A Python-based comparative study of convolutional neural network–based approaches for the early detection of breast cancer
Ishrat Jahan Mohima, Mosabber Uddin Ahmed · 2024
Breast cancer is a serious health problem for women worldwide, and timely identification is crucial for effective treatment. Convolutional neural networks (CNNs) have looked promising in the past few years for the timely identification of breast cancer. Here, a distinctive analysis of various CNN-based approaches for prior breast cancer detection is provided using histopathological image data. The proposed model comprises several layers, including CNN layers, which extract relevant features from the provided input data. Max pooling is included to downsample extracted features and for reduction of the dimensionality of the input data. The flattened layer reshapes the pooled attributes into a one-dimensional (1D) vector. Then a long short-term memory (LSTM) layer is included, and the output generated by the LSTM layer is integrated to some dense layers, which apply nonlinear transformations to the features. Finally, a classification layer is used to classify the input image. How well the model performs has been assessed using histopathological image dataset and achieved 88.13% accuracy, 88.15% precision, 88.14% recall, 88.14% F1 score, and 94.12% area under the curve (AUC) score. The results indicate how this suggested methodology may effectively identify breast cancer in its earliest stages.