Analysis of brain MRI images through spiking neural network technique

Brain tumours are pathological masses that develop abnormally in the brain, with the potential to cause significant damage and even life-threatening complications if left untreated. The early diagnosis of brain tumours is a critical factor in the success of the treatment process. In this context, deep learning-based models in biomedical imaging technologies are increasingly employed for the diagnosis and treatment of diseases. The present study proposes a novel approach that integrates Spiking Neural Networks (SNNs), inspired by biological neural networks, with the YOLOv7 algorithm for object detection. This integration has been undertaken to develop a system that is both energy efficient and highly accurate in detecting and labelling tumours in brain MRI images. The Br35H: Brain Tumor Detection 2021 dataset was analysed in this study, with a total of 3000 MRI images being evaluated. The tumour regions in the dataset were labelled by expert radiologists and the images were subjected to pre-processing techniques such as scaling before being applied as input to the model. The SNN model was trained using the Nengo framework, and the outputs were given as input to the YOLOv7 algorithm for object detection. The performance of the proposed system was evaluated using metrics such as Accuracy, Precision, Recall, F1 score and mean accuracy (mAP). The results of the tests demonstrate that the SNN model exhibits a 97% accuracy rate, while the YOLOv7 algorithm achieves 86% precision, 96% recall, 76% F1 score, and 81% mAP, respectively. The findings reveal that the proposed system offers a clinically applicable solution for tumour detection in brain MRI images thanks to its real-time object labelling structure. The study demonstrates that the integration of deep learning-based models with biomedical imaging technologies can be regarded as an innovative method in medical diagnosis and treatment processes.