MultiFeNet: Multi‐scale feature scaling in deep neural network for the brain tumour classification in MRI images

• Published in: International journal of imaging systems and technology 2024-01, Vol.34 (1), p.n/a
• Main Authors: Agrawal, Tarun, Choudhary, Prakash, Shankar, Achyut, Singh, Prabhishek, Diwakar, Manoj
• Format: Article
• Language: English
• Subjects: Ablation; Artificial neural networks; Brain; Brain cancer; brain tumour detection; Central nervous system; deep learning; Diagnosis; Feature extraction; hybrid pooling; Image classification; Image enhancement; Machine learning; Magnetic resonance imaging; Medical imaging; MRI images; multi‐scale feature; Neural networks; Performance evaluation; Scaling; Tumors
• ISSN: 0899-9457; 1098-1098
• DOI: 10.1002/ima.22956

One of the most fatal and prevalent diseases of the central nervous system is a brain tumour. Different subgrades exist for each type of brain tumour because of the broad variety of brain tumours and tumour pathologies. Manual diagnosis may be error‐prone and time‐consuming, both of which are becoming more challenging as the medical community's workload grows. There is a need for automatic diagnosis. In this study, we have proposed a deep learning model (MultiFeNet) based on a convolutional neural network for the classification of brain tumours. MultiFeNet uses multi‐scale feature scaling for feature extraction in magnetic resonance imaging (MRI) images. Multi‐scaling helps to learn the better feature representation of the MRI image for enhanced model performance. To evaluate the proposed model, 3064 MRI scans of three distinct categories of brain tumours (meningiomas, gliomas and pituitary tumours) were used. The MultiFeNet obtained 96.4% sensitivity, 96.4% F1‐score, 96.4% precision and 96.4% accuracy on the benchmark Figshare dataset. In addition, an ablation study is conducted with the objective of evaluating the role of multi‐scaling in model performance.