A Deep Learning Approach to Diagnostic Classification of Prostate Cancer Using Pathology–Radiology Fusion

Abstract

Background: A definitive diagnosis of prostate cancer requires a biopsy to obtain tissue for pathologic analysis, but this is an invasive procedure and is associated with complications. Purpose: To develop an artificial intelligence (AI)-based model (named AI-biopsy) for the early diagnosis of prostate cancer using magnetic resonance (MR) images labeled with histopathology information. Study Type: Retrospective. Population: Magnetic resonance imaging (MRI) data sets from 400 patients with suspected prostate cancer and with histological data (228 acquired in-house and 172 from external publicly available databases). Field Strength/Sequence: 1.5 to 3.0 Tesla, T2-weighted image pulse sequences. Assessment: MR images reviewed and selected by two radiologists (with 6 and 17 years of experience). The patient images were labeled with prostate biopsy including Gleason Score (6 to 10) or Grade Group (1 to 5) and reviewed by one pathologist (with 15 years of experience). Deep learning models were developed to distinguish 1) benign from cancerous tumor and 2) high-risk tumor from low-risk tumor. Statistical Tests: To evaluate our models, we calculated negative predictive value, positive predictive value, specificity, sensitivity, and accuracy. We also calculated areas under the receiver operating characteristic (ROC) curves (AUCs) and Cohen's kappa. Results: Our computational method (https://github.com/ih-lab/AI-biopsy) achieved AUCs of 0.89 (95% confidence interval [CI]: [0.86-0.92]) and 0.78 (95% CI: [0.74-0.82]) to classify cancer vs. benign and high-vs. low-risk of prostate disease, respectively. Data Conclusion: AI-biopsy provided a data-driven and reproducible way to assess cancer risk from MR images and a personalized strategy to potentially reduce the number of unnecessary biopsies. AI-biopsy highlighted the regions of MR images that contained the predictive features the algorithm used for diagnosis using the class activation map method. It is a fully automatic method with a drag-and-drop web interface (https://ai-biopsy.eipm-research.org) that allows radiologists to review AI-assessed MR images in real time. Level of Evidence: 1 Technical Efficacy Stage: 2

Automatic Classification of Cancer Pathology Reports: A Systematic Review

“…Listed in ascending ordered according to the year of publication 2021–2010. Citation Year Methodology Prediction Task Data Evaluation Khosravi et al 41 2021 Deep learning Classify cancer vs. benign and high vs. low-risk of prostate disease Local urology center database of 400 prostate cancer MRI images and pathology re-ports AUCs of 0.89 and 0.78 for classification of cancer vs benign and high vs low risk, respectively Gao et al 50 2020 Hierarchical deep learning Six cancer classification tasks: site, subsite, laterality , histology, behavior, and grade 546, 806 cancer (all types) pathology re-ports obtained from the SEER cancer registry program F1 Micro of 0.92, 0.64, 0.92, 0.8, 0.98, and 0.82 for site, subsite, laterality, histology, behavior, and grade, respectively Saib et al 51 2020 Hierarchical deep learning Classify 9 ICD-O morphology grading 1813 breast cancer pathology reports obtained from a local center database F1 Micro of 0.91 and F1 Macro of 0.69 for classification of 9 ICD-O codes Alawad et al 43 2020 Deep learning Two cancer classification tasks: subsite with 317 labels and histology with 556 labels 878,864 cancer (all types) pathology reports obtained from the SEER cancer registry program F1 Micro of 0.68 for subsite; F1 Micro of 0.79 for histology Glaser et al 22 2019 Rule-Based Extract stage, grade, and presence of muscularis propria 3,042 Transurethral Resection of the Bladder Tumor (TURBT) reports obtained from a local database Accuracy of 82%, 88% , and 100% for extracting stage, specimens and grade, respectively Soysal et al 23 2019 Rule-based Extract cancer-related information in pathology reports (e.g., tumor size, tumor stage, specimen, biomarkers, and others) 400 cancer (all types) pathology reports obtained from a local center database F1 average performance ranging from 0.87 to 0.99 for extracting cancer information Yoon et al 40 2019 Multi-task deep learning ...…”

“…We have found that many of the most recent NLP application in pathology are designed based on deep learning models and use a distributed representation of words to represent the data input. 37 , 38 , 39 , 40 , 41 John et al 38 investigated the usage of convolutional neural networks (CNN) 42 to classify ICD codes from breast and lung cancer pathology reports. The authors reported their best system with a F1-micro score of 0.722 over 12 ICD-O-3 topography codes.…”

MRI and Targeted Biopsy Essential Tools for an Accurate Diagnosis and Treatment Decision Making in Prostate Cancer

“…A retrospective study of 64 patients with PCa found 14 radiomics features that correlated with the Gleason score and 31 histogram and texture characteristics that correlated with different gene signatures. Although prospective trials are required, radiomics features with machine learning prediction models are promising markers for cancer aggressiveness [ 65 ].…”

Editorial for “A Deep Learning Approach to Diagnostic Classification of Prostate Cancer Using Pathology‐Radiology Fusion”

“…9 In this issue of Journal of Magnetic Resonance Imaging, Khosravi et al developed an AI-based model (named AIbiopsy) for the early detection and risk assessment of PCa. 10 This AI-biopsy technology used a CNN algorithm with transfer learning approach for prostate T2-weighted images to perform classifications. A total of 400 patients from multiple centers were included to construct and test the model.…”