Utilizing Convolutional Neural Networks for Breast Cancer Classification

Cancer denotes the anomalous proliferation of cells within bodily tissues, potentially exhibiting rapid dissemination throughout diverse anatomical regions. Prevalent manifestations encompass lung, colorectal, prostate, and notably, breast cancer. In the context of breast cancer, unrestrained growth transpires within adjacent breast tissue, frequently culminating in dire outcomes due to delayed identification and intervention. The diagnostic landscape of breast cancer contends with challenges in precise ailment discernment and categorization, largely attributed to the exponential surge in image volumes occasioned by escalating case numbers. The intricate likeness between certain tissue characteristics, coupled with the formation of clusters spanning 0.05mm to 1mm, exacerbates localization difficulties, thereby compromising classification accuracy. Although strides have been taken via machine learning methodologies such as support vector machines and decision trees, effectiveness on raw image data remains circumscribed due to the prerequisite of preliminary feature extraction. Consequently, an imperative exists for an automated system proficient in feature acquisition and precise prognostication. This research centers on modernizing medical image analysis through a customized adaptation of Convolutional Neural Networks (CNNs), facilitating erudition, detection, and classification of malignant and benign breast tissue with heightened precision. Leveraging transfer learning obviates model construction de novo, a strategy evincing efficacy in breast cancer categorization. This approach capitalizes on the AlexNet architecture, tailored to our specific task and augmented by reflection and rotation to amplify dataset diversity. Augmentation operations rectify image deficiencies prior to model input. Experimental evaluations on the test dataset manifest accuracy rates of 95.80%, 95.00%, 80.00%, 92.30%, and 93.63% for accuracy, sensitivity, specificity, precision, and F1 score respectively. Outcomes evince substantial enhancements in classification accuracy relative to prior deep learning approaches and the MIAS breast cancer dataset. This innovation promises to empower medical practitioners with precise classification, mitigating the temporal inefficiencies inherent in manual breast cancer scrutiny