ALNett: A cluster layer deep convolutional neural network for acute lymphoblastic leukemia classification

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Acute Lymphoblastic Leukemia (ALL) is a significant pediatric cancer, with over 6500 US cases annually.
• Early ALL diagnosis is crucial for improving clinical decisions and patient outcomes.
• Advancements in AI and big data analytics offer new avenues for rapid and accurate cancer detection.

Purpose of the Study:

• To develop and evaluate a deep neural network model, ALNett, for the classification of microscopic white blood cell images for ALL detection.
• To leverage depth-wise convolution with varying dilation rates for enhanced feature extraction in ALL diagnosis.
• To compare ALNett's performance against established pre-trained models.

Main Methods:

• A novel deep neural network, ALNett, was designed using depth-wise convolution and dilation rates.
• The model incorporates convolution, max-pooling, and normalization layers for robust feature extraction.
• ALNett's performance was benchmarked against VGG16, ResNet-50, GoogleNet, and AlexNet using key metrics.

Main Results:

• ALNett achieved the highest classification accuracy at 91.13%.
• The model demonstrated a superior F1 score of 0.96.
• ALNett exhibited lower computational complexity compared to the pre-trained models evaluated.

Conclusions:

• The proposed ALNett model shows significant promise for accurate Acute Lymphoblastic Leukemia categorization.
• ALNett outperforms existing pre-trained models in ALL classification from microscopic images.
• The model's efficiency and accuracy suggest its potential utility in clinical settings for aiding early ALL diagnosis.