Novel AI model combining visual understanding and natural language processing to generate comprehensive diagnostic reports with lesion annotations.
Use of Visual-Language Segmentation (VLS) model and SAM enables precise lesion segmentation and interpretable report generation.
AI-assisted ocular ultrasound reporting improves diagnostic accuracy and reduces reporting time, validated by senior and junior ophthalmologists.
Guideline-Based Recommendations
Diagnosis
Utilize ophthalmic ultrasound imaging for detailed structural assessment of ocular conditions.
Incorporate AI models that combine image analysis with natural language report generation to enhance diagnostic precision.
Management
Adopt AI-assisted reporting tools to support clinical decision-making and personalized patient care.
Leverage lesion segmentation outputs to guide treatment planning and monitoring.
Monitoring & Follow-up
Use AI-generated reports to track disease progression and response to therapy over time.
Regularly evaluate AI model performance and update with clinical feedback to maintain reliability.
Risks
Be aware of challenges in model interpretability and reliability in clinical settings.
Ensure AI outputs are reviewed by qualified ophthalmologists to prevent misinterpretation.
Patient & Prescribing Data
9670 patients across three hospitals with diverse ocular conditions, balanced gender distribution, and mean age around 50 years.
AI-assisted ultrasound reporting demonstrated higher diagnostic accuracy and significantly reduced reporting time, supporting its use as an auxiliary diagnostic tool.
Clinical Best Practices
Combine advanced Vision-Language Models with segmentation techniques for comprehensive ophthalmic ultrasound analysis.
Engage both senior and junior ophthalmologists in evaluating AI-generated reports to ensure clinical relevance.
Integrate AI tools into existing workflows to manage increasing ultrasound data volume efficiently.
Continuously validate AI model performance using diverse, real-world datasets to ensure generalizability.
