Researchers from Google Health found that using artificial intelligence (AI) to aid in the review of prostate biopsies improved the quality, efficiency, and consistency of cancer detection and grading.
In a prostate biopsy, tissue is removed and assessed for cell abnormalities that may be linked to prostate cancer. The standard grading system for this procedure is the Gleason grade (GG) system, involving classification into 1 of 5 prognostic groups. Expert-level AI algorithms for prostate biopsy grading, like this one from Gooogle Health, have recently been developed to combat interpathologist variability associated with grading.
In this diagnostic study, retrospective grading of prostate core needle biopsies was conducted at two medical laboratories in the US between October 2019 and January 2020. Twenty general pathologists reviewed prostate core needle biopsies from 240 patients. Pathologists were randomized to one of two cohorts. Each biopsy was reviewed by both cohorts in the opposite modality (with vs without AI assistance), with cohort modalities switching every 10 cases. After a minimum 4-week period for each batch, pathologists reviewed the cases for a second time using the opposite modality. The resulting grade group for each biopsy was compared with the majority opinion from urology subspecialists.
AI-assisted review was associated with a 5.6% increase (95% confidence interval [CI], 3.2%-7.9%; P < .001) in agreement with subspecialists (69.7% for unassisted reviews vs 75.3% for assisted reviews). Grade group 1 biopsies were associated with a 6.2% increase (95% CI, 2.7%-9.8%; P = .001) in agreement with subspecialists (72.3% unassisted vs 78.5% assisted). Follow-up analysis showed that AI assistance was also linked to improvements in tumor detection, mean review time, mean self-reported confidence, and interpathologist agreement.
“This diagnostic study indicated the potential ability of an AI-based assistive tool to improve the accuracy, efficiency, and consistency of prostate biopsy review by pathologists,” the authors concluded. “Additional efforts to optimize clinical workflow integration and to conduct prospective evaluation of AI-based tools in clinical settings remain important future directions.”
This study was published in JAMA Network Open.
— Andrew A. Borkowski (@tampapath) November 15, 2020