Turkish Chest X-Ray Report Generation Model Using the Swin Enhanced Yield Transformer (Model-SEY) Framework

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Natural Language Processing

Background:

• Automated medical report generation from chest X-rays is challenging.
• A significant gap exists in Turkish-language automatic reporting systems.
• Accurate interpretation of medical images is crucial for clinical decisions.

Purpose of the Study:

• To develop an AI model for automatic generation of Turkish medical reports from chest X-ray images.
• To address the unmet need for Turkish-language medical reporting systems.
• To enhance clinical decision-making support.

Main Methods:

• A deep learning model, Model-SEY, was developed.
• Swin Transformer was used for image feature extraction.
• CosmosGPT architecture, adapted for Turkish, handled text generation.

Main Results:

• A new dataset was created from Elazıg Fethi Sekin City Hospital and Indiana University.
• Model-SEY achieved scores of 0.6412 (BLEU-1) and 0.2240 (ROUGE).
• Generated reports were found to be meaningful and consistent.

Conclusions:

• Model-SEY represents a pioneering effort in Turkish deep learning-based autonomous medical reporting.
• The system shows potential to reduce diagnostic errors and aid physicians.
• This work advances AI applications in medical imaging for the Turkish language.