A novel framework for Breast cancer scoring based on machine learning techniques using Immunohistochemistry images

Razzaq, Hasanain Hayder; Ghazali, Rozaida; George, Loay E.; Zulqarnain, Muhammad

doi:10.22034/cmde.2024.63113.2810

A novel framework for Breast cancer scoring based on machine learning techniques using Immunohistochemistry images

Document Type : Research Paper

Authors

¹ 1. Faculty of Computer Science and Information Technology, University Tun Hussein Onn Malaysia, Batu Pahat, 86400, Johor, Malaysia. 2. College of Medicine, Jabir ibn Hayyan Medical University, Najaf, Iraq.

² Faculty of Computer Science and Information Technology, University Tun Hussein Onn Malaysia, Batu Pahat, 86400, Johor, Malaysia.

³ University of Information Technology and Communication (UoITC), 10001, Baghdad, Iraq.

⁴ Department of Computer Science & IT, Cholistan University of Veterinary and Animal Sciences, Bahawalpur, Punjab, Pakistan.

10.22034/cmde.2024.63113.2810

Abstract

Breast cancer is classified as a serious disease in the medical field, and there is no doubt that breast cancer detection requires effective and accurate techniques. Integrating deep learning (DL) and machine learning (ML) methods has shown promising results in this area. In this research, we introduce a hybrid approach for breast cancer diagnosis, which is centered on the analysis of immunohistochemical images. The proposed method encompasses algorithms for image pre-processing, segmentation, extracting informative indicators (such as relative cell area and intensity), and an algorithm for categorizing the molecular harmonic subtype of breast cancer. The number of the sample was 598, divided into training 70% and testing 30%. The 5-fold cross-validation was used to assess the proposed approach. Experimental results showcased the effectiveness of the proposed hybrid method in achieving superior performance in the detection of breast cancer, especially within breast cancer scoring systems. The accuracy of our proposed approach, which involved combining HSV integration with adaptive high boost filtering, reaches a peak at 96.5% when using SVC (linear kernel). Moreover, the precision, recall, F1-score, and specificity metrics are recorded at 95.29%, 99.99%, 95.59%, and 99.28%, respectively. Additionally, this study evaluated the efficacy of the proposed model in comparison to various other traditional breast cancer detection approaches.

Keywords