Advanced PTSVM Based Breast Cancer Classification with Weighted Feature Selection

In real time CAD system, it is important to consider different type of features of the image to diagnose the patients precisely. The authors have proposed an enhanced breast cancer classification model where all the textural, morphological, and graph features have been amalgamated and selected as per proposed weighted feature selection. Further, the model has been trained based on semi-supervised learning assuming that the labeled/annotated data is very less initially as the labeled image datasets for breast cancer is uncommon. In this paper, the authors propose a novel classification model that integrates diverse features to enhance the performance of CAD systems. The model is specifically designed to address the challenges faced by the scarcity of annotated data, a common issue in medical imaging, by optimizing the use of available resources and reducing the need for extensive expert annotations. Additionally, the model effectively minimizes inter- and intra-observer variability, offering reliable support to experts in early diagnosis. The proposed method demonstrates a high accuracy of 97.74%, with precision and recall rates of 96.14% and 99.80%, respectively, indicating its potential as a robust tool in clinical settings.