Breast Cancer Detection and Classification Using Artificial Intelligence

It is critical to find breast cancer quickly. This article introduces a novel approach to breast cancer categorization that makes use of deep learning together with other segmentation approaches. Here, we present a novel CAD system that can detect breast mammography pictures for benign and malignant mass tumours. Not one, but two separate segmentation methods are used by this CAD system. The threshold and area based strategy is used in one method, while manually determining the ROI is used in the other. The identifying traits are extracted using a deep convolutional neural network, or DCNN. Now, instead of one thousand, we may categorise two classes using AlexNet, a famous DCNN architecture. For added precision, the support vector machine (SVM) classifier is linked to a last fully connected (fc) layer. We utilise two open-source datasets-DDSM and CBIS-DDSM, which stands for the Curated Breast Imaging Subset of DDSM-to obtain the results. Accuracy can be enhanced through training on a bigger dataset. However, biological databases have a limit on the number of samples they can retain due to the high patient load. Thus, data augmentation is a method for increasing the amount of input data by generating new data from the existing data. Data augmentation comes in various forms; rotation is one of them. Manual ROI extraction from mammography results in an accuracy of 71.01 % for the newly trained DCNN architecture. The area under the curve (AUC) reached its maximum of 0.88, or 88%, for samples generated by both segmentation techniques. The DCNN's accuracy is enhanced to 73.6% when CBIS-DDSM samples are used, which is an additional perk. Remarkably, our SVM accuracy was 87.2% and our AUC was 0.94 (94% confidence range). Here, in comparison to earlier studies under these same settings, the AUC is at its maximum.