Optimized deep maxout for breast cancer detection: consideration of pre-treatment and in-treatment aspect

Breast cancer is one of the deadliest diseases, accounting for the second-highest rate of cancer mortality among females. Breast tissue begins to develop cancerous, malignant lumps as the disease progresses. Self-examinations and routine clinical checks aid in early diagnosis, which considerably increases the likelihood of survival. Because of this, we have created a revolutionary method for finding breast cancer that has the following four steps. Fuzzy filters are used in the initial pre-processing stage to reduce noise and improve outcomes from the incoming data. In the second stage, we have presented an Improved Hierarchical DBSCAN (Density-based clustering algorithm) for the segmentation of anomalous areas. Feature extraction will be carried out following segmentation. We have also developed a better kurtosis-based feature to complement traditional statistical and shape-based features and deliver better results. The Optimized Deep Maxout Neural Network is used for classification in the final step, with the suggested Shark Smell Indulged Shuffled Shepherd Optimization used to optimize the weight parameter (SSISSO). At 90% the learning percentage of the proposed model SSISSO model has achieved 0.984391 accuracy, which is superior to 22.54%, 28.46%, 17.44%, 17%, 15.04%, 13.28%, 29.45%, 28.59%, 21.58%, and 30.72% as compared to other methods like SVM-BS1, CNN-BS7, LSTM, NN, Bi-GRU, RNN, ARCHO, AOA, HGS, CMBO, SSOA, and SSO. Finally, the results of the proposed breast cancer detection technique are compared with conventional techniques.