Brain Tumor Classification for MRI-CT Fused Images using NAS-Optimized Bio-Inspired PSO Algorithm

Healthcare sector plays a vital role in industry contributing for the tremendous success of deep learning algorithms in the field of image processing in recent years. Medical image processing requires high accuracy in data prediction and classification that, brings up huge benefits to the healthcare...

Full description

Saved in:
Bibliographic Details
Published in:2025 9th International Conference on Inventive Systems and Control (ICISC) pp. 885 - 891
Main Authors: B, Ajith Jerom, S, Sarathambekai
Format: Conference Proceeding
Language:English
Published: IEEE 12.08.2025
Subjects:
Accuracy
Biological system modeling
Biomedical imaging
Brain tumors
Classification algorithms
Deep learning
EfficientB0
Image Fusion
Improved Chimp Optimization Algorithm (ICHOA)
Neural architecture search
Neural Architecture Search (NAS)
Optimization
Particle swarm optimization
Particle Swarm Optimization (PSO)
Wavelet Transform Fusion
Wavelet transforms
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Healthcare sector plays a vital role in industry contributing for the tremendous success of deep learning algorithms in the field of image processing in recent years. Medical image processing requires high accuracy in data prediction and classification that, brings up huge benefits to the healthcare. It has been largely affected by some rapid progress in particular image detection, image recognition, image segmentation, and computer aided diagnosis. Medical imaging is an essential for providing diagnostic information through different modalities. In this paper, MRI-CT brain tumor fused image is classified using NAS-driven Bio-Inspired PSO algorithm. Brain tumor classification necessitates highresolution imaging to guarantee correct localization and diagnosis. Scans of single images contains lack of ability to visualize tumor boundaries and areas of significance. Hence, to overcome this problem there is a need for an automated system for tumor classification without losing its essential features. Wavelet transform fusion is used to fuse the MRI and CT images of brain which, decomposes them into different frequency components and then selectively merge those data to form a single enhanced fused image. The EfficientNetB0, is a lightweight and accurate CNN, used for feature extraction due to efficient compound scaling and strong representation. The ICHOA algorithm is inspired by chimpanzees social behaviour designed for feature selection. Also, the proposed NAS is an automated method to identify optimal neural network architectures without manual intervention. By, exploring a defined search space, NAS enables, discovery of high-performing and accurate classification of tumors. Bio-Inspired PSO optimization reduces trail-and-error and computational burden.
DOI:10.1109/ICISC65841.2025.11188293