- Dear Medical Students: A new paper in Current Problems in Diagnostic Radiology directly addressed medical students’ concerns about AI’s impact on radiology, suggesting that they have been influenced by “misinformation,” and countering that AI will actually make radiology “a more attractive specialty.” The authors argued that AI will make radiologists more efficient, reduce burnout, improve accuracy, and increase its overall impact on healthcare. Uncoincidentally, this paper comes out a few weeks after another study provided solid evidence that more med students are deciding against becoming radiologists due to AI concerns.
- RadioPharm & MD Anderson’s Theranostics Venture: MD Anderson Cancer Center and RadioPharm Theranostics announced the creation of Radiopharm Ventures LLC, which will co-develop radiopharmaceutical cancer therapeutics. The joint venture will integrate MD Anderson’s proprietary technologies in antigen discovery and molecular imaging with RadioPharm’s experience in radiopharmaceuticals development, initially focusing on creating at least four therapeutic products.
- Self-Taught CXR AI: Harvard Medical researchers developed a self-supervised AI model (trained w/ unannotated images) that detected pathologies in chest X-rays with similar accuracy as radiologists. The researchers trained their CheXzeo model on 377k unannotated CXRs and 227k corresponding radiology reports, and tested it on two external CXR data sets, finding that the model was able to accurately detect conditions such as pneumonia, collapsed lungs, and lesions. There have been other self-supervised CXR AI models, but CheXzeo is the best performing, and provides the greatest assurance that AI might not face as big of an annotation barrier as some believe.
- FDG-PET Prediction Impact: A study in JNCCN revealed that FDG-PET can predict pancreatic cancer patients’ response to neoadjuvant therapy, aiding providers’ therapy and treatment decisions. Among 202 patients with borderline resectable/locally advanced pancreatic ductal adenocarcinoma, preoperative FDG-PET (w/ CT or MRI) was superior to biochemical assessments for predicting major pathologic response to the therapy (AUCs: 0.86 vs. 0.75).
- Neusoft’s IPO Do-Over: Neusoft Medical is planning to seek an IPO on the Hong Kong Stock Exchange after several failed attempts to join Shanghai’s STAR Market and the Hong Kong Exchange (in 2020 & 2021). However, market watchers suggest that Neusoft’s hopeful Hong Kong IPO might see similar challenges due to its low-end modality focus, lagging R&D investments, and rising costs.
- Radiographer Mammogram Reads: A Radiology study out of the UK found that radiographers performed comparably to radiologists in double reading screening mammograms. The study included 401 readers (224 radiologists, 177 radiographers) who double-read 1.4M screening mammograms, finding no significant differences between radiologists and radiographers’ cancer detection rates (7.84 vs 7.53 per 1k exams), recall rates (5.0% vs. 5.2%), and positive predictive value (17.1% vs. 16.1%). Although it’s common for radiographers to handle some interpretation duties in the UK, this study raised concerns among US-based rads.
- Subtle Medical’s GSA: Subtle Medical landed a spot on the US federal government’s GSA Multiple Awards Schedule, making its SubtleMR and SubtlePET image reconstruction solutions available to the VA and other government institutions. Given the VA’s massive scanner installed base, and its focus on efficiency and patient experience, SubtleMR and SubtlePET might find a solid fit in veteran care.
- Photon-Counting CT Plaque Characterization: A new study out of Switzerland highlighted photon-counting CT technology’s (PCD-CT, an emerging CT tech) advantage over standard CT for coronary plaque characterization. Researchers analyzed exams from 20 coronary syndrome patients undergoing ECG-gated coronary CT angiography (w/ PCD-CT), finding that PCD-CT reduced blooming artifacts, allowing for improved visualization of fibrotic and lipid-rich plaque components in ultra-high contrast mode.
- At-Home BCa Biomarker Test: Namida Lab announced the launch of its Auria at-home diagnostic test, which uses tears to identify protein biomarkers related to breast abnormalities. Targeted at women with low to average breast cancer risk, Auria provides a personalized score that tells women how soon they should schedule a mammogram. Although we’ve never seen a test like this, Auria joins a growing field of biomarker and home tests that might impact image-based screenings.
- Cardiac POCUS Fast Learners: An Indiana University study found that after a brief training session (video + 10min hands-on lesson), medical students with little cardiac POCUS experience were able to perform common exams. Among 132 second-year medical students who completed the training, 72% obtained PSLA cardiac views without instructor guidance and 62% were able to identify pericardial effusion and diagnose cardiac tamponade (all with simulated patients).
- Prostate MRI Variability: A new analysis in AJR showed that the cost of prostate MRI exams can vary widely depending on the healthcare facility. Researchers analyzed 37,073 prostate MRI exams conducted at 552 US facilities, finding that the median charge per exam was $4,419, with a 26-fold variation between the lowest and highest ($593 to $15,150).
