Abstract:
A serious neurological illness known as a brain tumour is one in which the brain or
skull's cells proliferate out of control. As the mortality rates for this condition continue
to climb, manual examination of Magnetic Resonance Images (MRIs) is insufficient
for accurate diagnosis, and early discovery is essential for patient survival. Employing
ResNet-50 as a tool for early detection, this project focuses on improving the accuracy
of brain tumour diagnosis. The proposed ResNet-50 is trained on a combined dataset
from three sources: Figshare, SARTAJ, and Br35H, which contains 7023 Magnetic
Resonance Images (MRI) scans belonging to four categories: Glioma, Meningioma,
No-tumour, and Pituitary. The model is trained using several pre-processing strategies,
resulting in a proposed ResNet-50 based Computer-Aided Diagnosis (CAD) system
that achieved an accuracy of 98.70%. The model is deployed using Flask, connecting
its Application Programming Interface (API) to the front-end. The user-friendly
interface, designed with Tailwind CSS and Next.js, enables seamless interaction with
the system and utilizing efficient brain tumour detection capabilities. The techniques
developed in this project have the capability to assist clinicians specialising in the
timely identification of brain tumours.
Keywords: Brain tumour detection, Deep learning, Magnetic resonance imaging
(MRI), ResNet-50, Computer-aided diagnosis (CAD), Flask, Next.js
