NLP Archives - The Imaging Wire

An AI algorithm that analyzed completed radiology reports found incidental cancers that weren’t being followed up by clinicians. The study, published in Journal of Surgical Research, offers a blueprint for how imaging practices can leverage AI to improve patient care and reduce malpractice risk.

Many of the early radiology AI applications focused on the technology’s ability to assist radiologists by identifying pathology hidden on medical images.

But AI is also being developed for natural language processing (NLP) tasks – sifting through massive amounts of text to find keywords that could indicate a high-risk incidental finding.

The new study falls into the latter category. Researchers at Banner MD Anderson Cancer Center in Arizona studied the implementation two years ago of Azra AI, software designed to track incidental oncology findings.

The algorithm identified high-risk lesions in outpatient and emergency reports through NLP analysis, using keywords provided by Banner researchers. It then provided a framework by which patients were navigated through follow-up care without falling through the cracks.

Over 13 months of use, Banner clinicians used Azra AI to review 961.3k radiology reports, finding…

High-risk lesions were identified in 5% (48.6k) of reports.
Clinician review confirmed 43.7k clinically significant findings, giving the software a 4.5% finding rate and 0.5% false-positive rate.
445 patients were navigated to multidisciplinary care.
After workup, 67 patients were diagnosed with cancer, representing 0.2% incidence among all validated reports and 15% incidence among navigated patients.

The researchers pointed out that working up this many incidental findings manually would have taken significantly longer, and many patients would have been overlooked entirely.

That would create significant malpractice risk for imaging providers, especially given a new report indicating that 80% of malpractice claims involve errors in judgment like missing diagnoses (see The Wire section).

Finally, it’s worth noting that the new study gives a boost to Azra AI, which is in the process of taking over AI platform company Blackford.

Azra AI executives see AI point-source algorithms available on Blackford Platform as key to their goal of offering a comprehensive AI-based framework for both detecting pathology and making sure patients are followed up.

The Takeaway

The new study on AI-based NLP analysis of radiology reports offers a tantalizing vision of how AI can help imaging providers identify patients who might otherwise fall through the cracks and get them the care they need. And there are 67 people who can vouch for that.

The last week brought two high profile studies underscoring radiology NLP’s potential to improve efficiency and accuracy, showing how the language-based technology can give radiologists a reporting head-start and allow them to enjoy the benefits of AI detection without the disruptions.

AI + NLP for Nodule QA – A new JACR study detailed how Yale New Haven Hospital combined AI and NLP to catch and report more incidental lung nodules in emergency CT scans, without impacting in-shift radiologists. The quality assurance program used a CT AI algorithm to detect suspicious nodules and an NLP tool to analyze radiology reports, flagging only the cases that AI marked as suspicious but the NLP tool marked as negative.

The AI/NLP program processed 19.2k CT exams over an 8-month period, flagging just 50 cases (0.26%) for a second review.
Those flagged cases led to 34 reporting changes and 20 patients receiving follow-up imaging recommendations.
Just as notably, this semi-autonomous process helped rads avoid “thousands of unnecessary notifications” for non-emergent nodules.

NLP Auto-Captions – JAMA highlighted an NLP model that automatically generates free-text captions describing CXR images, streamlining the radiology report writing process. A Shanghai-based team trained the model using 74k unstructured CXR reports labeled for 23 different abnormalities, and tested with 5,091 external CXRs alongside two other caption-generating models.

The NLP captions reduced radiology residents’ reporting times compared to when they used a normal captioning template or a rule-based captioning model (283 vs. 347 & 296 seconds), especially with abnormal exams (456 vs. 631 & 531 seconds).
The NLP-generated captions also proved to be most similar to radiologists’ final reports (mean BLEU scores: 0.69 vs. 0.37 & 0.57; on 0-1 scale).

The Takeaway

These are far from the first radiology NLP studies, but the fact that these implementations improved efficiency (without sacrificing accuracy) or improved accuracy (without sacrificing efficiency) deserves extra attention at a time when trade-offs are often expected. Also, considering that everyone just spent the last month marveling at what ChatGPT can do, it might be a safe bet that even more impressive language and text-based radiology solutions are on the way.

Get every issue of The Imaging Wire, delivered right to your inbox.