Hybrid neutrosophic enhanced MobileNetV2 model for leukemia blood cell classification

Area of Science:

• Medical Imaging Analysis
• Computational Pathology
• Artificial Intelligence in Medicine

Background:

• Leukemia diagnosis relies on manual inspection of blood cell images, which is time-consuming and prone to errors.
• Automated methods are needed to improve the accuracy and efficiency of leukemia classification.
• Challenges in image analysis include uncertainty, ambiguity, and poor contrast in blood cell representations.

Purpose of the Study:

• To develop a hybrid model for accurate leukemia classification using transfer learning and neutrosophic domain enhancement.
• To address image uncertainties and improve features for better leukemia identification.
• To enhance the performance of automated blood cell image analysis for medical diagnosis.

Main Methods:

• A hybrid model integrating transfer learning (MobileNetV2) and neutrosophic domain enhancement was proposed.
• Neutrosophic domain transformation separated RGB images into Truth, Falsity, and Indeterminacy components.
• Image augmentation techniques including wavelet sharpening, CLAHE, TVM, and denoising were applied to specific components.

Main Results:

• The model was trained and tested on 3,256 peripheral blood smear images across 4 classes.
• The neutrosophic-enhanced MobileNetV2 model achieved 98.36% overall testing accuracy.
• A macro F1-score of 0.98 was obtained, demonstrating significant multi-class leukemia classification enhancement.

Conclusions:

• Neutrosophic enhancement significantly improves classifier performance in leukemia detection.
• The proposed method achieves higher accuracy without increasing computational power.
• This approach offers a promising tool for more accurate and efficient leukemia diagnosis.