PCa who underwent mpMRI before radical prostatectomy. FIELD STRENGTH/SEQUENCE: 3.0 T, T2-weighted turbo spin echo imaging, a single-shot spin-echo EPI sequence diffusion-weighted imaging, and a gradient echo sequence dynamic contrast-enhanced MRI with an ultrafast 3D spoiled gradient-echo sequence. ASSESSMENT: The diagnostic performance of mpMRI in AA and W men was assessed using detection rates (DRs) and positive predictive values (PPVs) in zones defined by the PI-RADS v2.1 prostate sector map. Quantitative MRI parameters, including Ktrans and ve of clinically significant (cs) PCa (Gleason score ≥ 7) tumors were compared between AA and W sub-cohorts after matching age, prostate-specific antigen (PSA), and prostate volume. STATISTICAL TESTS: Weighted Pearson's chi-square and Mann-Whitney U tests with a statistically significant level of 0.05 were used to examine differences in DR and PPV and to compare parameters between AA and matched W men, respectively. RESULTS: A total number of 264 PCa lesions were identified in the study cohort. The PPVs in the peripheral zone (PZ) and posterior prostate of mpMRI for csPCa lesions were significantly higher in AA men than in matched W men (87.8% vs. 68.1% in PZ, and 89.3% vs. 69.6% in posterior prostate). The Ktrans of index csPCa lesions in AA men was significantly higher than in W men (0.25 ± 0.12 vs. 0.20 ± 0.08 min-1; P < 0.01). DATA CONCLUSION: This study demonstrated race-related differences in the diagnostic performances and quantitative MRI measures of csPCa that were not reflected in age, PSA, and prostate volume. EVIDENCE LEVEL: 3 TECHNICAL EFFICACY: Stage 2.

Towards a framework for interoperability and reproducibility of predictive models.

The development and deployment of machine learning (ML) models for biomedical research and healthcare currently lacks standard methodologies. Although tools for model replication are numerous, without a unifying blueprint it remains difficult to scientifically reproduce predictive ML models for any number of reasons (e.g., assumptions regarding data distributions and preprocessing, unclear test metrics, etc.) and ultimately, questions around generalizability and transportability are not readily answered. To facilitate scientific reproducibility, we built upon the Predictive Model Markup Language (PMML) to capture essential information. As a key component of the PREdictive Model Index and Exchange REpository (PREMIERE) platform, we present the Automated Metadata Pipeline (AMP) for conversion of a given predictive ML model into an extended PMML file that autocompletes an ML-based checklist, assessing model elements for interoperability and reproducibility. We demonstrate this pipeline on multiple test cases with three different ML algorithms and health-related datasets, providing a foundation for future predictive model reproducibility, sharing, and comparison.

Expanding Role of Advanced Image Analysis in CT-detected Indeterminate Pulmonary Nodules and Early Lung Cancer Characterization.

The implementation of low-dose chest CT for lung screening presents a crucial opportunity to advance lung cancer care through early detection and interception. In addition, millions of pulmonary nodules are incidentally detected annually in the United States, increasing the opportunity for early lung cancer diagnosis. Yet, realization of the full potential of these opportunities is dependent on the ability to accurately analyze image data for purposes of nodule classification and early lung cancer characterization. This review presents an overview of traditional image analysis approaches in chest CT using semantic characterization as well as more recent advances in the technology and application of machine learning models using CT-derived radiomic features and deep learning architectures to characterize lung nodules and early cancers. Methodological challenges currently faced in translating these decision aids to clinical practice, as well as the technical obstacles of heterogeneous imaging parameters, optimal feature selection, choice of model, and the need for well-annotated image data sets for the purposes of training and validation, will be reviewed, with a view toward the ultimate incorporation of these potentially powerful decision aids into routine clinical practice.

Variability Among Breast Cancer Risk Classification Models When Applied at the Level of the Individual Woman.

BACKGROUND: Breast cancer risk models guide screening and chemoprevention decisions, but the extent and effect of variability among models, particularly at the individual level, is uncertain. OBJECTIVE: To quantify the accuracy and disagreement between commonly used risk models in categorizing individual women as average vs. high risk for developing invasive breast cancer. DESIGN: Comparison of three risk prediction models: Breast Cancer Risk Assessment Tool (BCRAT), Breast Cancer Surveillance Consortium (BCSC) model, and International Breast Intervention Study (IBIS) model. SUBJECTS: Women 40 to 74 years of age presenting for screening mammography at a multisite health system between 2011 and 2015, with 5-year follow-up for cancer outcome. MAIN MEASURES: Comparison of model discrimination and calibration at the population level and inter-model agreement for 5-year breast cancer risk at the individual level using two cutoffs (≥ 1.67% and ≥ 3.0%). KEY RESULTS: A total of 31,115 women were included. When using the ≥ 1.67% threshold, more than 21% of women were classified as high risk for developing breast cancer in the next 5 years by one model, but average risk by another model. When using the ≥ 3.0% threshold, more than 5% of women had disagreements in risk severity between models. Almost half of the women (46.6%) were classified as high risk by at least one of the three models (e.g., if all three models were applied) for the threshold of ≥ 1.67%, and 11.1% were classified as high risk for ≥ 3.0%. All three models had similar accuracy at the population level. CONCLUSIONS: Breast cancer risk estimates for individual women vary substantially, depending on which risk assessment model is used. The choice of cutoff used to define high risk can lead to adverse effects for screening, preventive care, and quality of life for misidentified individuals. Clinicians need to be aware of the high false-positive and false-negative rates and variation between models when talking with patients.

Improving the Quantitative Analysis of Breast Microcalcifications: A Multiscale Approach.

Accurate characterization of microcalcifications (MCs) in 2D digital mammography is a necessary step toward reducing the diagnostic uncertainty associated with the callback of indeterminate MCs. Quantitative analysis of MCs can better identify MCs with a higher likelihood of ductal carcinoma in situ or invasive cancer. However, automated identification and segmentation of MCs remain challenging with high false positive rates. We present a two-stage multiscale approach to MC segmentation in 2D full-field digital mammograms (FFDMs) and diagnostic magnification views. Candidate objects are first delineated using blob detection and Hessian analysis. A regression convolutional network, trained to output a function with a higher response near MCs, chooses the objects which constitute actual MCs. The method was trained and validated on 435 screening and diagnostic FFDMs from two separate datasets. We then used our approach to segment MCs on magnification views of 248 cases with amorphous MCs. We modeled the extracted features using gradient tree boosting to classify each case as benign or malignant. Compared to state-of-the-art comparison methods, our approach achieved superior mean intersection over the union (0.670 ± 0.121 per image versus 0.524 ± 0.034 per image), intersection over the union per MC object (0.607 ± 0.250 versus 0.363 ± 0.278) and true positive rate of 0.744 versus 0.581 at 0.4 false positive detections per square centimeter. Features generated using our approach outperformed the comparison method (0.763 versus 0.710 AUC) in distinguishing amorphous calcifications as benign or malignant.

Medication adherence prediction through temporal modelling in cardiovascular disease management.

BACKGROUND: Chronic conditions place a considerable burden on modern healthcare systems. Within New Zealand and worldwide cardiovascular disease (CVD) affects a significant proportion of the population and it is the leading cause of death. Like other chronic diseases, the course of cardiovascular disease is usually prolonged and its management necessarily long-term. Despite being highly effective in reducing CVD risk, non-adherence to long-term medication continues to be a longstanding challenge in healthcare delivery. The study investigates the benefits of integrating patient history and assesses the contribution of explicitly temporal models to medication adherence prediction in the context of lipid-lowering therapy. METHODS: Data from a CVD risk assessment tool is linked to routinely collected national and regional data sets including pharmaceutical dispensing, hospitalisation, lab test results and deaths. The study extracts a sub-cohort from 564,180 patients who had primary CVD risk assessment for analysis. Based on community pharmaceutical dispensing record, proportion of days covered (PDC) [Formula: see text] 80 is used as the threshold for adherence. Two years (8 quarters) of patient history before their CVD risk assessment is used as the observation window to predict patient adherence in the subsequent 5 years (20 quarters). The predictive performance of temporal deep learning models long short-term memory (LSTM) and simple recurrent neural networks (Simple RNN) are compared against non-temporal models multilayer perceptron (MLP), ridge classifier (RC) and logistic regression (LR). Further, the study investigates the effect of lengthening the observation window on the task of adherence prediction. RESULTS: Temporal models that use sequential data outperform non-temporal models, with LSTM producing the best predictive performance achieving a ROC AUC of 0.805. A performance gap is observed between models that can discover non-linear interactions between predictor variables and their linear counter parts, with neural network (NN) based models significantly outperforming linear models. Additionally, the predictive advantage of temporal models become more pronounced when the length of the observation window is increased. CONCLUSION: The findings of the study provide evidence that using deep temporal models to integrate patient history in adherence prediction is advantageous. In particular, the RNN architecture LSTM significantly outperforms all other model comparators.

Discovering and interpreting transcriptomic drivers of imaging traits using neural networks.

MOTIVATION: Cancer heterogeneity is observed at multiple biological levels. To improve our understanding of these differences and their relevance in medicine, approaches to link organ- and tissue-level information from diagnostic images and cellular-level information from genomics are needed. However, these 'radiogenomic' studies often use linear or shallow models, depend on feature selection, or consider one gene at a time to map images to genes. Moreover, no study has systematically attempted to understand the molecular basis of imaging traits based on the interpretation of what the neural network has learned. These studies are thus limited in their ability to understand the transcriptomic drivers of imaging traits, which could provide additional context for determining clinical outcomes. RESULTS: We present a neural network-based approach that takes high-dimensional gene expression data as input and performs non-linear mapping to an imaging trait. To interpret the models, we propose gene masking and gene saliency to extract learned relationships from radiogenomic neural networks. In glioblastoma patients, our models outperformed comparable classifiers (>0.10 AUC) and our interpretation methods were validated using a similar model to identify known relationships between genes and molecular subtypes. We found that tumor imaging traits had specific transcription patterns, e.g. edema and genes related to cellular invasion, and 10 radiogenomic traits were significantly predictive of survival. We demonstrate that neural networks can model transcriptomic heterogeneity to reflect differences in imaging and can be used to derive radiogenomic traits with clinical value. AVAILABILITY AND IMPLEMENTATION: https://github.com/novasmedley/deepRadiogenomics. CONTACT: whsu@mednet.ucla.edu. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Automated identification and assignment of colonoscopy surveillance recommendations for individuals with colorectal polyps.

BACKGROUND AND AIMS: Determining surveillance intervals for patients with colorectal polyps is critical but time-consuming and challenging to do reliably. We present the development and assessment of a pipeline that leverages natural language processing techniques to automatically extract and analyze relevant polyp findings from free-text colonoscopy and pathology reports. Using this information, we categorized individual patients into 6 postcolonoscopy surveillance intervals defined by the U.S. Multi-Society Task Force on Colorectal Cancer. METHODS: Using a set of 546 randomly selected colonoscopy and pathology reports from 324 patients in a single health system, we used a combination of statistical classifiers and rule-based methods to extract polyp properties from each report type, associate properties with unique polyps, and classify a patient into 1 of 6 risk categories by integrating information from both report types. We then assessed the pipeline's performance by determining the positive predictive value (PPV), sensitivity, and F-score of the algorithm, compared with the determination of surveillance intervals by a gastroenterologist. RESULTS: The pipeline was developed using 346 reports (224 colonoscopy and 122 pathology) from 224 patients and evaluated on an independent test set of 200 reports (100 colonoscopy and 100 pathology) from 100 patients. We achieved an average PPV, sensitivity, and F-score of .92, .95, and .93, respectively, across targeted entities for colonoscopy. Pathology extraction achieved a PPV, sensitivity, and F-score of .95, .97, and .96. The system achieved an overall accuracy of 92% in assigning the recommended interval for surveillance colonoscopy. CONCLUSIONS: This study demonstrates the feasibility of using machine learning to automatically extract findings and classify patients to appropriate risk categories and corresponding surveillance intervals. Incorporating this system can facilitate proactive and timely follow-up after screening colonoscopy and enable real-time quality assessment of prevention programs and providers.

Influence of the Location and Zone of Tumor in Prostate Cancer Detection and Localization on 3-T Multiparametric MRI Based on PI-RADS Version 2.

OBJECTIVE. The objective of our study was to determine the performance of 3-T multiparametric MRI (mpMRI) for prostate cancer (PCa) detection and localization, stratified by anatomic zone and level, using Prostate Imaging Reporting and Data System version 2 (PI-RADSv2) and whole-mount histopathology (WMHP) as reference. MATERIALS AND METHODS. Multiparametric MRI examinations of 415 consecutive men were compared with thin-section WMHP results. A genitourinary radiologist and pathologist collectively determined concordance. Two radiologists assigned PI-RADSv2 scores and sector location to all detected foci by consensus. Tumor detection rates were calculated for clinical and pathologic (tumor location and zone) variables. Both rigid and adjusted sector-matching models were used to account for fixation-related issues. RESULTS. Of 863 PCa foci in 16,185 prostate sectors, the detection of overall and index PCa lesions in the midgland, base, and apex was 54.9% and 83.1%, 42.1% and 64.0% (p = 0.04, p = 0.02), and 41.9% and 71.4% (p = 0.001, p = 0.006), respectively. Tumor localization sensitivity was highest in the midgland compared with the base and apex using an adjusted match compared with a rigid match (index lesions, 71.3% vs 43.7%; all lesions, 70.8% vs 36.0%) and was greater in the peripheral zone (PZ) than in the transition zone. Three-Tesla mpMRI had similarly high specificity (range, 93.8-98.3%) for overall and index tumor localization when using both rigid and adjusted sector-matching approaches. CONCLUSION. For 3-T mpMRI, the highest sensitivity (83.1%) for detection of index PCa lesions was in the midgland, with 98.3% specificity. Multiparametric MRI performance for sectoral localization of PCa within the prostate was moderate and was best for index lesions in the PZ using an adjusted model.

External validation and recalibration of the Brock model to predict probability of cancer in pulmonary nodules using NLST data.

INTRODUCTION: We performed an external validation of the Brock model using the National Lung Screening Trial (NLST) data set, following strict guidelines set forth by the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis statement. We report how external validation results can be interpreted and highlight the role of recalibration and model updating. MATERIALS AND METHODS: We assessed model discrimination and calibration using the NLST data set. Adhering to the inclusion/exclusion criteria reported by McWilliams et al, we identified 7879 non-calcified nodules discovered at the baseline low-dose CT screen with 2 years of follow-up. We characterised differences between Pan-Canadian Early Detection of Lung Cancer Study and NLST cohorts. We calculated the slope on the prognostic index and the intercept coefficient by fitting the original Brock model to NLST. We also assessed the impact of model recalibration and the addition of new covariates such as body mass index, smoking status, pack-years and asbestos. RESULTS: While the area under the curve (AUC) of the model was good, 0.905 (95% CI 0.882 to 0.928), a histogram plot showed that the model poorly differentiated between benign and malignant cases. The calibration plot showed that the model overestimated the probability of cancer. In recalibrating the model, the coefficients for emphysema, spiculation and nodule count were updated. The updated model had an improved calibration and achieved an optimism-corrected AUC of 0.912 (95% CI 0.891 to 0.932). Only pack-year history was found to be significant (p<0.01) among the new covariates evaluated. CONCLUSION: While the Brock model achieved a high AUC when validated on the NLST data set, the model benefited from updating and recalibration. Nevertheless, covariates used in the model appear to be insufficient to adequately discriminate malignant cases.

PI-RADS Version 2 Category on 3 Tesla Multiparametric Prostate Magnetic Resonance Imaging Predicts Oncologic Outcomes in Gleason 3 + 4 Prostate Cancer on Biopsy.

PURPOSE: Three Tesla multiparametric magnetic resonance imaging with PI-RADS™ (Prostate Imaging Reporting and Data System) version 2 scoring is a common tool in prostate cancer diagnosis which informs the likelihood of a cancerous lesion. We investigated whether PI-RADS version 2 also predicts adverse pathology features mainly in patients with biopsy Gleason score 3 + 4 disease. MATERIALS AND METHODS: We reviewed the records of 326 consecutive men with a preoperative template and/or magnetic resonance imaging-ultrasound fusion biopsy Gleason score of 6-7 from a prospectively maintained database of men who underwent robotic radical prostatectomy. The primary analysis was done in patients with biopsy Gleason score 3 + 4 to assess the primary outcome of adverse pathology features on univariate and multivariate logistic regression. The secondary outcome was biochemical recurrence-free survival using the Kaplan-Meier method. Similar analysis was done in patients with a biopsy Gleason score of 6-7. RESULTS: Of men with Gleason score 3 + 4 findings 27%, 15%, 36% and 23% showed a PI-RADS version 2 score of 0-2, 3, 4 and 5, respectively. On univariate analysis PI-RADS version 2 category 5 predicted adverse pathology features vs categories 0-2 (OR 10.7, 95% CI 3.7-31, p ≤0.001). On multivariate analysis the PI-RADS version 2 category 5 was associated with adverse pathology when adjusting for preoperative magnetic resonance imaging targeted biopsy (OR 11.4, 95% CI 3.7-35, p ≤0.0001). In men with a targeted biopsy Gleason score of 3 + 4 prostate cancer PI-RADS version 2 category 5 was associated with adverse pathology (OR 14.7, 95% CI 1.5-146.9, p = 0.02). Of men with biopsy Gleason score 3 + 4 disease 92% and 58% with a PI-RADS version 2 score of 4 and 5, respectively, had 2-year biochemical recurrence-free survival. CONCLUSIONS: A PI-RADS version 2 category 5 lesion in patients with a biopsy Gleason score 3 + 4 lesion predicted adverse pathology features and biochemical recurrence-free survival. These findings suggest that preoperative 3 Tesla multiparametric magnetic resonance imaging may serve as a prognostic marker of treatment outcomes independently of biopsy Gleason score or biopsy type.

PI-RADS Version 2 Category on 3 Tesla Multiparametric Prostate Magnetic Resonance Imaging Predicts Oncologic Outcomes in Gleason 3 + 4 Prostate Cancer on Biopsy.

PURPOSE: Three Tesla multiparametric magnetic resonance imaging with PI-RADS™ (Prostate Imaging Reporting and Data System) version 2 scoring is a common tool in prostate cancer diagnosis which informs the likelihood of a cancerous lesion. We investigated whether PI-RADS version 2 also predicts adverse pathology features mainly in patients with biopsy Gleason score 3 + 4 disease. MATERIALS AND METHODS: We reviewed the records of 326 consecutive men with a preoperative template and/or magnetic resonance imaging-ultrasound fusion biopsy Gleason score of 6-7 from a prospectively maintained database of men who underwent robotic radical prostatectomy. The primary analysis was done in patients with biopsy Gleason score 3 + 4 to assess the primary outcome of adverse pathology features on univariate and multivariate logistic regression. The secondary outcome was biochemical recurrence-free survival using the Kaplan-Meier method. Similar analysis was done in patients with a biopsy Gleason score of 6-7. RESULTS: Of men with Gleason score 3 + 4 findings 27%, 15%, 36% and 23% showed a PI-RADS version 2 score of 0-2, 3, 4 and 5, respectively. On univariate analysis PI-RADS version 2 category 5 predicted adverse pathology features vs categories 0-2 (OR 10.7, 95% CI 3.7-31, p ≤0.001). On multivariate analysis the PI-RADS version 2 category 5 was associated with adverse pathology when adjusting for preoperative magnetic resonance imaging targeted biopsy (OR 11.4, 95% CI 3.7-35, p ≤0.0001). In men with a targeted biopsy Gleason score of 3 + 4 prostate cancer PI-RADS version 2 category 5 was associated with adverse pathology (OR 14.7, 95% CI 1.5-146.9, p = 0.02). Of men with biopsy Gleason score 3 + 4 disease 92% and 58% with a PI-RADS version 2 score of 4 and 5, respectively, had 2-year biochemical recurrence-free survival. CONCLUSIONS: A PI-RADS version 2 category 5 lesion in patients with a biopsy Gleason score 3 + 4 lesion predicted adverse pathology features and biochemical recurrence-free survival. These findings suggest that preoperative 3 Tesla multiparametric magnetic resonance imaging may serve as a prognostic marker of treatment outcomes independently of biopsy Gleason score or biopsy type.

Detection and Localization of Prostate Cancer at 3-T Multiparametric MRI Using PI-RADS Segmentation.

OBJECTIVE. The purpose of this study is to determine the overall and sector-based performance of 3-T multiparametric MRI for prostate cancer (PCa) detection and localization by using Prostate Imaging-Reporting and Data System version 2 (PI-RADSv2) scoring and segmentation compared with whole-mount histopathologic analysis. MATERIALS AND METHODS. Multiparametric 3-T MRI examinations of 415 consecutive men were compared with thin-section whole-mount histopathologic analysis. A genitourinary radiologist and pathologist collectively determined concordance. Two radiologists assigned PI-RADSv2 categories and sectoral location to all detected foci by consensus. Tumor detection rates were calculated for clinical and pathologic (Gleason score) variables. Both rigid and adjusted sector-matching models were used to account for fixation-related issues. RESULTS. The 415 patients had 863 PCa foci (52.7% had a Gleason score ≥ 7, 61.9% were ≥ 1 cm, and 90.4% (375/415) of index lesions were ≥ 1 cm) and 16,185 prostate sectors. Multiparametric MRI enabled greater detection of PCa lesions 1 cm or larger (all lesions vs index lesions, 61.6% vs 81.6%), lesions with Gleason score greater than or equal to 7 (all lesions vs index lesions, 71.4% vs 80.9%), and index lesions with both Gleason score greater than or equal to 7 and size 1 cm or larger (83.3%). Higher sensitivity was obtained for adjusted versus rigid tumor localization for all lesions (56.0% vs 28.5%), index lesions (55.4% vs 34.3%), lesions with Gleason score greater than or equal to 7 (55.7% vs 36.0%), and index lesions 1 cm or larger (56.1% vs 35.0%). Multiparametric 3-T MRI had similarly high specificity (96.0-97.9%) for overall and index tumor localization with adjusted and rigid sector-matching approaches. CONCLUSION. Using 3-T multiparametric MRI and PI-RADSv2, we achieved the highest sensitivity (83.3%) for the detection of lesions 1 cm or larger with Gleason score greater than or equal to 7. Sectoral localization of PCa within the prostate was moderate and was better with an adjusted model than with a rigid model.

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.

Pathological and 3 Tesla Volumetric Magnetic Resonance Imaging Predictors of Biochemical Recurrence after Robotic Assisted Radical Prostatectomy: Correlation with Whole Mount Histopathology.

PURPOSE: We sought to identify the clinical and magnetic resonance imaging variables predictive of biochemical recurrence after robotic assisted radical prostatectomy in patients who underwent multiparametric 3 Tesla prostate magnetic resonance imaging. MATERIALS AND METHODS: We performed an institutional review board approved, HIPAA (Health Insurance Portability and Accountability Act) compliant, single arm observational study of 3 Tesla multiparametric magnetic resonance imaging prior to robotic assisted radical prostatectomy from December 2009 to March 2016. Clinical, magnetic resonance imaging and pathological information, and clinical outcomes were compiled. Biochemical recurrence was defined as prostate specific antigen 0.2 ng/cc or greater. Univariate and multivariate regression analysis was performed. RESULTS: Biochemical recurrence had developed in 62 of the 255 men (24.3%) included in the study at a median followup of 23.5 months. Compared to the subcohort without biochemical recurrence the subcohort with biochemical recurrence had a greater proportion of patients with a high grade biopsy Gleason score, higher preoperative prostate specific antigen (7.4 vs 5.6 ng/ml), intermediate and high D'Amico classifications, larger tumor volume on magnetic resonance imaging (0.66 vs 0.30 ml), higher PI-RADS® (Prostate Imaging-Reporting and Data System) version 2 category lesions, a greater proportion of intermediate and high grade radical prostatectomy Gleason score lesions, higher pathological T3 stage (all p <0.01) and a higher positive surgical margin rate (19.3% vs 7.8%, p = 0.016). On multivariable analysis only tumor volume on magnetic resonance imaging (adjusted OR 1.57, p = 0.016), pathological T stage (adjusted OR 2.26, p = 0.02), positive surgical margin (adjusted OR 5.0, p = 0.004) and radical prostatectomy Gleason score (adjusted OR 2.29, p = 0.004) predicted biochemical recurrence. CONCLUSIONS: In this cohort tumor volume on magnetic resonance imaging and pathological variables, including Gleason score, staging and positive surgical margins, significantly predicted biochemical recurrence. This suggests an important new imaging biomarker.

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.

Comparative risks for cancer associated with use of calcitonin, bisphosphonates or selective estrogen receptor modulators among osteoporosis patients: a population-based cohort study.

BACKGROUND: This population-based cohort study was to compare the risks of incident cancer in osteoporosis patients who used bisphosphonates, calcitonin or selective estrogen receptor modulators (SERMs). METHODS: We identified 9995 patients who were diagnosed with osteoporosis and prescribed osteoporosis drugs (bisphosphonate (n = 4675), calcitonin (n = 3993) and SERMs (n = 1327)) between 1 January 2000 and 31 December 2006 in Taiwan's National Health Insurance Research Database. Date of first prescription of osteoporosis drugs was assigned as the index date. The outcome measurement was incident cancer, defined by a first-ever inpatient visit with a primary diagnosis of cancer. All patients were followed until the occurrence of cancer. For those who did not develop cancer, we censored them at 1 year after their last prescription of osteoporosis drugs. Cox proportional hazard models were used to examine the association between risk of cancer and use of calcitonin, bisphosphonates or SERMs. RESULTS: The incidence rate of cancer was 68.8, 34.0 and 29.6 per 1000 person years in the calcitonin, SERMs and bisphosphonate cohorts, respectively. Compared with bisphosphonate users, calcitonin users were associated with an increased risk of cancer (adjusted hazard ratio (HR) 2.11, 95% confidence interval (CI) 2.01-2.21, P < 0.001). SERM users were associated with an increased risk of cancer (adjusted HR 1.20, 95% CI 1.13-1.28, P < 0.001). CONCLUSION: Our findings suggest that calcitonin is associated with an increased risk of cancer than bisphosphonate, supporting the recent warning issued by the European Medicines Agency and US Food and Drug Administration. SERMs is found to be associated with an increased risk of cancer than bisphosphonate.

Utilizing time-driven activity-based costing to understand the short- and long-term costs of treating localized, low-risk prostate cancer.

BACKGROUND: Given the costs of delivering care for men with prostate cancer remain poorly described, this article reports the results of time-driven activity-based costing (TDABC) for competing treatments of low-risk prostate cancer. METHODS: Process maps were developed for each phase of care from the initial urologic visit through 12 years of follow-up for robotic-assisted laparoscopic prostatectomy (RALP), cryotherapy, high-dose rate (HDR) and low-dose rate (LDR) brachytherapy, intensity-modulated radiation therapy (IMRT), stereotactic body radiation therapy (SBRT), and active surveillance (AS). The last modality incorporated both traditional transrectal ultrasound (TRUS) biopsy and multiparametric-MRI/TRUS fusion biopsy. The costs of materials, equipment, personnel, and space were calculated per unit of time and based on the relative proportion of capacity used. TDABC for each treatment was defined as the sum of its resources. RESULTS: Substantial cost variation was observed at 5 years, with costs ranging from $7,298 for AS to $23,565 for IMRT, and they remained consistent through 12 years of follow-up. LDR brachytherapy ($8,978) was notably cheaper than HDR brachytherapy ($11,448), and SBRT ($11,665) was notably cheaper than IMRT, with the cost savings attributable to shorter procedure times and fewer visits required for treatment. Both equipment costs and an inpatient stay ($2,306) contributed to the high cost of RALP ($16,946). Cryotherapy ($11,215) was more costly than LDR brachytherapy, largely because of increased single-use equipment costs ($6,292 vs $1,921). AS reached cost equivalence with LDR brachytherapy after 7 years of follow-up. CONCLUSIONS: The use of TDABC is feasible for analyzing cancer services and provides insights into cost-reduction tactics in an era focused on emphasizing value. By detailing all steps from diagnosis and treatment through 12 years of follow-up for low-risk prostate cancer, this study has demonstrated significant cost variation between competing treatments.