Published October 20, 2023 | Version Published
Journal Article Open

Focalizing regions of biomarker relevance facilitates biomarker prediction on histopathological images

Abstract

Image-based AI has thrived as a potentially revolutionary tool for predicting molecular biomarker statuses, which aids in categorizing patients for appropriate medical treatments. However, many methods using hematoxylin and eosin-stained (H&E) whole-slide images (WSIs) have been found to be inefficient because of the presence of numerous uninformative or irrelevant image patches. In this study, we introduced the region of biomarker relevance (ROB) concept to identify the morphological areas most closely associated with biomarkers for accurate status prediction. We actualized this concept within a framework called saliency ROB search (SRS) to enable efficient and effective predictions. By evaluating various lung adenocarcinoma (LUAD) biomarkers, we showcased the superior performance of SRS compared to current state-of-the-art AI approaches. These findings suggest that AI tools, built on the ROB concept, can achieve enhanced molecular biomarker prediction accuracy from pathological images.

Copyright and License

© 2023 The Author(s). Under a Creative Commons license Attribution-NonCommercial-NoDerivs 4.0 International.

Acknowledgement

The authors would like to thank the two anonymous reviewers for their valuable comments and suggestions. This work was supported by China Hainan Provincial Major Science and Technology Project ZDKJ2021028 (D.Y.), The University of Texas MD Anderson Lung Moon Shot Program, The University of Texas MD Anderson Cancer Center Core Grant P30 CA01667, the National Institutes of Health (NIH) grant R00CA218667, R01CA234629, the AACR-Johnson & Johnson Lung Cancer Innovation Science Grant (18-90-52-ZHAN), the Rexanna's Foundation for Fighting Lung Cancer, Sabin Family Fund, Rydin Family Research Fund.

Contributions

J.G., H.W., X.W., Y.C., G.W., and T.X. conceived and designed the study. J.G., H.W., Y.C., and T.X. did the literature search. Y.C., W.Z., M.W., D.Y., Y.C., G.C., and G.W. contributed to lung adenocarcinoma subtype annotation and immunohistochemistry (IHC) on PD-L1 biomarker. J.G., Y.C., and H.W. preprocessed the data. J.G. and H.W. did the deep learning model development and performance evaluation. J.G. analyzed and interpreted the data, and drafted the manuscript. H.W., Z.H., K.M., H.Z., Y.B., Z.Z., A.Y., B.X., J.L., X.G., C.C., J.W., J.Z., and T.X. critically revised the manuscript. All authors had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Data Availability

  • Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request.
  • The original code will be available at https://github.com/ganjf/biomarkerPrediction.
  • This paper used both existing LUAD WSI from TCGA, CPTAC and NLST datasets and newly acquired LUAD WSI from WHTJ dataset to train and validate the computional framework. The accession numbers for theses publicy available datasets of TCGA, CPTAC and NLST are listed in the key resources table. The newly acquired WHTJ dataset will be shared by the lead contact upon request.

Conflict of Interest

The authors declare no competing interest.

Files

1-s2.0-S2589004223013202-main.pdf

Files (10.4 MB)

Name Size Download all
md5:b5955039fb2c40383c161f1149c50acf
4.6 MB Preview Download
md5:6a8136cce2f11410912f1bcb369e1358
5.9 MB Preview Download

Additional details

Identifiers

Funding

The University of Texas MD Anderson Cancer Center
National Institutes of Health
P30 CA01667
National Institutes of Health
R00CA218667
National Institutes of Health
R01CA234629
American Association For Cancer Research
18-90-52-ZHAN