A novel deep learning based approach with hyperparameter selection using grey wolf optimization for leukemia...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Traditional leukemia diagnosis relies on manual microscopic analysis of blood smears, which is subjective and error-prone.
• Automated methods are needed to improve the accuracy and efficiency of leukemia diagnosis.

Purpose of the Study:

• To develop and evaluate a deep learning framework for automated leukemia classification from microscopic images.
• To enhance image quality and leverage advanced AI models for improved diagnostic performance.

Main Methods:

• A novel lightweight algorithm using hyperbolic sine function for contrast enhancement.
• A customized Convolutional Neural Network (CNN) model incorporating a parallel inverted dual self-attention network (PIDSAN4) and a tiny Vision Transformer (ViT).
• Hyperparameter tuning using grey wolf optimization for model training.

Main Results:

• The proposed model achieved high performance metrics: 0.913 accuracy, 0.892 sensitivity, 0.925 specificity, 0.883 precision, 0.894 F-measure, and 0.901 G-mean.
• Comparison with state-of-the-art pre-trained models demonstrated superior accuracy of the proposed framework.
• The automated system offers a more objective and potentially faster diagnostic approach.

Conclusions:

• The developed deep learning framework shows significant potential for accurate and automated leukemia diagnosis.
• This approach can overcome the limitations of traditional manual methods, leading to better patient outcomes.
• Further research can explore integration into clinical workflows for real-world application.