Article: Deep learning-based mitosis detection using genetic optimal feature set selection Journal: International Journal of Bio-Inspired Computation (IJBIC) 2022 Vol.19 No.3 pp.189 - 198 Abstract: Mitosis detection in breast cancer is considered to be a vital factor in cancer progression. The significance of identifying mitotic count will be more helpful to estimate the aggressiveness and proliferation rate of the tumour. The manual mitosis detection process is prone to intra-observer variability and also a challenging task. To alleviate this limitation, we present a deep convolution neural network-based genetic optimiser to detect mitosis signature from histopathology images. In this study, the proposed model is designed to solve the problems of feature dimensionality, computational cost and misclassification rate. The deep learning-based genetic optimiser consists of two phases: first, deep convolutional neural network and second is genetic optimiser. It is compared to state-of-the-art algorithms using MITOS-ATYPIA-14 dataset. The proposed architecture achieved an accuracy of 98.7% with 91% precision, 89% recall and 92% F-score. Results are obtained from experiments conducted on 760 histopathology breast cancer images in which 415 images are used for training and 345 images are taken for testing. Significantly, the proposed model will intelligently assist and help pathologists to do their jobs more efficiently. Finally, the model could help pathologists, medical practitioner to understand the progression of the cancer stages. Inderscience Publishers - linking academia, business and industry through research

Title: Deep learning-based mitosis detection using genetic optimal feature set selection

Authors: B. Lakshmanan; S. Anand

Addresses: Department of Computer Science and Engineering, Mepco Schlenk Engineering College, Sivakasi, Tamil Nadu, India ' Department of Electronics and Communication Engineering, Mepco Schlenk Engineering College, Sivakasi, Tamil Nadu, India

Abstract: Mitosis detection in breast cancer is considered to be a vital factor in cancer progression. The significance of identifying mitotic count will be more helpful to estimate the aggressiveness and proliferation rate of the tumour. The manual mitosis detection process is prone to intra-observer variability and also a challenging task. To alleviate this limitation, we present a deep convolution neural network-based genetic optimiser to detect mitosis signature from histopathology images. In this study, the proposed model is designed to solve the problems of feature dimensionality, computational cost and misclassification rate. The deep learning-based genetic optimiser consists of two phases: first, deep convolutional neural network and second is genetic optimiser. It is compared to state-of-the-art algorithms using MITOS-ATYPIA-14 dataset. The proposed architecture achieved an accuracy of 98.7% with 91% precision, 89% recall and 92% F-score. Results are obtained from experiments conducted on 760 histopathology breast cancer images in which 415 images are used for training and 345 images are taken for testing. Significantly, the proposed model will intelligently assist and help pathologists to do their jobs more efficiently. Finally, the model could help pathologists, medical practitioner to understand the progression of the cancer stages.

Keywords: breast cancer; histopathology images; convolutional neural network; CNN; genetic optimiser; mitosis detection.

DOI: 10.1504/IJBIC.2022.123115

International Journal of Bio-Inspired Computation, 2022 Vol.19 No.3, pp.189 - 198

Accepted: 25 Jun 2021
Published online: 30 May 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article

Title: Deep learning-based mitosis detection using genetic optimal feature set selection

Keep up-to-date