An integrated radiology-pathology machine learning classifier for outcome prediction following radical prostatectomy: Preliminary findings

Objectives

To evaluate the added benefit of integrating features from pre-treatment MRI (radiomics) and digitized post-surgical pathology slides (pathomics) in prostate cancer (PCa) patients for prognosticating outcomes post radical-prostatectomy (RP) including a) rising prostate specific antigen (PSA), and b) extraprostatic-extension (EPE).

Methods

Multi-institutional data (N = 58) of PCa patients who underwent pre-treatment 3-T MRI prior to RP were included in this retrospective study. Radiomic and pathomic features were extracted from PCa regions on MRI and RP specimens delineated by expert clinicians. On training set (D1, N = 44), Cox Proportional-Hazards models MR, MP and MRaP were trained using radiomics, pathomics, and their combination, respectively, to prognosticate rising PSA (PSA > 0.03 ng/mL). Top features from MRaP were used to train a model to predict EPE on D1 and test on external dataset (D2, N = 14). C-index, Kalplan-Meier curves were used for survival analysis, and area under ROC (AUC) was used for EPE. MRaP was compared with the existing post-treatment risk-calculator, CAPRA (MC).

Results

Patients had median follow-up of 34 months. MRaP (c-index = 0.685 ± 0.05) significantly outperformed MR (c-index = 0.646 ± 0.05), MP (c-index = 0.631 ± 0.06) and MC (c-index = 0.601 ± 0.071) (p < 0.0001). Cross-validated Kaplan-Meier curves showed significant separation among risk groups for rising PSA for MRaP (p < 0.005, Hazard Ratio (HR) = 11.36) as compared to MR (p = 0.64, HR = 1.33), MP (p = 0.19, HR = 2.82) and MC (p = 0.10, HR = 3.05). Integrated radio-pathomic model MRaP (AUC = 0.80) outperformed MR (AUC = 0.57) and MP (AUC = 0.76) in predicting EPE on external-data (D2).

Conclusions

Results from this preliminary study suggest that a combination of radiomic and pathomic features can better predict post-surgical outcomes (rising PSA and EPE) compared to either of them individually as well as extant prognostic nomogram (CAPRA).

Multi-scale statistical deformation based co-registration of prostate MRI and post-surgical whole mount histopathology

BACKGROUND

Accurate delineations of regions of interest (ROIs) on multi-parametric magnetic resonance imaging (mpMRI) are crucial for development of automated, machine learning-based prostate cancer (PCa) detection and segmentation models. However, manual ROI delineations are labor-intensive and susceptible to inter-reader variability. Histopathology images from radical prostatectomy (RP) represent the "gold standard" in terms of the delineation of disease extents, for example, PCa, prostatitis, and benign prostatic hyperplasia (BPH). Co-registering digitized histopathology images onto pre-operative mpMRI enables automated mapping of the ground truth disease extents onto mpMRI, thus enabling the development of machine learning tools for PCa detection and risk stratification. Still, MRI-histopathology co-registration is challenging due to various artifacts and large deformation between in vivo MRI and ex vivo whole-mount histopathology images (WMHs). Furthermore, the artifacts on WMHs, such as tissue loss, may introduce unrealistic deformation during co-registration.

PURPOSE

This study presents a new registration pipeline, MSERgSDM, a multi-scale feature-based registration (MSERg) with a statistical deformation (SDM) constraint, which aims to improve accuracy of MRI-histopathology co-registration.

METHODS

In this study, we collected 85 pairs of MRI and WMHs from 48 patients across three cohorts. Cohort 1 (D1), comprised of a unique set of 3D printed mold data from six patients, facilitated the generation of ground truth deformations between ex vivo WMHs and in vivo MRI. The other two clinically acquired cohorts (D2 and D3) included 42 patients. Affine and nonrigid registrations were employed to minimize the deformation between ex vivo WMH and ex vivo T2-weighted MRI (T2WI) in D1. Subsequently, ground truth deformation between in vivo T2WI and ex vivo WMH was approximated as the deformation between in vivo T2WI and ex vivo T2WI. In D2 and D3, the prostate anatomical annotations, for example, tumor and urethra, were made by a pathologist and a radiologist in collaboration. These annotations included ROI boundary contours and landmark points. Before applying the registration, manual corrections were made for flipping and rotation of WMHs. MSERgSDM comprises two main components: (1) multi-scale representation construction, and (2) SDM construction. For the SDM construction, we collected N = 200 reasonable deformation fields generated using MSERg, verified through visual inspection. Three additional methods, including intensity-based registration, ProsRegNet, and MSERg, were also employed for comparison against MSERgSDM.

RESULTS

Our results suggest that MSERgSDM performed comparably to the ground truth (p > 0.05). Additionally, MSERgSDM (ROI Dice ratio = 0.61, landmark distance = 3.26 mm) exhibited significant improvement over MSERg (ROI Dice ratio = 0.59, landmark distance = 3.69 mm) and ProsRegNet (ROI Dice ratio = 0.56, landmark distance = 4.00 mm) in local alignment.

CONCLUSIONS

This study presents a novel registration method, MSERgSDM, for mapping ex vivo WMH onto in vivo prostate MRI. Our preliminary results demonstrate that MSERgSDM can serve as a valuable tool to map ground truth disease annotations from histopathology images onto MRI, thereby assisting in the development of machine learning models for PCa detection on MRI.

Identifying and improving the "ground truth" of race in disparities research through improved EMR data reporting. A systematic review

BACKGROUND

Studies about racial disparities in healthcare are increasing in quantity; however, they are subject to vast differences in definition, classification, and utilization of race/ethnicity data. Improved standardization of this information can strengthen conclusions drawn from studies using such data. The objective of this study is to examine how data related to race/ethnicity are recorded in research through examining articles on race/ethnicity health disparities and examine problems and solutions in data reporting that may impact overall data quality.

METHODS

In this systematic review, Business Source Complete, Embase.com, IEEE Xplore, PubMed, Scopus and Web of Science Core Collection were searched for relevant articles published from 2000 to 2020. Search terms related to the concepts of electronic medical records, race/ethnicity, and data entry related to race/ethnicity were used. Exclusion criteria included articles not in the English language and those describing pediatric populations. Data were extracted from published articles. This review was organized and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement for systematic reviews.

FINDINGS

In this systematic review, 109 full text articles were reviewed. Weaknesses and possible solutions have been discussed in current literature, with the predominant problem and solution as follows: the electronic medical record (EMR) is vulnerable to inaccuracies and incompleteness in the methods that research staff collect this data; however, improved standardization of the collection and use of race data in patient care may help alleviate these inaccuracies.

INTERPRETATION

Conclusions drawn from large datasets concerning peoples of certain race/ethnic groups should be made cautiously, and a careful review of the methodology of each publication should be considered prior to implementation in patient care.

Delta radiomic patterns on serial bi-parametric MRI are associated with pathologic upgrading in prostate cancer patients on active surveillance: preliminary findings

Objective

The aim of this study was to quantify radiomic changes in prostate cancer (PCa) progression on serial MRI among patients on active surveillance (AS) and evaluate their association with pathologic progression on biopsy.

Methods

This retrospective study comprised N = 121 biopsy-proven PCa patients on AS at a single institution, of whom N = 50 at baseline conformed to the inclusion criteria. ISUP Gleason Grade Groups (GGG) were obtained from 12-core TRUS-guided systematic biopsies at baseline and follow-up. A biopsy upgrade (AS+) was defined as an increase in GGG (or in number of positive cores) and no upgrade (AS-) was defined when GGG remained the same during a median period of 18 months. Of N = 50 patients at baseline, N = 30 had MRI scans available at follow-up (median interval = 18 months) and were included for delta radiomic analysis. A total of 252 radiomic features were extracted from the PCa region of interest identified by board-certified radiologists on 3T bi-parametric MRI [T2-weighted (T2W) and apparent diffusion coefficient (ADC)]. Delta radiomic features were computed as the difference of radiomic feature between baseline and follow-up scans. The association of AS+ with age, prostate-specific antigen (PSA), Prostate Imaging Reporting and Data System (PIRADS v2.1) score, and tumor size was evaluated at baseline and follow-up. Various prediction models were built using random forest (RF) classifier within a threefold cross-validation framework leveraging baseline radiomics (Cbr), baseline radiomics + baseline clinical (Cbrbcl), delta radiomics (CΔr), delta radiomics + baseline clinical (CΔrbcl), and delta radiomics + delta clinical (CΔrΔcl).

Results

An AUC of 0.64 ± 0.09 was obtained for Cbr, which increased to 0.70 ± 0.18 with the integration of clinical variables (Cbrbcl). CΔr yielded an AUC of 0.74 ± 0.15. Integrating delta radiomics with baseline clinical variables yielded an AUC of 0.77 ± 0.23. CΔrΔclresulted in the best AUC of 0.84 ± 0.20 (p < 0.05) among all combinations.

Conclusion

Our preliminary findings suggest that delta radiomics were more strongly associated with upgrade events compared to PIRADS and other clinical variables. Delta radiomics on serial MRI in combination with changes in clinical variables (PSA and tumor volume) between baseline and follow-up showed the strongest association with biopsy upgrade in PCa patients on AS. Further independent multi-site validation of these preliminary findings is warranted.

"Shortcuts" Causing Bias in Radiology Artificial Intelligence: Causes, Evaluation, and Mitigation

Despite the expert-level performance of artificial intelligence (AI) models for various medical imaging tasks, real-world performance failures with disparate outputs for various subgroups limit the usefulness of AI in improving patients' lives. Many definitions of fairness have been proposed, with discussions of various tensions that arise in the choice of an appropriate metric to use to evaluate bias; for example, should one aim for individual or group fairness? One central observation is that AI models apply "shortcut learning" whereby spurious features (such as chest tubes and portable radiographic markers on intensive care unit chest radiography) on medical images are used for prediction instead of identifying true pathology. Moreover, AI has been shown to have a remarkable ability to detect protected attributes of age, sex, and race, while the same models demonstrate bias against historically underserved subgroups of age, sex, and race in disease diagnosis. Therefore, an AI model may take shortcut predictions from these correlations and subsequently generate an outcome that is biased toward certain subgroups even when protected attributes are not explicitly used as inputs into the model. As a result, these subgroups became nonprivileged subgroups. In this review, the authors discuss the various types of bias from shortcut learning that may occur at different phases of AI model development, including data bias, modeling bias, and inference bias. The authors thereafter summarize various tool kits that can be used to evaluate and mitigate bias and note that these have largely been applied to nonmedical domains and require more evaluation for medical AI. The authors then summarize current techniques for mitigating bias from preprocessing (data-centric solutions) and during model development (computational solutions) and postprocessing (recalibration of learning). Ongoing legal changes where the use of a biased model will be penalized highlight the necessity of understanding, detecting, and mitigating biases from shortcut learning and will require diverse research teams looking at the whole AI pipeline.

Novel radiomic analysis on bi-parametric MRI for characterizing differences between MR non-visible and visible clinically significant prostate cancer

Background

around one third of clinically significant prostate cancer (CsPCa) foci are reported to be MRI non-visible (MRI─).

Objective

To quantify the differences between MR visible (MRI+) and MRI^─ CsPCa using intra- and peri-lesional radiomic features on bi-parametric MRI (bpMRI).

Methods

This retrospective and multi-institutional study comprised 164 patients with pre-biopsy 3T prostate multi-parametric MRI from 2014 to 2017. The MRI^─ CsPCa referred to lesions with PI-RADS v2 score < 3 but ISUP grade group > 1. Three experienced radiologists were involved in annotating lesions and PI-RADS assignment. The validation set (D_v) comprised 52 patients from a single institution, the remaining 112 patients were used for training (D_t). 200 radiomic features were extracted from intra-lesional and peri-lesional regions on bpMRI.Logistic regression with least absolute shrinkage and selection operator (LASSO) and 10-fold cross-validation was applied on D_t to identify radiomic features associated with MRI^─ and MRI⁺ CsPCa to generate corresponding risk scores R^MRI─ and R^MRI+. R_bpMRI was further generated by integrating R^MRI─ and R^MRI+. Statistical significance was determined using the Wilcoxon signed-rank test.

Results

Both intra-lesional and peri-lesional bpMRI Haralick and CoLlAGe radiomic features were significantly associated with MRI^─ CsPCa (p < 0.05). Intra-lesional ADC Haralick and CoLlAGe radiomic features were significantly different among MRI^─ and MRI⁺ CsPCa (p < 0.05). R_bpMRI yielded the highest AUC of 0.82 (95 % CI 0.72-0.91) compared to AUCs of R^MRI+ 0.76 (95 % CI 0.63-0.89), and PI-RADS 0.58 (95 % CI 0.50-0.72) on D_v. R_bpMRI correctly reclassified 10 out of 14 MRI^─ CsPCa on D_v.

Conclusion

Our preliminary results demonstrated that both intra-lesional and peri-lesional bpMRI radiomic features were significantly associated with MRI^─ CsPCa. These features could assist in CsPCa identification on bpMRI.

Ten quick tips for computational analysis of medical images

Medical imaging is a great asset for modern medicine, since it allows physicians to spatially interrogate a disease site, resulting in precise intervention for diagnosis and treatment, and to observe particular aspect of patients' conditions that otherwise would not be noticeable. Computational analysis of medical images, moreover, can allow the discovery of disease patterns and correlations among cohorts of patients with the same disease, thus suggesting common causes or providing useful information for better therapies and cures. Machine learning and deep learning applied to medical images, in particular, have produced new, unprecedented results that can pave the way to advanced frontiers of medical discoveries. While computational analysis of medical images has become easier, however, the possibility to make mistakes or generate inflated or misleading results has become easier, too, hindering reproducibility and deployment. In this article, we provide ten quick tips to perform computational analysis of medical images avoiding common mistakes and pitfalls that we noticed in multiple studies in the past. We believe our ten guidelines, if taken into practice, can help the computational-medical imaging community to perform better scientific research that eventually can have a positive impact on the lives of patients worldwide.

Prostate Surface Distension and Tumor Texture Descriptors From Pre-Treatment MRI Are Associated With Biochemical Recurrence Following Radical Prostatectomy: Preliminary Findings

Objective

To derive and evaluate the association of prostate shape distension descriptors from T2-weighted MRI (T2WI) with prostate cancer (PCa) biochemical recurrence (BCR) post-radical prostatectomy (RP) independently and in conjunction with texture radiomics of PCa.

Methods

This retrospective study comprised 133 PCa patients from two institutions who underwent 3T-MRI prior to RP and were followed up with PSA measurements for ≥3 years. A 3D shape atlas-based approach was adopted to derive prostate shape distension descriptors from T2WI, and these descriptors were used to train a random forest classifier (C_S) to predict BCR. Texture radiomics was derived within PCa regions of interest from T2WI and ADC maps, and another machine learning classifier (C_R) was trained for BCR. An integrated classifier C_S+R was then trained using predictions from C_S and C_R. These models were trained on D₁ (N = 71, 27 BCR+) and evaluated on independent hold-out set D₂ (N = 62, 12 BCR+). C_S+R was compared against pre-RP, post-RP clinical variables, and extant nomograms for BCR-free survival (bFS) at 3 years.

Results

C_S+R resulted in a higher AUC (0.75) compared to C_R (0.70, p = 0.04) and C_S (0.69, p = 0.01) on D₂ in predicting BCR. On univariable analysis, C_S+R achieved a higher hazard ratio (2.89, 95% CI 0.35–12.81, p < 0.01) compared to other pre-RP clinical variables for bFS. C_S+R, pathologic Gleason grade, extraprostatic extension, and positive surgical margins were associated with bFS (p < 0.05). C_S+R resulted in a higher C-index (0.76 ± 0.06) compared to CAPRA (0.69 ± 0.09, p < 0.01) and Decipher risk (0.59 ± 0.06, p < 0.01); however, it was comparable to post-RP CAPRA-S (0.75 ± 0.02, p = 0.07).

Conclusions

Radiomic shape descriptors quantifying prostate surface distension complement texture radiomics of prostate cancer on MRI and result in an improved association with biochemical recurrence post-radical prostatectomy.

Integrating pathomics with radiomics and genomics for cancer prognosis: A brief review

In the last decade, the focus of computational pathology research community has shifted from replicating the pathological examination for diagnosis done by pathologists to unlocking and discovering "sub-visual" prognostic image cues from the histopathological image. While we are getting more knowledge and experience in digital pathology, the emerging goal is to integrate other-omics or modalities that will contribute for building a better prognostic assay. In this paper, we provide a brief review of representative works that focus on integrating pathomics with radiomics and genomics for cancer prognosis. It includes: correlation of pathomics and genomics; fusion of pathomics and genomics; fusion of pathomics and radiomics. We also present challenges, potential opportunities, and avenues for future work.

An integrated nomogram combining deep learning, Prostate Imaging-Reporting and Data System (PI-RADS) scoring, and clinical variables for identification of clinically significant prostate cancer on biparametric MRI: a retrospective multicentre study

Background

Biparametric MRI (comprising T2-weighted MRI and apparent diffusion coefficient maps) is increasingly being used to characterise prostate cancer. Although previous studies have combined Prostate Imaging–Reporting & Data System (PI-RADS)-based MRI findings with routinely available clinical variables and with deep learning-based imaging predictors, respectively, for prostate cancer risk stratification, none have combined all three. We aimed to construct an integrated nomogram (referred to as ClaD) combining deep learning-based imaging predictions, PI-RADS scoring, and clinical variables to identify clinically significant prostate cancer on biparametric MRI.

Methods

In this retrospective multicentre study, we included patients with prostate cancer, with histopathology or biopsy reports and a screening or diagnostic MRI scan in the axial view, from four cohorts in the USA (from University Hospitals Cleveland Medical Center, Icahn School of Medicine at Mount Sinai, Cleveland Clinic, and Long Island Jewish Medical Center) and from the PROSTATEx Challenge dataset in the Netherlands. We constructed an integrated nomogram combining deep learning, PI-RADS score, and clinical variables (prostate-specific antigen, prostate volume, and lesion volume) using multivariable logistic regression to identify clinically significant prostate cancer on biparametric MRI. We used data from the first three cohorts to train the nomogram and data from the remaining two cohorts for independent validation. We compared the performance of our ClaD integrated nomogram with that of integrated nomograms combining clinical variables with either the deep learning-based imaging predictor (referred to as DIN) or PI-RADS score (referred to as PIN) using area under the receiver operating characteristic curves (AUCs). We also compared the ability of the nomograms to predict biochemical recurrence on a subset of patients who had undergone radical prostatectomy. We report cross-validation AUCs as means for the training set and used AUCs with 95% CIs to assess the performance on the test set. The difference in AUCs between the models were tested for statistical significance using DeLong’s test. We used log-rank tests and Kaplan-Meier curves to analyse survival.

Findings

We investigated 592 patients (823 lesions) with prostate cancer who underwent 3T multiparametric MRI at five hospitals in the USA between Jan 8, 2009, and June 3, 2017. The training data set consisted of 368 patients from three sites (the PROSTATEx Challenge cohort [n=204], University Hospitals Cleveland Medical Center [n=126], and Icahn School of Medicine at Mount Sinai [n=38]), and the independent validation data set consisted of 224 patients from two sites (Cleveland Clinic [n=151] and Long Island Jewish Medical Center [n=73]). The ClaD clinical nomogram yielded an AUC of 0·81 (95% CI 0·76–0·85) for identification of clinically significant prostate cancer in the validation data set, significantly improving performance over the DIN (0·74 [95% CI 0·69–0·80], p=0·0005) and PIN (0·76 [0·71–0·81], p<0·0001) nomograms. In the subset of patients who had undergone radical prostatectomy (n=81), the ClaD clinical nomogram resulted in a significant separation in Kaplan-Meier survival curves between patients with and without biochemical recurrence (HR 5·92 [2·34–15·00], p=0·044), whereas the DIN (1·22 [0·54–2·79], p=0·65) and PIN nomograms did not (1·30 [0·62–2·71], p=0·51).

Interpretation

Risk stratification of patients with prostate cancer using the integrated ClaD nomogram could help to identify patients with prostate cancer who are at low risk, very low risk, and favourable intermediate risk, who might be candidates for active surveillance, and could also help to identify patients with lethal prostate cancer who might benefit from adjuvant therapy.

Funding

National Cancer Institute of the US National Institutes of Health, National Institute for Biomedical Imaging and Bioengineering, National Center for Research Resources, US Department of Veterans Affairs Biomedical Laboratory Research and Development Service, US Department of Defense, US National Institute of Diabetes and Digestive and Kidney Diseases, The Ohio Third Frontier Technology Validation Fund, Case Western Reserve University, Dana Foundation, and Clinical and Translational Science Collaborative.

Computer extracted gland features from H&E predicts prostate cancer recurrence comparably to a genomic companion diagnostic test: a large multi-site study

Existing tools for post-radical prostatectomy (RP) prostate cancer biochemical recurrence (BCR) prognosis rely on human pathologist-derived parameters such as tumor grade, with the resulting inter-reviewer variability. Genomic companion diagnostic tests such as Decipher tend to be tissue destructive, expensive, and not routinely available in most centers. We present a tissue non-destructive method for automated BCR prognosis, termed "Histotyping", that employs computational image analysis of morphologic patterns of prostate tissue from a single, routinely acquired hematoxylin and eosin slide. Patients from two institutions (n = 214) were used to train Histotyping for identifying high-risk patients based on six features of glandular morphology extracted from RP specimens. Histotyping was validated for post-RP BCR prognosis on a separate set of n = 675 patients from five institutions and compared against Decipher on n = 167 patients. Histotyping was prognostic of BCR in the validation set (p < 0.001, univariable hazard ratio [HR] = 2.83, 95% confidence interval [CI]: 2.03-3.93, concordance index [c-index] = 0.68, median years-to-BCR: 1.7). Histotyping was also prognostic in clinically stratified subsets, such as patients with Gleason grade group 3 (HR = 4.09) and negative surgical margins (HR = 3.26). Histotyping was prognostic independent of grade group, margin status, pathological stage, and preoperative prostate-specific antigen (PSA) (multivariable p < 0.001, HR = 2.09, 95% CI: 1.40-3.10, n = 648). The combination of Histotyping, grade group, and preoperative PSA outperformed Decipher (c-index = 0.75 vs. 0.70, n = 167). These results suggest that a prognostic classifier for prostate cancer based on digital images could serve as an alternative or complement to molecular-based companion diagnostic tests.

A novel imaging based Nomogram for predicting post-surgical biochemical recurrence and adverse pathology of prostate cancer from pre-operative bi-parametric MRI

Background

We developed and validated an integrated radiomic-clinicopathologic nomogram (RadClip) for post-surgical biochemical recurrence free survival (bRFS) and adverse pathology (AP) prediction in men with prostate cancer (PCa). RadClip was further compared against extant prognostics tools like CAPRA and Decipher.

Methods

A retrospective study of 198 patients with PCa from four institutions who underwent pre-operative 3 Tesla MRI followed by radical prostatectomy, between 2009 and 2017 with a median 35-month follow-up was performed. Radiomic features were extracted from prostate cancer regions on bi-parametric magnetic resonance imaging (bpMRI). Cox Proportional-Hazards (CPH) model warped with minimum redundancy maximum relevance (MRMR) feature selection was employed to select bpMRI radiomic features for bRFS prediction in the training set (D₁, N = 71). In addition, a bpMRI radiomic risk score (RadS) and associated nomogram, RadClip, were constructed in D₁ and then compared against the Decipher, pre-operative (CAPRA), and post-operative (CAPRA-S) nomograms for bRFS and AP prediction in the testing set (D₂, N = 127).

Findings

“RadClip yielded a higher C-index (0.77, 95% CI 0.65-0.88) compared to CAPRA (0.68, 95% CI 0.57-0.8) and Decipher (0.51, 95% CI 0.33-0.69) and was found to be comparable to CAPRA-S (0.75, 95% CI 0.65-0.85). RadClip resulted in a higher AUC (0.71, 95% CI 0.62-0.81) for predicting AP compared to Decipher (0.66, 95% CI 0.56-0.77) and CAPRA (0.69, 95% CI 0.59-0.79).”

Interpretation

RadClip was more prognostic of bRFS and AP compared to Decipher and CAPRA. It could help pre-operatively identify PCa patients at low risk of biochemical recurrence and AP and who therefore might defer additional therapy.

Funding

The National Institutes of Health, the U.S. Department of Veterans Affairs, and the Department of Defense.

Test-retest repeatability of a deep learning architecture in detecting and segmenting clinically significant prostate cancer on apparent diffusion coefficient (ADC) maps

OBJECTIVES

To evaluate short-term test-retest repeatability of a deep learning architecture (U-Net) in slice- and lesion-level detection and segmentation of clinically significant prostate cancer (csPCa: Gleason grade group > 1) using diffusion-weighted imaging fitted with monoexponential function, ADC_m.

METHODS

One hundred twelve patients with prostate cancer (PCa) underwent 2 prostate MRI examinations on the same day. PCa areas were annotated using whole mount prostatectomy sections. Two U-Net-based convolutional neural networks were trained on three different ADC_m b value settings for (a) slice- and (b) lesion-level detection and (c) segmentation of csPCa. Short-term test-retest repeatability was estimated using intra-class correlation coefficient (ICC(3,1)), proportionate agreement, and dice similarity coefficient (DSC). A 3-fold cross-validation was performed on training set (N = 78 patients) and evaluated for performance and repeatability on testing data (N = 34 patients).

RESULTS

For the three ADC_m b value settings, repeatability of mean ADC_m of csPCa lesions was ICC(3,1) = 0.86-0.98. Two CNNs with U-Net-based architecture demonstrated ICC(3,1) in the range of 0.80-0.83, agreement of 66-72%, and DSC of 0.68-0.72 for slice- and lesion-level detection and segmentation of csPCa. Bland-Altman plots suggest that there is no systematic bias in agreement between inter-scan ground truth segmentation repeatability and segmentation repeatability of the networks.

CONCLUSIONS

For the three ADC_m b value settings, two CNNs with U-Net-based architecture were repeatable for the problem of detection of csPCa at the slice-level. The network repeatability in segmenting csPCa lesions is affected by inter-scan variability and ground truth segmentation repeatability and may thus improve with better inter-scan reproducibility.

KEY POINTS

• For the three ADC_m b value settings, two CNNs with U-Net-based architecture were repeatable for the problem of detection of csPCa at the slice-level. • The network repeatability in segmenting csPCa lesions is affected by inter-scan variability and ground truth segmentation repeatability and may thus improve with better inter-scan reproducibility.

Repeatability of radiomics and machine learning for DWI: Short-term repeatability study of 112 patients with prostate cancer

PURPOSE

To evaluate repeatability of prostate DWI-derived radiomics and machine learning methods for prostate cancer (PCa) characterization.

METHODS

A total of 112 patients with diagnosed PCa underwent 2 prostate MRI examinations (Scan1 and Scan2) performed on the same day. DWI was performed using 12 b-values (0-2000 s/mm² ), post-processed using kurtosis function, and PCa areas were annotated using whole mount prostatectomy sections. A total of 1694 radiomic features including Sobel, Kirch, Gradient, Zernike Moments, Gabor, Haralick, CoLIAGe, Haar wavelet coefficients, 3D analogue to Laws features, 2D contours, and corner detectors were calculated. Radiomics and 4 feature pruning methods (area under the receiver operator characteristic curve, maximum relevance minimum redundancy, Spearman's ρ, Wilcoxon rank-sum) were evaluated in terms of Scan1-Scan2 repeatability using intraclass correlation coefficient (ICC)(3,1). Classification performance for clinically significant and insignificant PCa with Gleason grade groups 1 versus >1 was evaluated by area under the receiver operator characteristic curve in unseen random 30% data split.

RESULTS

The ICC(3,1) values for conventional radiomics and feature pruning methods were in the range of 0.28-0.90. The machine learning classifications varied between Scan1 and Scan2 with % of same class labels between Scan1 and Scan2 in the range of 61-81%. Surface-to-volume ratio and corner detector-based features were among the most represented features with high repeatability, ICC(3,1) >0.75, consistently high ranking using all 4 feature pruning methods, and classification performance with area under the receiver operator characteristic curve >0.70.

CONCLUSION

Surface-to-volume ratio and corner detectors for prostate DWI led to good classification of unseen data and performed similarly in Scan1 and Scan2 in contrast to multiple conventional radiomic features.

Association of radiomic features from prostate bi-parametric MRI with Decipher risk categories to predict risk for biochemical recurrence post-prostatectomy

e16561

Background: Decipher test examines expression levels of 22 metastasis-related RNA sequences and provides a risk score (DS) for assessing metastasis risk within 5 years after radical prostatectomy (RP). Our previous work has shown that radiomics or computer extracted texture features are associated with DS on multiparametric MRI (mpMRI). DS associated radiomics (RDS) may capture signatures related to poor prognosis of prostate cancer (PCa) following RP. Thus, we sought to evaluate whether RDS were also able to predict the risk of biochemical recurrence (BCR), a surrogate clinical endpoint of metastasis, after RP. Methods: Two patient cohorts from two institutions who underwent 3T mpMRI prior to RP were used in this study - I1: 70 PCa patients who had a Decipher test following RP; I2: 120 PCa patients with at least 5-year follow-up after RP. I1 was investigated as a training set to identify RDS and develop a multivariate logistic regression model (CL). I2 was used as a testing set to validate CL for predicting BCR. A total of 150 radiomics were extracted within each lesion from T2-weighted MRI (T2WI) and apparent diffusion coefficient (ADC) maps. Area under the receiver operating characteristic curve (AUC) and Kaplan-Meier (KM) analysis were used to evaluate performance of CL. Results: Radiomics quantifying lesion spot and ripple texture patterns on ADC maps and T2WI intensity gradient changes were able to differentiate Decipher high and low/intermediate groups in I1 (AUC = 0.92). The same model CL was then applied to I2 resulting in a significant difference in BCR-free survival time between the predicted risk groups (p < 0.05, hazard ratio (HR) = 2.54) independent of Gleason Grade Group (GGG, p < 0.05, HR = 7.61) and PIRADS-v2 (p < 0.05, HR = 5.64) in multivariate testing. Especially, for patients with GGG > 2 (n = 57), CL predicted BCR risk groups shown HR = 3.1 (p < 0.05) while GGG alone was not prognostic (p > 0.05, HR = 1.93). Conclusions: Our radiomic model was able to predict DS and was found to also be prognostic of BCR-free survival. Further work will involve evaluating whether these radiomic features are predictive of PCa metastasis. [Table: see text]

Computerized histomorphometric features of glandular architecture predict risk of biochemical recurrence following radical prostatectomy: A multisite study

5060

Background: Following a radical prostatectomy (RP) for prostate cancer, a patient may experience biochemical recurrence (BCR), defined as two consecutive prostate specific antigen (PSA) readings > 0.2 ng/mL. BCR is correlated with metastasis and disease specific survival. Extant molecular based companion diagnostic tests for predicting risk of BCR and disease progression tend to be expensive and tissue destructive. We sought to evaluate whether computer extracted features of glandular architecture from routine digitized H&E slides could predict post-RP BCR risk. Methods: RP specimens from 683 patients (184 with BCR, 499 without) with post-surgical PSA follow-up information were gathered from six sites. Median non-BCR follow-up was 3.2 years. A representative tumor area was annotated on the diagnostic H&E slide of each patient. 324 (131 BCR) patients from two sites formed the training set. The other 359 (53 BCR) patients formed the validation set. Glands were segmented by a deep learning model. 216 features describing gland arrangement, shape, and disorder were then extracted. An elastic net Cox proportional hazards model was constructed from the training set using the top 10 stable features identified via feature selection. Risk score thresholds were chosen on the training set to stratify patients into low-, medium-, or high-risk. Validation set results were evaluated by the log-rank test and hazard ratio. For the 172 (37 BCR) patients for whom Gleason grade and preoperative PSA values were available, risk classifications were compared using Cox proportional hazards regression. Results: Nine of the top features were gland shape features and one was a gland arrangement feature. The hazard ratio between the low- and high-risk groups on the validation set was 3.04 (p < 0.05). The histomorphometric classifier was predictive of BCR (p < 0.05, hazard ratio = 1.63) independent of Gleason grade group and preoperative PSA in multivariate testing. Conclusions: Computer extracted features of gland morphology can stratify post-RP patients by BCR risk. Our computerized histomorphometric model could serve as a prognostic tool in the post-RP setting.[Table: see text]

Radiomic features from pretreatment biparametric MRI predict prostate cancer biochemical recurrence: Preliminary findings

BACKGROUND

Radiomics or computer-extracted texture features derived from MRI have been shown to help quantitatively characterize prostate cancer (PCa). Radiomics have not been explored depth in the context of predicting biochemical recurrence (BCR) of PCa.

PURPOSE

To identify a set of radiomic features derived from pretreatment biparametric MRI (bpMRI) that may be predictive of PCa BCR.

STUDY TYPE

Retrospective.

SUBJECTS

In all, 120 PCa patients from two institutions, I₁ and I₂ , partitioned into training set D₁ (N = 70) from I₁ and independent validation set D₂ (N = 50) from I₂ . All patients were followed for ≥3 years.

SEQUENCE

3T, T₂ -weighted (T₂ WI) and apparent diffusion coefficient (ADC) maps derived from diffusion-weighted sequences.

ASSESSMENT

PCa regions of interest (ROIs) on T₂ WI were annotated by two experienced radiologists. Radiomic features from bpMRI (T₂ WI and ADC maps) were extracted from the ROIs. A machine-learning classifier (C_BCR ) was trained with the best discriminating set of radiomic features to predict BCR (p_BCR ).

STATISTICAL TESTS

Wilcoxon rank-sum tests with P < 0.05 were considered statistically significant. Differences in BCR-free survival at 3 years using p_BCR was assessed using the Kaplan-Meier method and compared with Gleason Score (GS), PSA, and PIRADS-v2.

RESULTS

Distribution statistics of co-occurrence of local anisotropic gradient orientation (CoLlAGe) and Haralick features from T₂ WI and ADC were associated with BCR (P < 0.05) on D₁ . C_BCR predictions resulted in a mean AUC = 0.84 on D₁ and AUC = 0.73 on D₂ . A significant difference in BCR-free survival between the predicted classes (BCR + and BCR-) was observed (P = 0.02) on D₂ compared to those obtained from GS (P = 0.8), PSA (P = 0.93) and PIRADS-v2 (P = 0.23).

DATA CONCLUSION

Radiomic features from pretreatment bpMRI can be predictive of PCa BCR after therapy and may help identify men who would benefit from adjuvant therapy.

LEVEL OF EVIDENCE

4 Technical Efficacy: Stage 5 J. Magn. Reson. Imaging 2018;48:1626-1636.

Radiomic features on MRI enable risk categorization of prostate cancer patients on active surveillance: Preliminary findings

BACKGROUND

Radiomic analysis is defined as computationally extracting features from radiographic images for quantitatively characterizing disease patterns. There has been recent interest in examining the use of MRI for identifying prostate cancer (PCa) aggressiveness in patients on active surveillance (AS).

PURPOSE

To evaluate the performance of MRI-based radiomic features in identifying the presence or absence of clinically significant PCa in AS patients.

STUDY TYPE

Retrospective.

SUBJECTS MODEL

MRI/TRUS (transperineal grid ultrasound) fusion-guided biopsy was performed for 56 PCa patients on AS who had undergone prebiopsy.

FIELD STRENGTH/SEQUENCE

3T, T2 -weighted (T2 w) and diffusion-weighted (DW) MRI.

ASSESSMENT

A pathologist histopathologically defined the presence of clinically significant disease. A radiologist manually delineated lesions on T2 w-MRs. Then three radiologists assessed MRIs using PIRADS v2.0 guidelines. Tumors were categorized into four groups: MRI-negative-biopsy-negative (Group 1, N = 15), MRI-positive-biopsy-positive (Group 2, N = 16), MRI-negative-biopsy-positive (Group 3, N = 10), and MRI-positive-biopsy-negative (Group 4, N = 15). In all, 308 radiomic features (First-order statistics, Gabor, Laws Energy, and Haralick) were extracted from within the annotated lesions on T2 w images and apparent diffusion coefficient (ADC) maps. The top 10 features associated with clinically significant tumors were identified using minimum-redundancy-maximum-relevance and used to construct three machine-learning models that were independently evaluated for their ability to identify the presence and absence of clinically significant disease.

STATISTICAL TESTS

Wilcoxon rank-sum tests with P < 0.05 considered statistically significant.

RESULTS

Seven T2 w-based (First-order Statistics, Haralick, Laws, and Gabor) and three ADC-based radiomic features (Laws, Gradient and Sobel) exhibited statistically significant differences (P < 0.001) between malignant and normal regions in the training groups. The three constructed models yielded overall accuracy improvement of 33, 60, 80% and 30, 40, 60% for patients in testing groups, when compared to PIRADS v2.0 alone.

DATA CONCLUSION

Radiomic features could help in identifying the presence and absence of clinically significant disease in AS patients when PIRADS v2.0 assessment on MRI contradicted pathology findings of MRI-TRUS prostate biopsies.

LEVEL OF EVIDENCE

3 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2018.

A combination of computer extracted measurements of prostate capsule shape and tumor texture on MRI to predict biochemical recurrence post treatment

e16554

Background: Prostate cancer (PCa) biochemical recurrence (BCR) occurs in a significant proportion of men after treatment and is associated with increased mortality. Identifying PCa patients at risk of BCR following definitive therapy may help identify patients who might benefit from adjuvant therapy. Multi-parametric MRI (mpMRI) is being increasingly used in pre-treatment staging and risk stratification of PCa. In this work, we sought to explore whether a combination of computer extracted features relating to prostate capsule shape and tumor texture from pre-treatment mpMRI is predictive of BCR post treatment. Methods: This single center and retrospective study included 80 men with PCa who underwent pre-treatment 3T mpMRI and were followed for > 3 years post treatment. These men were grouped into a training set D1 (N = 50, 25 each of BCR+ and BCR-) and an independent validation set D2 (N = 30, 10 BCR+ and 20 BCR-). The prostate capsule and cancer region of interest (ROI) were annotated on mpMRI by a single experienced radiologist. Shape features (curvature and orientation) were extracted from a surface of interest where statistically significant differences were observed between BCR+ and BCR- patients in the training set. Radiomic features for capturing tumor textural patterns (including 1st and 2nd order statistics, Gabor and Haralick) were extracted from within the radiologist annotated cancer ROIs. Features from D1 were used to train random forest classifiers, one each with shape (Cs) and radiomic (CR) features. A fused Bayesian classifier (CR+S) was created by integrating decisions from both Cs and CR. Results: The classifiers Cs, CR and CR+S were evaluated on independent validation set D2, resulting in area under the curve (AUCs) of 0.71, 0.81 and 0.84 respectively. Cs added value in patients who had extra prostatic spread of PCa, but suffered from mpMRI intensity artefacts that affected performance of CR. Conclusions: Integrating prostate capsule shape and tumor radiomic features from pre-treatment mpMRI enabled prediction of PCa BCR after treatment. Multi-site validation is needed to establish the robustness of the approach.

Computer extracted shape features of prostate capsule from MRI to predict biochemical recurrence of prostate cancer post-treatment

e16579

Background: Pre-treatment identification of biochemical recurrence (BCR) from MRI may enable the use of aggressive neo-adjuvant therapies for prostate cancer patients to improve prognosis. BCR is often associated with aggressive cancer growth and/or extra prostatic extension resulting in an irregular bulge and focal capsular retraction. This may induce differences in the shape of the prostate capsule between BCR positive (BCR+) and BCR negative (BCR-) patients as observed on MRI. In this work, we show that computer extracted shape features of the prostate capsule on MRI can identify patients that are at a risk of BCR post-treatment Methods: In a single centre IRB approved study, from a registry of 874 patients, availability of complete image datasets (T1w, T2w and ADC map); no treatment for PCa before MRI; presence of clinically localised PCa; availability of Gleason score; and data available for post-treatment PSA and follow-up for at least 3 years in patients without BCR were used as inclusion criteria to select 80 patients. The prostate capsule was manually segmented on T2w MRI by an experienced radiologist. Two atlases A+ and A- were created for BCR+ and BCR- patients respectively with similar Gleason score (6 to 9), similar numbers in each cohort (25 each) and similar tumor stage (T2 to T3). A t-test based analysis corrected for multiple comparison revealed statistically significant prostate shape differences between A+ and A- in surface of interest (SOI). Curvature features (magnitude and surface normal orientations) were extracted from SOI of the two cohorts. A random forest classifier was trained using the 50 training images (from A+ and A-) and validated using a hold-out validation set of 30 patients. Results: The inter-quartile range, variance, skewness and kurtosis of curvature magnitude and normal orientations were found to be predictive of BCR. The RF classifier trained using these features could predict BCR with an accuracy of 78% and an AUC of 0.71 in the validation set. Conclusions: Curvature magnitude and orientation features of the prostate capsule from the SOI may be predictive of BCR. In future a multi centre independent datasets will be used to validate the findings.

Radiomics based targeted radiotherapy planning (Rad-TRaP): a computational framework for prostate cancer treatment planning with MRI

BACKGROUND

Radiomics or computer - extracted texture features have been shown to achieve superior performance than multiparametric MRI (mpMRI) signal intensities alone in targeting prostate cancer (PCa) lesions. Radiomics along with deformable co-registration tools can be used to develop a framework to generate targeted focal radiotherapy treatment plans.

METHODS

The Rad-TRaP framework comprises three distinct modules. Firstly, a module for radiomics based detection of PCa lesions on mpMRI via a feature enabled machine learning classifier. The second module comprises a multi-modal deformable co-registration scheme to map tissue, organ, and delineated target volumes from MRI onto CT. Finally, the third module involves generation of a radiomics based dose plan on MRI for brachytherapy and on CT for EBRT using the target delineations transferred from the MRI to the CT.

RESULTS

Rad-TRaP framework was evaluated using a retrospective cohort of 23 patient studies from two different institutions. 11 patients from the first institution were used to train a radiomics classifier, which was used to detect tumor regions in 12 patients from the second institution. The ground truth cancer delineations for training the machine learning classifier were made by an experienced radiation oncologist using mpMRI, knowledge of biopsy location and radiology reports. The detected tumor regions were used to generate treatment plans for brachytherapy using mpMRI, and tumor regions mapped from MRI to CT to generate corresponding treatment plans for EBRT. For each of EBRT and brachytherapy, 3 dose plans were generated - whole gland homogeneous ([Formula: see text]) which is the current clinical standard, radiomics based focal ([Formula: see text]), and whole gland with a radiomics based focal boost ([Formula: see text]). Comparison of [Formula: see text] against conventional [Formula: see text] revealed that targeted focal brachytherapy would result in a marked reduction in dosage to the OARs while ensuring that the prescribed dose is delivered to the lesions. [Formula: see text] resulted in only a marginal increase in dosage to the OARs compared to [Formula: see text]. A similar trend was observed in case of EBRT with [Formula: see text] and [Formula: see text] compared to [Formula: see text].

CONCLUSIONS

A radiotherapy planning framework to generate targeted focal treatment plans has been presented. The focal treatment plans generated using the framework showed reduction in dosage to the organs at risk and a boosted dose delivered to the cancerous lesions.