Dave Steiner MD PhD Clinical Research Scientist Google Health, Palo Alto, California

AI Improves Accuracy of Prostate Cancer Grading

MedicalResearch.com Interview with:

Dave Steiner MD PhD Clinical Research Scientist Google Health, Palo Alto, California

Dr. Steiner

Dave Steiner MD PhD
Clinical Research Scientist
Google Health, Palo Alto, California

MedicalResearch.com: What is the background for this study?

Response: For prostate cancer patients, the grading of cancer in prostate biopsies by pathologists is central to risk stratification and treatment decisions. However, the grading process can be subjective, often resulting in variability among pathologists. This variability can complicate diagnostic and treatment decisions. As an initial step towards addressing this problem, we and others in the field have recently developed artificial intelligence (AI) algorithms that perform on-par with expert pathologists for prostate cancer grading. Such algorithms have the potential to improve the quality and efficiency of prostate biopsy grading, but the impact of these algorithms when used by pathologists has not been well studied. In the current study, we developed and evaluated an AI-based assistant tool for use by pathologists while reviewing prostate biopsies.

MedicalResearch.com: What are the main findings? 

Response: In our study, use of the AI-assistant tool improved the accuracy and consistency of prostate cancer grading by pathologists. Grading agreement of pathologists with urologic pathologists (prostate cancer experts) increased from 70% to 75%, and time spent per biopsy decreased by approximately 13% on average.

MedicalResearch.com: What should readers take away from your report?

Response: This study illustrates a successful example of combining AI and pathologists to develop effective AI-based assistance in pathology. Thoughtful clinical and human-computer interaction research was required to design a tool that is able to help the user know when to rely on the AI interpretations versus when to rely on their own expertise.

At the same time, we note that this was a controlled research study using retrospective cases. Understanding and optimizing the impact of AI-assistance in routine clinical workflows remains an important and exciting next step. 

MedicalResearch.com: What recommendations do you have for future research as a result of this work?

Response: In addition to clinical workflow integration, understanding and minimizing the impact of AI “errors” is an important area for future research. While we found that AI-assistance increased overall accuracy and that pathologists were often able to effectively ignore AI errors, this was not always the case. Understanding how AI errors interact with pathologist uncertainty and further developing strategies to minimize overreliance will increase the value and utility of AI in pathology.

This study also used the grading of urologic subspecialist pathologists as the reference standard, but disagreement even among experts remains a challenge in the field. Thus understanding if AI-assistance can lead to grading that more accurately predicts patient outcomes remains another important future research topic.

For additional information you can also see our related research blog post.

Disclosures: Many of the authors of this study are Google employees and own Alphabet stock.

Citation:

Steiner DF, Nagpal K, Sayres R, et al. Evaluation of the Use of Combined Artificial Intelligence and Pathologist Assessment to Review and Grade Prostate Biopsies. JAMA Netw Open. 2020;3(11):e2023267. doi:10.1001/jamanetworkopen.2020.23267

[subscribe]

Last Modified: [last-modified]

The information on MedicalResearch.com is provided for educational purposes only, and is in no way intended to diagnose, cure, or treat any medical or other condition. Always seek the advice of your physician or other qualified health and ask your doctor any questions you may have regarding a medical condition. In addition to all other limitations and disclaimers in this agreement, service provider and its third party providers disclaim any liability or loss in connection with the content provided on this website.

 

Last Updated on November 14, 2020 by Marie Benz MD FAAD