OPTIMIZING MRI-BASED MEDICAL DIAGNOSIS: COMPARATIVE ANALYSIS OF EFFICIENTNET PERFORMANCE WITH VARYING LEARNING RATES | ||||
| Journal of the Egyptian Society of Tribology | ||||
| Volume 21, Issue 2, April 2024, Page 105-119 PDF (925.55 K) | ||||
| Document Type: Original Article | ||||
| DOI: 10.21608/jest.2024.278570.1085 | ||||
 View on SCiNiTO | ||||
| Authors | ||||
Wael Abouelwafa Ahmed1; Mohamed Ahmed Massoud  1; Mohamed E. El-Bouridy2
| ||||
| 1Department of Biomedical Engineering, Faculty of Engineering, Minia University, Minia, 61519, EGYPT. | ||||
| 2Alexandria Higher Institute of Engineering &Technology, Alex, Egypt. | ||||
| Abstract | ||||
| Magnetic Resonance Imaging (MRI) has emerged as a fundamental tool in the field of medical diagnostics, offering detailed insights into anatomical structures. As the demand for efficient and accurate diagnosis increases, leveraging deep learning techniques becomes imperative, among which the learning rate stands out as a pivotal factor influencing model convergence and generalization. In this research, we investigate the influence of varying learning rates on the efficacy of the EfficientNet B0 model, a cutting-edge convolutional neural network design acclaimed for its efficiency and proficiency in tasks related to image classification. Our comparative analysis unveils the profound influence of learning rates on the diagnostic accuracy and efficiency of the model. Specifically, we observe that optimal learning rates significantly enhance the convergence speed and overall performance of EfficientNet in medical image. In conclusion, this research highlighting the importance of learning rates in improving diagnostic precision and efficacy. We observed a wide range of outcomes in terms of training and validation accuracy, as well as training and validation losses. Notably, Trial 1 and Trial 2, which utilized lower initial learning rates (0.001 and 0.01, respectively), achieved higher validation accuracy compared to Trial 3, where the initial learning rate was set to 0.1. This suggests that tuning learning rates may lead to better convergence and generalization in the training process | ||||
| Keywords | ||||
| MRI; EfficientNet; classification models; learning rates; medical diagnosis | ||||
|
Statistics Article View: 214 PDF Download: 320 |
||||
