Year: 2024
Type: article
Abstract: As recent advancements in image processing and computer vision have transformed the health technology scene of today, improvements in precision diagnostics come with colossal cost and time efficiency. MRI is a necessary tool for the expert team of radiologists, who would be able to detect minor abnormalities in the brain. The present work analyses the YOLOv10 framework of state-of-the-art deep architectures capable of classifying and detecting multiple brain tumors in MRI scans. In this work, we have included three different classes in the dataset: meningiomas, gliomas, and pituitary tumours. For exact tumour segmentation, we are applying advanced preprocessing technique with effective mask alignment. We evaluate YOLOv10 performance against the pre-processed dataset, which generates an optimistic result in tumour detection and classification. Preliminary results indicate a recall score of 0.912 for bounding box detection and 0.914 for mask segmentation with precision scores of 0.941 and 0.938, respectively. These results are therefore indicative of the promise of YOLOv10 in advancing the diagnosis of brain tumors toward better patient care in a clinical context. Keywords— Brain tumour, deep learning, image processing, MRI, YOLO.
DOI: 10.55041/ijsrem40304
Authors: Dr. G. Mary Swarnalatha, Gangam Nihithani, Shiva Prasad Gadila, Vyshnavi Keerthi
Language: English
Subfield: Neurology
Field: Neuroscience
Domain: Life Sciences
Open Access status: bronze
OpenAlex ID: W4405917461