Machine learning classification of diagnostic accuracy in pathologists interpreting breast biopsies - Research study

Aug 5, 2026

Upload by Oliver

Objective: This study evaluates whether machine learning can predict diagnostic accuracy by analyzing pathologists’ spatiotemporal viewing behavior while interpreting digital breast biopsy whole-slide images. Materials and Methods: Data were collected from 140 pathologists with varying experience who reviewed 14 digital whole-slide images; zooming and panning actions were recorded, 30 viewing features extracted, and four classifiers trained. Results: Random Forest achieved the best performance with 0.81 test accuracy and 0.86 AUC, with attention distribution and critical-region focus as key predictors. Incorporating case- and pathologist-level information further improved performance. The findings support automated feedback and decision support for training and clinical use.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Objective

Materials and Methods

Results

Discussion

Conclusion

Background and significance

UCLA UCLA Previously Published Works Title Machine learning classification of diagnostic accuracy in pathologists interpreting breast biopsies. Permalink [https://escholarship.org/uc/item/1d62s5j9](https://escholarship.org/uc/item/1d62s5j9) Journal A Scholarly Journal of Informatics in Health and Biomedicine, 31(3) Authors Brunyé, Tad Booth, Kelsey Hendel, Dalitet al. Publication Date 2024-02-16 DOI 10.1093/jamia/ocad232 Peer reviewed [eScholarship.org](eScholarship.org) Powered by the California Digital Library University of California Research and Applications Machine learning classification of diagnostic accuracy in pathologists interpreting breast biopsies Tad T. Bruny, PhD1,2, 􀀃 , Kelsey Booth, MS1, Dalit Hendel, MS1, Kathleen F. Kerr , PhD3, Hannah Shucard, MS3, Donald L. Weaver, MD4, Joann G. Elmore, MD5 1Center for Applied Brain and Cognitive Sciences, Tufts University, Medford, MA 02155, United States, 2Department of Psychology, Tufts University, Medford, MA 02155, United States, 3Department of Biostatistics, University of Washington, Seattle, WA 98105, United States, 4Department of Pathology and Laboratory Medicine, Larner College of Medicine, University of Vermont and Vermont Cancer Center, Burlington, VT 05405, United States, 5Department of Medicine, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, CA 90095, United States 􀀃 Corresponding author: Tad T. Bruny, PhD, Center for Applied Brain and Cognitive Sciences, Tufts University, 177 College Ave., Suite 090, Medford, MA 02155 ([tbruny01@tufts.edu](tbruny01@tufts.edu)) Abstract Objective: This study explores the feasibility of using machine learning to predict accurate versus inaccurate diagnoses made by pathologists based on their spatiotemporal viewing behavior when evaluating digital breast biopsy images. Materials and Methods: The study gathered data from 140 pathologists of varying experience levels who each reviewed a set of 14 digital whole slide images of breast biopsy tissue. Pathologists’ viewing behavior, including zooming and panning actions, was recorded during image evaluation. A total of 30 features were extracted from the viewing behavior data, and 4 machine learning algorithms were used to build classifiers for predicting diagnostic accuracy. Results: The Random Forest classifier demonstrated the best overall performance, achieving a test accuracy of 0.81 and area under the receiver-operator characteristic curve of 0.86. Features related to attention distribution and focus on critical regions of interest were found to be important predictors of diagnostic accuracy. Further including case-level and pathologist-level information incrementally improved classifier performance. Discussion: Results suggest that pathologists’ viewing behavior during digital image evaluation can be leveraged to predict diagnostic accuracy, affording automated feedback and decision support systems based on viewing behavior to aid in training and, ultimately, clinical practice. They also carry implications for basic research examining the interplay between perception, thought, and action in diagnostic decision-making. Conclusion: The classifiers developed herein have potential applications in training and clinical settings to provide timely feedback and support to pathologists during diagnostic decision-making. Further research could explore the generalizability of these findings to other medical domainsand varied levels of expertise. Key words: breast pathology; medical education; medical residency training; machine learning; diagnostic decision-making; medical image interpretation; diagnostic accuracy. Background and significance Over 1 million breast biopsies are estimated to occur annually in the United States and are interpreted and diagnosed by pathologists.1–3 Accurate pathological diagnosis of biopsy tissue is the linchpin for appropriate patient care, yet the perceptual an

FAQs

What is the objective of the study?

The study explores whether machine learning can predict whether pathologists’ diagnoses are accurate or inaccurate using their spatiotemporal viewing behavior during digital breast biopsy interpretation.

How was the dataset collected and processed?

Viewing behavior was recorded from 140 pathologists as they reviewed 14 digital whole-slide images. Zooming and panning actions were converted into 30 extracted features used for classifier training.

Which model performed best and what features mattered most?

The Random Forest classifier performed best, with 0.81 test accuracy and 0.86 AUC. Predictors included features related to attention distribution and focusing on critical regions of interest.

Aug 5, 2026

Upload by Oliver