Medical imaging extensively uses artificial intelligence (AI) to evaluate radiology reports. From chest X-rays to bone fractures, AI is being used to reduce the burden on radiologists. However, with the recent breakthrough by Oxipit, the use of AI in medical imaging may increase in multi folds.
Oxipit, a Lithuanian startup, has developed autonomous AI reporting that can identify normal chest X-rays (CXRs) without the need for radiologist intervention. The startup has also received Class IIb regulatory approval for its ‘autonomous AI’ technology.
This news has sent shockwaves through the radiology community, eliciting both awe and scepticism.
But what difference can autonomous AI make? Will AI be the new radiologist? How is Oxipit’s technology revolutionising the medical imaging landscape? Let’s find out.
ChestLink: Oxipit’s autonomous AI technology
ChestLink is Oxipit’s fully autonomous AI medical imaging product. It evaluates CXRs without radiologist involvement. Here’s how it works:
- If ChestLink confidently determines that a CXR is normal and has no actionable findings, it autonomously generates a report. This means no trained radiologist will review the diagnosis before the report is produced.
- If ChestLink cannot definitively rule out the presence of actionable findings and suspects abnormalities in the CXR, the report is directed to a radiologist for further review and reporting.
But, current AI radiology systems already perform autonomous image analysis. How does this make ChestLink different?
Current AI systems evaluate CXRs and report them to radiologists. But ChestLink is the first and only fully autonomous reporting product that confidently predicts a ‘normal’ chest x-ray without the need for radiologist verification.
How does autonomous reporting help radiologists and diagnostic centres?
Automating the scanning process will not replace radiologists. Instead, it’ll significantly reduce the workload on radiologists. It’ll address the global shortage of radiologists and burnout of existing radiologists due to increased workload.
Oxipit’s ChestLink holds promising potential for primary care and diagnostic centres, where up to 80% of chest X-rays are normal. By using ChestLink, the centres can automate 15% to 40% of daily reporting workflows and free up radiologists for serious and abnormal cases, prioritising patient care with efficiency.
ChestEye, Oxipit’s AI Imaging suite, can identify 75 radiological findings, covering 90% of daily abnormalities, that assist radiologists in better diagnosis. It also has the potential for large-scale screening projects like global tuberculosis detection.
Is ChestLink Safe?
To demonstrate its commitment to safety, Oxipit has obtained the CE mark certification in the EU (Class IIb approval) and is actively pursuing FDA clearance. It is also fully GDPR compliant.
During pilot applications, it claims zero important errors and an impressive 99% accuracy in detecting normal cases. The startup recently published an impartial study in the Radiology journal.
Oxipit’s system has also undergone rigorous regulatory scrutiny by TUV Rheinland, and it met their requirements. They proposed a three-stage framework—retrospective audit, supervised operation and autonomous reporting—which seems sensible.
From a regulatory standpoint, the device meets safety standards and will be monitored post-market. Oxipit has registered the device on EUDAMED and will update its clinical evaluation report annually with post-market data. Overall, the device is considered safe, but further scrutiny and data sharing are crucial for a comprehensive assessment.
An important note: Oxipit’s autonomy claim is specific to normal studies. Any suspicion of abnormality is still left for the radiologist’s judgement. While their marketing may state it as the first fully autonomous AI, it would be more accurate to consider it as partial autonomy or fully autonomous normal reporting.
More about Oxipit
Oxipit is a Lithuania-based AI startup founded in 2017 by Naglis Ramanauskas, Jonas Bialopetravičius, Darius Barušauskas, Vytautas Naujalis and Gediminas Pekšys. It specialises in creating AI tools for global medical imaging.
By harnessing AI and machine learning, Oxipit aims to address the global shortage of radiologists and enhance the quality of medical diagnostics.
Since its inception, Oxipit has secured a total funding of $6.6 million through three funding rounds, and the latest fundraising took place in January 2023.
Apart from ChestLink, Oxipit also offers imaging audits to revolutionise the use of AI in medical imaging.
Oxipit Quality: Oxipit’s AI-powered imaging audit
Oxipit Quality is an AI double-reading medical imaging tool that analyses radiologists’ reports to identify errors or discrepancies. It is a retrospective imaging audit that allows diagnostic centres to identify common mistakes, calling for attention or additional training to mitigate risks.
With Oxipit Quality, the startup aims to improve diagnostic performance and improve patient outcomes. Along with chest X-rays, Quality software also supports musculoskeletal, mammography and CT Pulmonary embolism study modalities.
ChestLink competitors and challenges
With an overwhelming backlog of CXRs awaiting reports, several academics, AI startups and even the tech giant, Google, are striving to address the challenges. Three other companies—Qure.ai’s qXR, Thirona’s CAD4TB and Annalise Enterprise CXR—have also received Class IIb certification for CXR AI.
While many have ventured into the realm of automated classification, none have dared to fully embrace the concept of autonomous reporting. As a result, Oxipit stands out for its assertion of complete autonomy.
Dr Nicola Strickland, former President of the UK Royal College of Radiologists, hails this development as a potential game-changer for the NHS, given the mounting CXR backlogs.
Moving forward, Oxipit needs to navigate complex legislative frameworks like GDPR, IR(ME)R, EURATOM and the CQC for the widespread adoption of ChestLink in healthcare systems. Legislative changes are needed to address practical and legal complexities and highlight the benefits of reduced human workload and potential financial and safety advantages.
While the availability of the first autonomous AI system is exciting, there are still hurdles to overcome. The speed at which existing laws can be amended or bypassed remains uncertain, and only time will reveal its impact.
What do you think? Let us know in the comments.
