A Swift and Accurate Approach to Breast Cancer Recognition Using CNN and Decision Tree Method

Radiological examination using MRI is critical in the early diagnosis and in planning the therapy in handling cancer. Complexity and heterogeneity of tumor presentation in MRI scans usually necessitates the use of extreme machine learning strategies because the problem is beyond the capacity of conventional methodologies. It is in this work that the author puts forward a double strategy that aims at enhancing the ability of MRI image cancer diagnosis by combining CNNs and DTs. To ensure the best outcome of the classification, features the CNN recognizes patterns in the MRI scans are divided into the DT classifier. The CNN is applied for feature extraction. This method enhances the detection accuracy by a large extent by combine both the efficiencies and strengths of CNNs and DTs. For classification purpose, the dataset containing 10 thousand MRI scans split equally between the scans having cancer and those not, was collected from TCIA for the validation of the model. CNN+DT proved to be better than conventional CNN only approach with the test accuracy of 92% after the extensive trial. In identifying the cancerous regions of the mammogram, the model was competent and equally airtight in avoiding false identification stemming from false positives where records of 80% were confirmed on precision, recall and, F1-score measures. Thus, these results suggest that using CNNs and DTs might be a way to improve the cancer diagnosis in MRI images. As for the advantage of this hybrid technique, it provides a better accuracy than the radial basis function model while giving better interpretability of a model, which, to the clinician, may be very important.