ViTCAI: A Vision-Language Model for Automated Triage and Captioning of Postoperative Incision Images Submitted by...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Computer Vision

Background:

• Surgical site infections (SSIs) are frequent postoperative complications impacting patient health and healthcare expenses.
• Manual review of patient-submitted wound images for remote monitoring is inefficient due to variability and volume.
• Automated tools are needed for accurate and efficient assessment of postoperative incision images.

Purpose of the Study:

• To develop and evaluate an automated vision-language model for triage and captioning of postoperative incision images.
• To improve the accuracy and consistency of identifying surgical incisions and SSIs from patient-submitted images.
• To support clinical decision-making and enhance diagnostic efficiency in postoperative care.

Main Methods:

• Development of Vision Transformer for CAptioning of Incision Images (ViTCAI), a vision-language model.
• Fine-tuning ViTCAI on a clinically annotated dataset of postoperative incision images.
• Evaluating ViTCAI's performance in descriptive accuracy for identifying surgical incisions and SSIs.

Main Results:

• ViTCAI demonstrates improved descriptive accuracy in identifying surgical incisions and SSIs.
• The model provides consistent and detailed captions for postoperative images.
• Automated captioning assists in reducing manual review workload and enhancing diagnostic efficiency.

Conclusions:

• ViTCAI offers a promising automated solution for postoperative wound monitoring.
• The model can aid clinicians in timely detection and management of surgical site infections.
• Automated image analysis can significantly improve efficiency and accuracy in postoperative care.