Improved DeepMitosisNet framework for detection of mitosis in histopathology images

• Published in: Multimedia tools and applications 2024-04, Vol.83 (14), p.43303-43324
• Main Authors: B, Lakshmanan, S, Anand, P.S, Vivek Raja, B, Selvakumar
• Format: Article
• Language: English
• Subjects: Cell division; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Deep learning; Histopathology; Lymphocytes; Machine learning; Medical imaging; Mitosis; Multimedia Information Systems; Soft tissues; Special Purpose and Application-Based Systems; Track 2: Medical Applications of Multimedia
• ISSN: 1573-7721; 1380-7501; 1573-7721
• DOI: 10.1007/s11042-023-16830-8

Mitosis detection in the histopathology images is one of the important prognostic factors that is helpful for cancer grading. Mitosis counting from whole slide images is challenging due to shape variations and stain variations. The key to solving the issue is to find more discriminative features corresponding to object boundaries and other similar objects like lymphocytes, cells with dense nuclei, and various soft tissues with the same grey level. However, small mitotic cells are difficult to detect using traditional machine learning methods. In this paper, we present an improved deep learning combined teaching learning-based optimizer (TLBO) model called DeepMitosisNet for accurate detection of mitotic figures. The proposed model helps reduce misclassification rates and computational costs. Comprehensive experiments are performed on the MITOS-ATYPIA 14 dataset for both Aperio and Hamamatsu scanners. The proposed model achieved an F-score of 96%, precision of 93.7% and recall of 98%, showing that the proposed deep framework with TLBO shows substantial improvements over other state-of-the-art mitosis detection techniques.