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

• DOI: 10.1016/j.heliyon.2024.e40452
• Publication Date: 2024-11-15T02:28:38Z
• Authors: Myeong Suk Yim; Yun Heung Kim; Hyeon Sang Bark; Seung Jae Oh; Inhee Maeng; Jin-Kyoung Shim; Jong Hee Chang; Seok-Gu Kang; Byeong Cheol Yoo; Jae Gwang Kwon; Jungsup Byun; Woon-Ha Yeo; Seung-Hwan Jung; Han-Cheol Ryu; Se Hoon Kim; Hyun Ju Choi; Young Bin Ji
• Keywords: Digital Pathology

We used deep learning methods to develop an AI model capable of autonomously delineating cancerous regions in digital pathology images (H&E-stained images). By using a transgenic brain tumor derived from the TS13-64 cell line, we digitized total 187 H&E-stained and annotated these compile dataset. A approach was executed through DEEP:PHI, which abstracts Python coding complexities, thereby simplifying execution training protocols for users. employing Image Crop with Mask technique patch generation method, not only maintained appropriate data class balance but also overcame challenge limited computing resources. This enabled us successfully that segments areas. enables provision guiding determining areas minimal assistance neuropathologists. In addition, high-quality, large dataset curated proposed contributes development novel terahertz imaging-based cancer diagnosis technologies accelerates technological advancements.