Optimal Deep Transfer Learning-Based Human-Centric Biomedical Diagnosis for Acute Lymphoblastic Leukemia Detection

Area of Science:

• Biomedical Engineering
• Medical Imaging Analysis
• Artificial Intelligence in Healthcare

Background:

• Human-centric biomedical diagnosis (HCBD) is crucial for clinical decision-making, particularly for diseases like leukemia.
• Deep learning (DL) models show promise in medical image analysis, but require large datasets for training.
• Transfer learning offers a solution for feature extraction in DL models when data is limited.

Purpose of the Study:

• To develop an optimal deep transfer learning-based model for acute lymphoblastic leukemia detection (ODLHBD-ALLD).
• To accurately detect and classify acute lymphoblastic leukemia using blood smear images.
• To improve the efficiency and accuracy of computer-aided diagnosis in healthcare.

Main Methods:

• Utilized Gabor filtering (GF) for noise reduction in blood smear images.
• Employed a modified fuzzy c-means (MFCM) algorithm for image segmentation.
• Integrated Competitive Swarm Optimization (CSO) with EfficientNetB0 for feature extraction.
• Applied an attention-based long-short term memory (ABiLSTM) model for classification.

Main Results:

• The ODLHBD-ALLD model demonstrated superior performance in detecting and classifying acute lymphoblastic leukemia.
• Simulations on an open-access dataset confirmed the model's effectiveness.
• Comparative analysis showed significant improvements over existing diagnostic approaches.

Conclusions:

• The proposed ODLHBD-ALLD model offers an effective deep transfer learning solution for acute lymphoblastic leukemia detection.
• The integration of GF, MFCM, CSO-EfficientNetB0, and ABiLSTM enhances diagnostic accuracy.
• This approach contributes to advancing computer-aided detection in biomedical diagnosis.