Deep learning-driven macroscopic AI segmentation model for brain tumor detection via digital pathology: Foundations for terahertz imaging-based AI diagnostics - Research article

Sep 21, 2026

We used deep learning methods to develop an AI model that autonomously delineates cancerous regions in digital pathology images. A transgenic brain tumor model derived from the TS13-64 brain tumor cell line supported digitization of 187 H&E-stained images with annotated cancer regions to build a training dataset. Training was executed through DEEP:PHI to simplify AI protocol execution, while the Image Crop with Mask and patch generation preserved class balance and addressed limited computing resources. The resulting model provides guidance images for minimal neuropathologist assistance and supports terahertz imaging-based AI diagnostic development.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages11

Reading Time28 min

AccessFree to download

Abstract

Introduction

Heliyon 10 (2024) e40452 Contents lists available at ScienceDirect Heliyon journal [homepage:](homepage: www.cell.com/heliyon)[ www.cell.com/heliyon](homepage: www.cell.com/heliyon) | Research article
Deep learning-driven macroscopic AI segmentation model for brain tumor detection via digital pathology: Foundations for terahertz imaging-based AI diagnostics
Myeong Suk Yimb,1, Yun Heung Kim c,1, Hyeon Sang Bark a,1, Seung Jae Oh d, Inhee Maengd, Jin-Kyoung Shime, Jong Hee Change, Seok-Gu Kange, Byeong Cheol Yooc, Jae Gwang Kwon a, Jungsup Byunb, Woon-Ha Yeof, Seung-Hwan Jung f, Han-Cheol Ryu f, Se Hoon Kim g,***, Hyun Ju Choi c,**, Young Bin Jia,b,*
a Advanced Photonics Research Institute (APRI), Gwangju Institute of Science and Technology (GIST), Gwangju, 61005, Republic of Korea b Gimhae Biomedical Center, Gimhae Biomedical Industry Promotion Agency (GBIA), Gimhae, 05969, Republic of Korea
c DX Business Division, Deepnoid.Inc, Seoul, 08376, Republic of Korea
d YUHS-KRIBB Medical Convergence Research Institute, Yonsei University College of Medicine, Seoul, 03722, Republic of Korea e Department of Neurosurgery, Brain Tumor Center, Severance Hospital, Yonsei University College of Medicine, Seoul, 03722, Republic of Korea f Department of Artificial Intelligence Convergence, Sahmyook University, Seoul, 01795, Republic of Korea
g Department of Pathology, Severance Hospital, Yonsei University College of Medicine, Seoul, 03722, Republic of Korea | | --- | | A B S T R A C T | | We used deep learning methods to develop an AI model capable of autonomously delineating cancerous regions in digital pathology images (H&Estained images). By using a transgenic brain tumor model derived from the TS13-64 brain tumor cell line, we digitized a total of 187 H&E-stained images and annotated the cancerous regions in these images to compile a dataset. A deep learning approach was executed through DEEP:PHI, which abstracts Python coding complexities, thereby simplifying the execution of AI training protocols for users. By employing the Image Crop with Mask technique and patch generation method, we not only maintained an appropriate data class balance but also overcame the challenge of limited computing resources. This approach enabled us to successfully develop an AI training model that autonomously segments cancerous areas. This AI model enables the provision of guiding images for determining cancerous areas with minimal assistance from neuropathologists. In addition, the high-quality, large dataset curated for training using the proposed approach contributes to the development of novel terahertz imaging-based AI cancer diagnosis technologies and accelerates technological advancements. | 1. Introduction Cancer remains one of the major public health issues globally, with its incidence and mortality rates persistently rising due to population growth, increasing average lifespan, and lifestyle changes [1]. Consequently, the importance of research in cancer * Corresponding author. Advanced Photonics Research Institute (APRI), Gwangju Institute of Science and Technology (GIST), Gwangju, 61005, Republic of Korea. ** Corresponding author. *** Corresponding author. E-mail addresses: [paxco@yuhs.ac](paxco@yuhs.ac) (S.H. Kim), [hjchoi@deepnoid.com](hjchoi@deepnoid.com) (H.J. Choi), [ybji.thz@gist.ac.kr](ybji.thz@gist.ac.kr) (Y.B. Ji). 1 These authors contributed equally to this work. [https://doi.org/10.1016/j.heliyon.2024.e40452](https://doi.org/10.1016/j.heliyon.2024.e40452) Received 9 September 2024; Received in revised form 5 November 2024; Accepted 14 November 2024 Available online 15 November 2024 2405-8440/© 2024 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license ([http://creativecommons.org/licenses/by/4.0/](http://creativecommons.org/licenses/by/4.0/)). prevention and early diagnosis and the development of effective treatments is attracting increasing attentio

FAQs

What AI capability does the model provide in digital pathology images?

The model autonomously delineates cancerous regions in digital pathology images, enabling segmentation of tumor areas.

How was the training dataset constructed?

The study digitized 187 H&E-stained images from a transgenic brain tumor model and annotated the cancerous regions to compile the dataset.

What techniques were used to improve training efficiency and data handling?

DEEP:PHI was used to abstract Python coding complexity, and Image Crop with Mask plus patch generation helped maintain class balance and mitigate limited computing resources.

Sep 21, 2026

Uploaded by Clementine