Tropical Convolutional Neural Networks (TCNNs) Based Methods for Breast Cancer Diagnosis

Abstract

One of the leading causes of mortality for women worldwide is breast cancer. The likelihood of breast cancer-related mortality can be decreased by early identification and rapid treatment. Machine learning-based predictive technologies provide ways to detect breast cancer earlier. Several analytical techniques, such as breast MRI, X-ray, thermography, mammography, ultrasound, etc., may be used to find it. Accuracy metrics are the most extensively used approach for performance evaluation, and the Tropical Convolutional Neural Networks (TCNNs) model for breast cancer detection is the most precise and popular model. The proposed approach was examined using the Kaggle Breast Cancer Datasets (KBCD). The data set is partitioned into training and testing. We suggest a new class of CNNs called Tropical Convolutional Neural Networks (TCNNs), which are based on tropical convolutions and replace the multiplications and additions in traditional convolutional layers with additions and min/max operations, respectively, in order to reduce the number of multiplications. The results of the review demonstrated that the Tropical Convolutional Neural Networks (TCNNs) is the most successful and popular model for detecting breast cancer, and that accuracy metrics is the most popular approach for evaluating performance. It is amazing how deep learning is being used to so many different real-world problems. Additionally, because tropical convolution operators are basically nonlinear operators, we anticipate that TCNNs will be better at nonlinear fitting than traditional CNNs. The Kaggle Breast Cancer Datasets (KBCD) findings demonstrate that TCNN can reach more expressive power than regular convolutional layers.