Authors:
B. Gunapriya, V. Revathi, Thirumalraj Karthikeyan, S. Gopikha, Sheila Agnes Vidot

Addresses:
Department of Electrical and Electronics Engineering, New Horizon College of Engineering, Bengaluru, Karnataka, India. Department of Research and Development, New Horizon College of Engineering, Bengaluru, Karnataka, India. Department of Artificial Intelligence, Trichy Research Labs, Quest Technologies, Tiruchirappalli, Tamil Nadu, India. Department of Information Technology, St. Joseph’s College of Engineering, Chennai, Tamil Nadu, India. Department of Medical Consultant, IPS Health, Mahe, Victoria, Seychelles.

Abstract:

Deep learning and artificial neural networks can diagnose brain tumours as benign or malignant. Brain disease research requires image categorisation and automatic BT segmentation from reconstructed microwave (RMW) brain images. The tumour's architecture requires manual tumour identification, classification, and segmentation, which is highly time-consuming yet vital. RMB images were originally utilised to create a sensor-based microwave brain imaging (SMBI) image library. The 1320 photos include 300 "normal" photographs, 215 "malignant" images, 200 "double benign" and "double malignant" images, and 190 "single benign" and "single malignant" images. Images were resized and normalised. The dataset was expanded to 13,200 training photos per fold for 5-fold cross-validation. Though promising in BT classification, deep learning techniques cannot extract global features and preserve long-range associations. Vision Transformer (ViT) uses a self-attention mechanism to model long-range associations, which are essential for BT categorisation. For the BT organisation, the researcher employs a TECNN-based model with a CNN for extraction and an attention instrument in the transformer for global feature extraction. The Dung Beetle Optimisation Algorithm (DBO) chooses the best weight value here. The TECNN model achieved 89%-91% six-class categorisation accuracy after training on raw RMB images. The SMBI scheme's RMB photo classification can use the suggested model to diagnose tumours accurately.

Keywords: Brain Classification; Reconstructed Microwave; Vision Transformer; Double Malignant; Microwave Images; Dung Beetle Optimisation Algorithm; Global Pooling.

Received on: 22/11/2024, Revised on: 31/01/2025, Accepted on: 23/03/2025, Published on: 07/09/2025

DOI: 10.69888/FTSHSL.2025.000501

FMDB Transactions on Sustainable Health Science Letters, 2025 Vol. 3 No. 3, Pages: 148-164

• Views : 161
• Downloads : 22

Download PDF