|
Multimodal Thyroid AI | Learning from CAD
February 27, 2022
|
|
|
|
|
Together with
|
|
|
|
“As long as we’re all trying to do the best thing for the patient, we’re going to come out okay.”
|
|
Triad Radiology Associates’ Lauren Golding, MD on how to prepare for radiology’s value-based future.
|
|
|
I’m happy to share the latest Imaging Wire Show featuring Triad Radiology Associates CEO, Lauren Golding, MD. Join us for a great discussion covering TRA’s strategic pillars, AI in radiology, how to outsource intelligently, and preparing for radiology’s value-based future.
|
|
|
|
An MGH and Harvard Medical team developed a multimodal ultrasound AI platform that applies an interesting mix of AI techniques to accurately detect and stage thyroid cancer, potentially improving diagnosis and treatment planning.
The Platform – The platform combines radiomics, topological data analysis (TDA), ML-based TI-RADS assessments, and deep learning, allowing them to capture more data, minimize noise, and improve prediction accuracy.
The Study – Starting with 1,346 ultrasound images from 784 patients, the researchers trained the multimodal AI platform with 362 nodules (103 malignant) and validated it against a pair of internal (51 malignant, 98 benign) and external (270 malignant, 50 benign) datasets, finding that:
- The platform predicted 98.7% of internal dataset malignancies (0.99 AUC)
- The platform predicted 91.4% of external dataset malignancies (0.94 AUC)
- The individual AI methods were far less accurate (80% to 89% w/ internal)
- A version of the platform accurately predicted nodal pathological stages (93% for T-stage, 89% for N-stage, 98% for extrathyroidal extension)
- The platform predicted BRAF mutations with 96% accuracy
Next Steps – The researchers plan to validate their multimodal platform in prospective multicenter clinical trials, including in low-resource countries where it might be particularly helpful.
The Takeaway
We cover plenty of ultrasound AI and thyroid cancer imaging studies, but this team’s multi-AI approach is unique and appears promising. A multimodal AI platform like this might make thyroid cancer diagnosis more efficient and less subjective, avoid unnecessary biopsies, allow non-invasive staging and mutation assessment, and lead to more personalized treatments. That would be a major accomplishment, and might suggest that similar multimodal AI platforms could be developed for other cancers and imaging modalities.
|
|
|
Einstein & Bayer’s Injection System Upgrade
See how Einstein Healthcare Network reduced its syringe expenses, enhanced its syringe loading, and improved its contrast documentation when it upgraded to Bayer Radiology’s MEDRAD Stellant FLEX CT Injection System.
|
|
IHIE & Nuance Cut Study Retrieval Times
Do your radiologists want faster and less manual access to imaging studies? See how the Indiana Health Information Exchange (IHIE), the largest inter-organizational clinical data repository in the US, cut its imaging study retrieval time by 94% when it adopted Nuance PowerShare.
|
|
- Learning from CAD: A new JAMA editorial warned that AI might do more harm than good if we don’t apply lessons learned from mammography CAD. The authors recounted how regulatory and reimbursement decisions drove widespread CAD adoption, but created a range of unintended consequences (false positives, unnecessary tests, higher care costs). To avoid history repeating, the authors suggested that we: (1) Test how physicians are influenced by AI software design; (2) Base AI reimbursements on real-world patient outcomes; (3) Revise the FDA clearance process to encourage ongoing AI improvements; (4) Address how AI impacts medical-legal risks for image interpretation.
- Second Class Second-Opinions: A new JACR study detailed the significant radiologist labor required for second-opinion breast imaging interpretations, arguing that these studies should be billed at the same wRVU values as original interpretations. The authors analyzed 2,216 re-interpretations performed over a 3-year period, estimating that the second opinions required far more radiologist effort than they were able to bill with the current per-report credit system (2,660 vs. 284 wRVUs).
- Gesund Emerges: Gesund emerged from stealth last week, announcing a $2M pre-seed round and an interesting strategy that positions it as an imaging AI Contract Research Organization. Gesund’s “federated validation” platform allows developers to test and validate their AI models against Gesund partner hospitals’ on-site clinical data (without exposing patient info or requiring 3rd party cloud services). Gesund appears to provide an answer to growing calls for better AI validation, potentially with less labor and compliance challenges than current processes.
- Radiologist Pay Gap: A new JAMA study (54k academic physicians, 45 specialties) revealed that female academic radiologists start their careers with 9% lower salaries than their male colleagues ($373k vs. $400k avg.) and are paid 3% less during their 10th year ($432k vs. $447k avg.). Radiology has a slimmer gender pay gap than most other specialties, as the average female physician starts with a 10% lower salary and is paid 9% less during her 10th year.
- MIT’s Nanophotonic Scintillators: MIT researchers discovered a way to improve X-ray scintillator efficiency “by at least tenfold, and perhaps even a hundredfold,” which could significantly improve X-ray and CT image quality and radiation dosage. Unlike most recent scintillator research that’s focused on adopting new materials, the MIT team changed the surface patterns on existing scintillator materials (e.g. adding wave-like ridges) to create “nanophotonic scintillators” with optical properties that allow significantly more emissions.
- Viz ANEURYSM: Viz.ai announced the FDA clearance of its new Viz ANEURYSM module, which uses AI to automatically detect suspected cerebral aneurysms in CTAs, and then leverages the Viz Platform for care and follow-up coordination. The clearance continues Viz.ai’s care coordination platform expansion, which began with a focus on stroke detection / coordination, before adding new modules for PE, aortic disease, and cerebral aneurysms in the last three months.
- Charlotte Hack: North Carolina’s Charlotte Radiology announced that hackers gained access to its servers in December 2021, exposing documents containing a range of patient information (e.g. names, address, DOB, physicians, diagnosis/treatment info, and some SSNs). Although health system ransomware attacks justifiably get all the publicity, at least six US practices and imaging centers have disclosed security incidents in the last year.
- UHG’s Change Acquisition Block: The DOJ recently filed a suit to block UnitedHealth Group’s Optum subsidiary from acquiring Change Healthcare, alleging that the acquisition would significantly reduce competition by giving UHG a 75% share of the healthcare claims processing market. While UHG called the DOJ’s argument “deeply flawed” and committed to fighting the case, the suit suggests that the Biden administration is following through on its plans to scrutinize healthcare deals more closely, meaning that a demonstrable consumer benefit will be an essential component of future merger proposals.
- Ovarian Cancer Ultrasound AI: A team of China-based researchers developed an ultrasound AI model that accurately detects ovarian cancer, both independently and when used by radiologists. The researchers trained the DL model with pelvic ultrasound exams from 135k patients (34,488 w/ ovarian cancer) and validated it against one internal dataset set and two external sets. The model detected cancer in the internal and first external datasets more accurately than 35 radiologists without AI (Internal: 88.8% vs. 85.7%, External #1: 86.9% vs. 81.1%), while six less-experienced radiologists significantly improved their diagnostic accuracy when they used the AI model (87.6% vs. 78.3%).
- Publicizing the Patient Experience: A new AJR study suggests that posting patient survey results online might encourage radiologists to prioritize the patient experience. The researchers analyzed 71,885 Brigham and Women’s Hospital patient surveys that were distributed after outpatient procedures (2,703 for radiologists), finding that experience scores for radiologists and other physicians (94.2 to 97.1; 95.7 to 96.3) increased during the 19 months after BWH began posting scores online. The radiologist surveys were only performed following image-guided procedures, but these results suggest that publicizing survey results might increase most specialties’ emphasis on the patient experience.
- Healthcare Resilience: A recent USA Today / Ipsos survey of 1,170 adult healthcare workers found that many are showing resilience two years into the pandemic, with 80% reporting that they are satisfied with their current job and a majority feeling “hopeful” (59%), “motivated” (59%), or “optimistic” (56%) about going to work. Despite these positives, there were several warning signs of the pandemic’s ongoing strains, such as the fact that the 58% “hopeful” stat is down from 76% last year and over a third (39%) of the respondents agreed that “the American healthcare system is on the verge of collapse.”
|
|
- CD burning issues? Check out this one-minute video showing how Novarad’s CryptoChart image sharing solution allows patients to easily access and share their medical images using personalized, highly secure QR codes (not CDs).
- The American Medical Association recently added new CPT III codes for quantitative CT tissue characterization, paving the way for more health systems to adopt Nanox AI’s HealthCCSng CAC scoring population health solution.
- Check out this Blackford Analysis video detailing how its AI platform streamlines AI adoption and workflows, allowing radiology teams to achieve AI’s clinical benefits without operational sacrifices.
- Over 9 out of 10 people who should be screened for lung cancer aren’t, and nearly 50% of lung cancer cases are caught in the advanced stages. We know from prostate and breast cancer screening that clear guidelines and increased screening saves lives. But lung cancer screening has been challenging. Riverain strives to make everything about the lungs clearer, so they assembled this resource page for anyone interested in starting or improving their lung screening program.
|
|
|
|
|
