Image Segmentation Using Neural Networks for Tumour Detection in Medical Imaging

Magnetic Resonance Imaging (MRI) is a vital tool for detecting and diagnosing brain tumors due to its ability to provide high-resolution, detailed images of brain structures. However, Prior models, have struggled to take full advantage of MRI for accurate tumor identification. The presented work is a qualitative study that investigates the way MRI images can be effectively used in Multimodal Brain Tumor Segmentation (BraTS) dataset to train a model for long term. Results: Detection algorithm, to automatically detect and extract Region of Interest (ROI) from MR images. This model outperformed prior methods used in the BraTS dataset and this case, described in more detail here is unique due to both training and testing being done on said two-dimensional test set from a single new case. On the BraTS 2020 validation dataset, it achieved a training accuracy of 97.68% and a test score equal to that obtained during training 96.21%. These results prove the efficacy of model to segment brain tumors in various modalities. This ability to adapt is also highlighted by comparative studies demonstrating applicability in different tumor shapes, sizes and locations. Again, the analysis proves that MRI data along with a well-geared model can provide accurate and efficient brain tumor detection.