Related Experiment Video
Updated: Sep 8, 2025
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
6.9K
Lightweight convolutional neural networks using nonlinear Lévy chaotic moth flame optimisation for brain tumour
Amin Abdollahi Dehkordi1, Mehdi Neshat2,3, Alireza Khosravian4
1Department of Computer Engineering, Najafabad Branch, Islamic Azad University, Najafabad, Iran. amin.abdollahi.dehkordi@gmail.com.
Scientific Reports
|July 2, 2025
Summary
This study introduces a fast, efficient deep learning model for medical image classification. The novel approach combines lightweight convolutional neural networks (CNNs) with advanced hyperparameter optimization, achieving superior accuracy in brain tumor detection.
Area of Science:
- Artificial Intelligence
- Medical Imaging
- Machine Learning
Background:
- Deep convolutional neural networks (CNNs) excel in medical image classification but demand significant computational resources and complex hyperparameter optimization.
- Optimizing deep CNN hyperparameters is computationally intensive, posing challenges for researchers without high-performance computing access.
Purpose of the Study:
- To develop a computationally efficient and highly accurate model for medical image classification.
- To address the limitations of traditional deep CNNs by introducing a lightweight CNN combined with an advanced optimizer for hyperparameter tuning.
Main Methods:
- Development of lightweight CNNs integrated with the Nonlinear Lévy chaotic moth flame optimiser (NLCMFO) for automatic hyperparameter optimization.
- NLCMFO enhances exploration and exploitation phases using Lévy flight, chaotic parameters, and nonlinear control mechanisms.
- Empirical analysis using a dataset of 2314 brain tumor detection images.
Main Results:
- The proposed CNN_NLCMFO model achieved 92.40% accuracy, outperforming a non-optimized CNN by 5%.
- It surpassed established models like DarkNet19 (96.41%), EfficientNetB0 (96.32%), Xception (96.41%), ResNet101 (92.15%), and InceptionResNetV2 (95.63%).
- Performance gains ranged from 1% to 5.25% compared to existing models.
Conclusions:
- The lightweight CNN combined with NLCMFO offers a computationally efficient and accurate solution for medical image classification.
- This approach effectively addresses the challenges posed by traditional deep CNNs in terms of computational cost and hyperparameter optimization.
- The findings highlight the potential of optimized lightweight CNNs for practical applications in medical diagnostics.
