Probabilistic GA-NARX based Approach for Brain Tumor Classification Via Tumor Region Partition and Augmentation

The suggested method offers a sophisticated classification technique for distinguishing between normal and abnormal brain MRI data. Diagnostic imaging techniques including EKG, MRI, and CT scans are becoming increasingly important in modern medicine. Manual assessment of tumor based on visual inspection by radiologist or physician is the standard approach, although this might lead to inaccurate classification when a significant number of MRIs must be assessed. Brain tumors in humans are a major killer. The patient's prognosis improves with rapid and precise tumor diagnosis. Before training a model, data must be prepared, cleaned, segmented, and features extracted. Preprocessing procedures, such as image scaling and Histogram equalization, can yield sharper and more defined final results. K-means clustering is the most convenient and straightforward method for segmentation. The (GLCM) is a pattern matrix used in feature extraction to recognize the texture pattern present in a photograph by characterizing texture as a two-dimensional array of gray levels. Let's start using GA-NARX to train our models. In comparison to LSTM and GA, the two current leaders, the proposed technique has been proved to perform better.