An interpretable deep learning model for detecting BRCA pathogenic variants of breast cancer from hematoxylin and eosin-stained pathological images

View article
Bioinformatics and Genomics

Main article text

 

Introduction

Materials and Methods

Patient selection

Image preprocessing and sample preparation

Segmentation network

Prediction network

Tissue features visualization and cell features quantitative analysis

Statistical analysis

Results

Clinical characteristics of patients

Performance of the segmentation network

Performance of BiAMIL network

Interpretability analysis of tissue features

Interpretability analysis of cell features

Discussion

Supplemental Information

Raw data: Patient details from Table 1.

DOI: 10.7717/peerj.18098/supp-1

Deep learning algorithm.

DOI: 10.7717/peerj.18098/supp-3

Raw data: predicted probability of slide in BiAMIL model.

DOI: 10.7717/peerj.18098/supp-4

Raw data:predicted probability of slid in ResNet34 model.

DOI: 10.7717/peerj.18098/supp-5

The performance of the BiAMIL model in the five-fold cross-validation in the training set.

DOI: 10.7717/peerj.18098/supp-6

The performance of the ResNet 34 model in the five-fold cross-validation in the training set.

DOI: 10.7717/peerj.18098/supp-7

STARD-Checklist.

DOI: 10.7717/peerj.18098/supp-8

Performance of the color normalization network.

(A). Workflow of sparse stain matrix decomposition constraint Cycle GAN color normalization model (SDCC-GAN). (B). Representative tiles of the color normalization network. The original tiles are in the first and second rows; The corresponding color-normalized tiles are in the third and fourth rows.

DOI: 10.7717/peerj.18098/supp-9

(A). The workflow of 5-fold cross-validation. (B). The loss curve of the model during the 5-fold cross-validation.

DOI: 10.7717/peerj.18098/supp-10

Additional Information and Declarations

Competing Interests

The authors declare that they have no competing interests.

Author Contributions

Yi Li conceived and designed the experiments, performed the experiments, analyzed the data, prepared figures and/or tables, authored or reviewed drafts of the article, and approved the final draft.

Xiaomin Xiong conceived and designed the experiments, analyzed the data, prepared figures and/or tables, authored or reviewed drafts of the article, and approved the final draft.

Xiaohua Liu performed the experiments, analyzed the data, authored or reviewed drafts of the article, and approved the final draft.

Yihan Wu analyzed the data, prepared figures and/or tables, and approved the final draft.

Xiaoju Li analyzed the data, prepared figures and/or tables, and approved the final draft.

Bo Liu performed the experiments, prepared figures and/or tables, and approved the final draft.

Bo Lin conceived and designed the experiments, performed the experiments, analyzed the data, authored or reviewed drafts of the article, and approved the final draft.

Yu Li conceived and designed the experiments, analyzed the data, prepared figures and/or tables, and approved the final draft.

Bo Xu conceived and designed the experiments, performed the experiments, authored or reviewed drafts of the article, and approved the final draft.

Human Ethics

The following information was supplied relating to ethical approvals (i.e., approving body and any reference numbers):

The study received approval from the Ethics Committee of Chongqing University Cancer Hospital (Ethics number: CZLS2023213-A).

Data Availability

The following information was supplied regarding data availability:

The code and raw measurements are available in the Supplemental Files.

The source code and pre-trained models are available at GitHub and Zenodo:

- https://github.com/LIYI0720/BiAMIL.

- LIYI0720. (2024). LIYI0720/BiAMIL: code to BiAMIL (BiAMIL). Zenodo. https://doi.org/10.5281/zenodo.13638785.

Funding

This work was supported by the National Natural Science Foundation of China (81974464, 61906022), the Chongqing Natural Science Foundation (cstc2020jcyj-msxmX0482), the Chongqing University Research Fund (2021CDJXKJC004), and the Chongqing Technology Innovation and Application Development Project (CSTB2023TIAD-KPX0050). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

647 Visitors 654 Views 27 Downloads