1234P - HistoMate: Automated preprocessing software for digital histopathology image to enhance deep learning

Background

Recently, the field of digital pathology has seen an increase in research utilizing deep learning techniques. This allows for predictions such as cancer diagnosis, prognosis, and treatment response, which previously required additional analysis, such as genetic analysis that was both time-consuming and costly, to be made based on pathological imaging. However, digital pathology images often contain noise, such as inconsistent staining and inconsistent annotations. To address this issue, image preprocessing is essential in deep learning analysis of digital pathology images, and various preprocessing techniques have been proposed to address issues such as bias, overfitting, and robust deep learning model development. However, automated preprocessing methods for digital pathology images have not yet been fully developed.

Methods

To address this, we have developed a user-friendly tool for image preprocessing analysis called HistoMate.

Results

HistoMate provides GUI-based image segmentation, image tiling, color normalization, and deep learning-based data augmentation to automate the preprocessing process. It also provides functionality to evaluate image quality and select appropriate patches.

Conclusions

In conclusion, HistoMate provides an automated preprocessing tool for pathological image-based research, accelerating digital pathology-based research.

Clinical trial identification

Editorial acknowledgement

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean Government (MSIT) (grant no. NRF-2021R1C1C1013706), and research fund by Seoul National University Bundang Hospital (grant no. 14-2018-0013).

Legal entity responsible for the study

The authors.

Funding

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean Government (MSIT) (grant no. NRF-2021R1C1C1013706), and research fund by Seoul National University Bundang Hospital (grant no. 14-2018-0013).

Disclosure

All authors have declared no conflicts of interest.