Medical Image Captioning Using Optimized Deep Learning Model.

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Natural Language Processing

Background:

• Medical image captioning generates textual descriptions for medical images.
• Current methods, like the show, attend, and tell model (ATM), face challenges with initial parameter sensitivity.
• Efficiently evaluating visual-textual similarity is crucial for accurate medical image captioning.

Purpose of the Study:

• To enhance the performance of the show, attend, and tell model (ATM) for medical image captioning.
• To address the sensitivity of the ATM to its initial parameters.
• To improve the generation of natural language descriptions for medical images.

Main Methods:

• Implementation of a novel show, attend, and tell model (ATM) incorporating a visual attention mechanism via an encoder-decoder architecture.
• Optimization of the ATM's initial parameters using the Strength Pareto Evolutionary Algorithm-II (SPEA-II).
• Experimental evaluation using benchmark datasets and comparison with existing medical image captioning techniques.

Main Results:

• The SPEA-II-optimized ATM demonstrated superior performance compared to conventional models.
• The proposed method achieved significant improvements in medical image captioning accuracy.
• The visual attention mechanism effectively guided the caption generation process.

Conclusions:

• The integration of SPEA-II with the ATM offers a robust solution for medical image captioning.
• Optimizing initial parameters significantly enhances the effectiveness of attention-based captioning models.
• This approach represents a substantial advancement in generating accurate and relevant descriptions for medical images.