Breast Cancer Examination in Digitized Mammograms using Integrated K-Means Clustering with Garbor Filter and Shrunk Kernel KNN Method
V. Sridevi · Indian Journal of Science and Technology · 2024
Objectives: To suggest an intelligent classification system for efficient breast cancer diagnosis that distinguishes between benign and malignant breast cancer. The goal of the research is to develop a unique CAD system for the detection & classification of breast cancer using novel K-Means clustering (KMC) with Gabor Filter (GF) and Shrunk Kernel K-Nearest Neighbor (KNN) classifier. Methods: Two different sorts of perspectives, such as Craniocaudal (CC) and Mediolateral oblique (MLO) mammograms are employed to improve diagnostic effectiveness. Utilizing an adaptive K-means clustering technique to segment the tumor. The Gabor filter is used in conjunction with the k-means clustering method to extract the features of the CC and MLO perspectives. The mammography image is finally classified into benign and malignant using a unique Shrunk Kernel K-Nearest Neighbor (SKKNN) classifier. The biopsy-proven annotated mammograms from the CBIS-DDSM dataset are used in this study. There were 6156 occurrences in the dataset with MLO and CC view of 1331 normal, 858 benign and 889 malignant mammograms. Findings: The experimental findings showed that the suggested model KMC-GF and SKKNN can accurately detect breast cancer at an early stage. The accuracy, sensitivity , specificity, AUC, precision, F1-measure for SKKNN was 92.56%, 93.8%, 92.75%, 95.2%, 93.93%, and 94.5% which are higher comparing single view features. Novelty: This technique could be employed in the medical field to diagnose breast cancer and also produce few false positive results. This method reduces the workload for radiologists while still providing reliable diagnostics without the need for expensive procedures or a lot of equipment. Keywords: SKKNN, Adaptive K-means segmentation, Gabor filter, MLO and CC Mammogram, KMC-GF