Real-time generation of radiology reports using generative AI
This technology is a software platform that uses generative AI and computer vision for real-time, automated generation of radiology reports.
Unmet Need: Low-cost, efficient generation of radiology reports
Current gold standard methods for generating radiology reports, such as medical report dictation systems used by teleradiologists, require human engagement to turn an acquired image into an interpretable report for providers and patients. This process is costly, time-consuming, and can be susceptible to human error. Reducing the turnover time of report generation can offer faster, smarter, and more cost-effective medical analyses, enabling streamlined medical decisions for both providers and patients.
The Technology: Real-time radiology report generation with generative AI and computer vision
This technology is a software platform that incorporates generative AI and computer vision to take in radiology images and generate reports in under 5 minutes, compared to the current 24-hour turnaround time. This technology incorporates and optimizes human expertise to prioritize AI safety and minimizes hallucination risk by giving providers the ability to access and validate the generated reports in the cloud. The reports are broadly interpretable, enabling access to non-experts, and can be automatically translated into 120 languages. Further, entity tagging and redaction processes ensure compliance with Protected Health Information.
Applications:
- Automated generation of radiology reports
- Generation of interpretable radiology reports for non-experts
- Potential integration with other radiology cameras
- Scalable with additional training data
- Telehealth applications for low-resource environments
- Research tool for training AI models in radiology
Advantages:
- Automated and unbiased report generation
- Delivers reports faster than current methods, from 24 hours to under 5 minutes
- Cost-effective
- Generates interpretable reports for non-experts
- Translates reports into 120 languages
Lead Inventors:
Shahab Jalalvand, Ph.D.
Related Publications:
Tech Ventures Reference:
IR CU25102
Licensing Contact: Greg Maskel