Development and Validation of a Deep Learning Algorithm for Gleason Grading of Prostate Cancer From Biopsy Specimens

Abstract

IMPORTANCE For prostate cancer, Gleason grading of the biopsy specimen plays a pivotal role in determining case management. However, Gleason grading is associated with substantial interobserver variability, resulting in a need for decision support tools to improve the reproducibility of Gleason grading in routine clinical practice. OBJECTIVE To evaluate the ability of a deep learning system (DLS) to grade diagnostic prostate biopsy specimens. DESIGN, SETTING, AND PARTICIPANTS The DLS was evaluated using 752 deidentified digitized images of formalin-fixed paraffin-embedded prostate needle core biopsy specimens obtained from 3 institutions in the United States, including 1 institution not used for DLS development. To obtain the Gleason grade group (GG), each specimen was first reviewed by 2 expert urologic subspecialists from a multi-institutional panel of 6 individuals (years of experience: mean, 25 years; range, 18-34 years). A third subspecialist reviewed discordant cases to arrive at a majority opinion. To reduce diagnostic uncertainty, all subspecialists had access to an immunohistochemical-stained section and 3 histologic sections for every biopsied specimen. Their review was conducted from December 2018 to June 2019. MAIN OUTCOMES AND MEASURES The frequency of the exact agreement of the DLS with the majority opinion of the subspecialists in categorizing each tumor-containing specimen as 1 of 5 categories: nontumor, GG1, GG2, GG3, or GG4-5. For comparison, the rate of agreement of 19 general pathologists' opinions with the subspecialists' majority opinions was also evaluated. RESULTS For grading tumor-containing biopsy specimens in the validation set (n = 498), the rate of agreement with subspecialists was significantly higher for the DLS (71.7%; 95% CI, 67.9%-75.3%) than for general pathologists (58.0%; 95% CI, 54.5%-61.4%) (P < .001). In subanalyses of biopsy specimens from an external validation set (n = 322), the Gleason grading performance of the DLS remained similar. For distinguishing nontumor from tumor-containing biopsy specimens (n = 752), the rate of agreement with subspecialists was 94.3% (95% CI, 92.4%-95.9%) for the DLS and similar at 94.7% (95% CI, 92.8%-96.3%) for general pathologists (P = .58). CONCLUSIONS AND RELEVANCE In this study, the DLS showed higher proficiency than general pathologists at Gleason grading prostate needle core biopsy specimens and generalized to an independent institution. Future research is necessary to evaluate the potential utility of using the DLS as a decision support tool in clinical workflows and to improve the quality of prostate cancer grading for therapy decisions.

An Artificial Intelligence–based Support Tool for Automation and Standardisation of Gleason Grading in Prostate Biopsies

“…In this study, we present an algorithm for automated Gleason grading that could be used as a screening tool, improving the accuracy of PCa diagnosis and reducing pathologists' workload. Several studies on this subject have been published in recent years [10,12,[15][16][17]20], indicating that there is a strong interest and need for the development of such tools. The present study is the first to evaluate an AI algorithm on a cohort scanned on two different slide scanners, comparing their results pixel by pixel.…”

“…We demonstrate its robustness by obtaining very similar results on scans from two different scanners. Nagpal et al [17] show encouraging results but lack testing on different scanners. Ström et al [15] tested their algorithm on an external set scanned on a different scanner, but the slides in this set differed from those used in the testing cohort.…”

“…Initial studies on PCa tissue were aimed at distinguishing cancer from noncancer regions [5][6][7]. In recent years, more studies on automated Gleason grading have emerged, using image patches [8,9], tissue microarrays [10,11], and prostatectomy specimens [12] as well as biopsies [4,[13][14][15][16][17], but all have some limitations such as small training datasets, no separate testing sets, training on un-annotated images, and lack of testing on images from different scanners. Several groups have tested external cohorts scanned on different scanners, but, to the best of our knowledge, no one has compared the same slides on different scanners.…”

Artificial Intelligence & Tissue Biomarkers: Advantages, Risks and Perspectives for Pathology

“…This is also confirmed by the increasing number of papers on the topic. Nagpal et al [ 34 ] developed a DL algorithm using as reference a group of expert uropathologists, and reported that their model performed significantly better than general pathologists on tumor grading in prostatic biopsy (71.7% versus 58.0%; p < 0.001). Interestingly enough, the AI algorithm achieved the same results of general pathologists in the recognizing the presence of tumor (94.3% for the AI and 94.7% for general pathologists).…”

Grading Evolution and Contemporary Prognostic Biomarkers of Clinically Significant Prostate Cancer

“…The use of proportionate grading is becoming easier with better access to image repositories and related data, as well as increased computational ability for complex artificial intelligence (AI) models. Research has shown that the accuracy and consistency of AI algorithms, when it comes to grading, are similar or even better than that of the general pathologists [ 23 , 24 , 25 ]. AI could make cancer detection and grade assessment more efficient and less prone to human error in the future.…”