A Robust Malaria Cell Detection Framework Using Adaptive and Atrous Convolution-Based Recurrent Mobilenetv2 with...

Area of Science:

• Medical Imaging
• Parasitology
• Artificial Intelligence

Background:

• Malaria, a deadly mosquito-borne disease, requires early detection for effective treatment.
• Current diagnostic methods like Rapid Diagnostic Tests (RDTs) and microscopy have limitations in accuracy, cost, and accessibility, especially in endemic regions.
• Existing deep learning models for malaria detection often demand significant computational resources.

Purpose of the Study:

• To develop an advanced deep learning-based adaptive method for accurate malaria cell detection in medical images.
• To overcome the limitations of traditional and current deep learning approaches in terms of accuracy, processing power, and cost.
• To improve early diagnosis capabilities for malaria.

Main Methods:

• A novel adaptive deep learning framework was designed, incorporating image segmentation and cell recognition.
• Abnormality segmentation was performed using a developed Trans-MobileUNet++ (T-MUnet++) network, leveraging its global context capture for segmentation tasks.
• Malaria cell recognition was achieved using an Adaptive and Atrous Convolution-based Recurrent MobilenetV2 (AA-CRMV2) model.
• The AA-CRMV2 model's parameters were optimized using the Updated Random Parameter-based Fennec Fox Optimization (URP-FFO) algorithm.

Main Results:

• The developed Trans-MobileUNet++ effectively segmented abnormalities in medical images.
• The AA-CRMV2 model, optimized by URP-FFO, demonstrated high efficacy in recognizing malaria cells.
• Experimental analyses showed the proposed adaptive deep learning approach outperformed classical techniques in malaria detection.

Conclusions:

• The proposed deep learning-based adaptive method offers a promising, accurate, and efficient solution for malaria detection from medical images.
• This approach has the potential to enhance early diagnosis, particularly in resource-limited settings.
• The integration of advanced segmentation and optimized recognition models represents a significant advancement in automated malaria diagnosis.