Selective identification and localization of indolent and aggressive prostate cancers via CorrSigNIA: an MRI-pathology correlation and deep learning framework

被引：43

作者：

Bhattacharya, Indrani ^{[1
,4
]}

Seetharaman, Arun ^{[2
]}

Kunder, Christian ^{[3
]}

Shao, Wei ^{[1
]}

Chen, Leo C. ^{[4
]}

Soerensen, Simon J. C. ^{[4
,5
]}

Wang, Jeffrey B. ^{[1
]}

Teslovich, Nikola C. ^{[4
]}

Fan, Richard E. ^{[4
]}

Ghanouni, Pejman ^{[1
,4
]}

Brooks, James D. ^{[4
]}

Sonn, Geoffrey A. ^{[1
,4
]}

Rusu, Mirabela ^{[1
]}

机构：

[1] Stanford Univ, Dept Radiol, 300 Pasteur Dr, Stanford, CA 94305 USA

[2] Stanford Univ, Dept Elect Engn, 350 Jane Stanford Way, Stanford, CA 94305 USA

[3] Stanford Univ, Dept Pathol, 300 Pasteur Dr, Stanford, CA 94305 USA

[4] Stanford Univ, Dept Urol, 300 Pasteur Dr, Stanford, CA 94305 USA

[5] Aarhus Univ Hosp, Dept Urol, Aarhus, Denmark

来源：

MEDICAL IMAGE ANALYSIS | 2022年 / 75卷

基金：

美国国家卫生研究院;

关键词：

Prostate cancer; Computer-aided diagnosis; Correlated feature learning; Radiology-pathology fusion; REGISTRATION; IMAGES; BIOPSY;

D O I：

10.1016/j.media.2021.102288

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

Automated methods for detecting prostate cancer and distinguishing indolent from aggressive disease on Magnetic Resonance Imaging (MRI) could assist in early diagnosis and treatment planning. Existing automated methods of prostate cancer detection mostly rely on ground truth labels with limited accu-racy, ignore disease pathology characteristics observed on resected tissue, and cannot selectively identify aggressive (Gleason Pattern >= 4) and indolent (Gleason Pattern = 3) cancers when they co-exist in mixed lesions. In this paper, we present a radiology-pathology fusion approach, CorrSigNIA, for the selective identification and localization of indolent and aggressive prostate cancer on MRI. CorrSigNIA uses regis-tered MRI and whole-mount histopathology images from radical prostatectomy patients to derive accu-rate ground truth labels and learn correlated features between radiology and pathology images. These correlated features are then used in a convolutional neural network architecture to detect and localize normal tissue, indolent cancer, and aggressive cancer on prostate MRI. CorrSigNIA was trained and val-idated on a dataset of 98 men, including 74 men that underwent radical prostatectomy and 24 men with normal prostate MRI. CorrSigNIA was tested on three independent test sets including 55 men that underwent radical prostatectomy, 275 men that underwent targeted biopsies, and 15 men with normal prostate MRI. CorrSigNIA achieved an accuracy of 80% in distinguishing between men with and without cancer, a lesion-level ROC-AUC of 0.81 +/- 0.31 in detecting cancers in both radical prostatectomy and biopsy cohort patients, and lesion-levels ROC-AUCs of 0.82 +/- 0.31 and 0.86 +/- 0.26 in detecting clinically significant cancers in radical prostatectomy and biopsy cohort patients respectively. CorrSigNIA consistently outper-formed other methods across different evaluation metrics and cohorts. In clinical settings, CorrSigNIA may be used in prostate cancer detection as well as in selective identification of indolent and aggressive com-ponents of prostate cancer, thereby improving prostate cancer care by helping guide targeted biopsies, reducing unnecessary biopsies, and selecting and planning treatment. (C) 2021 The Authors. Published by Elsevier B.V.

引用

页数：16