Transfer Learning Based Breast cancer Classification using One-Hot Encoding Technique

Early diagnosis of breast cancer can be curable with precise techniques and improve patients prognosis with cancer. Most people failed to detect cancer early, leading to an increased death rate. In recent years, several studies have developed in various imaging modalities to predict breast cancer. The medical practitioner sometimes diagnosis diseases mistakenly due to misinterpretation. The automated assistance for practitioners using different computer-aided diagnosis (CAD) will give an accurate prediction for critical diseases. This paper presented a CAD system to perform automatic breast cancer diagnosis by employing the one-hot encoding technique. The system has experimented with BreakHis dataset partitioned into training and testing sets. The system performance can be measured with accuracy, sensitivity, specificity, precision, and recall. The system acquired 98.62% accuracy using the one-hot encoding technique, which is better than state-of-art methods. The proposed system outperformed the existing system by utilizing various performance metrics.