RAIQC supports evaluation of Lunit INSIGHT CXR AI tool for enhancing clinician chest X-ray interpretation

10 December 2024

The team at RAIQC is pleased to have supported a recent study evaluating the effectiveness of the Lunit INSIGHT CXR AI tool in improving clinician accuracy for interpreting inpatient and emergency chest X-rays (CXRs). The study utilised RAIQC’s online capabilities to facilitate an efficient evaluation process, enabling remote participation from clinicians across specialties and experience levels.

CXRs are a critical diagnostic tool, particularly in emergency and inpatient settings, where timely and accurate interpretation can directly affect patient outcomes. The study explored the performance of the Lunit INSIGHT CXR AI tool in assisting clinicians across multiple specialties and varying levels of experience. The specialties include: radiology, emergency medicine, general internal medicine, and intensive care.

Using a multi-reader, multi-case design, the study tasked 30 clinicians to interpret 500 CXRs. The readers, grouped into three postgraduate experience levels (1–4 years, 5–8 years, and over 8 years), identified the presence or absence of ten pathologies and assigned confidence scores for their interpretations. The process was carried out once without AI assistance and then again with AI, following a four-week washout period. Ground truth for the cases was established by thoracic radiology consultants, with arbitration by a senior consultant in cases of disagreement.

The entire study was conducted remotely through RAIQC, allowing readers to participate from various locations. This remote setup eliminated logistical challenges and ensured timely completion, showcasing the flexibility and efficiency of RAIQC's platform for large-scale clinical research.

Key findings from the study include:

• Standalone AI performance: The AI correctly identified abnormalities in CXRs with an accuracy range of 83–99%. For eight out of ten pathologies, the AI achieved an accuracy above 90%.
• Reader performance with AI assistance: The accuracy of clinicians interpreting CXRs improved significantly for 80% of the pathologies when using AI assistance. For example, in the detection of fibrosis, AI assistance boosted accuracy by nearly 20%.
• Clinical relevance: The study highlights the potential for AI to help clinicians from patient-facing specialties make better decisions. As hospitals face increasing demands, AI can bridge gaps in expertise and improve timely decision-making.

The results were presented by Dr. Ruchir Shah from Oxford University Hospitals at RSNA 2024 and were awarded the Trainee Research Prize. 

RAIQC’s contribution to the study also included hosting and organising the imaging data, managing reader progress, and assisting with the collection and analysis of results. By offering the tools and platform needed to support a complex, multi-site study, RAIQC played a key role in enabling the study's completion.

The study provides valuable insights into the role of AI in addressing the challenges clinicians face in interpreting chest X-rays, particularly in settings where resources are stretched. By improving diagnostic confidence and accuracy, AI has the potential to become a key partner in patient care.
 

About RAIQC

RAIQC is a web-based platform that simulates day-to-day practice, allowing healthcare professionals, students and educators to review and report on diagnostic quality medical images in a secure online environment. Using over 6000 real-world clinical cases, RAIQC offers structured reporting study lists for training and assessment for individuals and healthcare providers across a range of imaging modalities and disease areas. The platform also provides hosting for clinical research and AI validation studies that require review of medical imaging.

For more information visit www.raiqc.com or email info@raiqc.com.

RAIQC Ltd is registered in England and Wales with company number 09863569 and its registered address at 1 & 3 Kings Meadow, Osney Mead, Oxford, OX2 0DP.