Lori A Carr

Centers for Medicare & Medicaid Services (CMS) and the Food and Drug Administration (FDA) using AI to fight fraud: The CMS is using tree-based artificial intelligence models and deep learning approaches in a pilot program to identify fraudulent Medicare claims, HHS Chief Information Officer Karl Mathias says. At the FDA, computer vision AI is being put to use to detect counterfeit drugs within seconds, and the NIH is using AI to audit research proposals, Mathias says.

Smartphone health app developers investing in clinical studies: Advances in technology are slowly turning smartphones into medical devices that measure signs of health and transmit data to clinicians; however, the accuracy and clinical utility of most health-related smartphone tools and apps still must be validated, says Eugene Yang of the University of Washington. Technology that has not been validated against approved medical devices can yield false positives and negatives that drive health spending higher, so some developers have been investing in clinical trials.

Blood test for colorectal cancer DNA predicts recurrence: The presence of circulating tumor DNA strongly predicted colorectal cancer recurrence after resection, and testing for ctDNA could help oncologists identify patients who are likely to benefit from chemotherapy and which ones may be able to forgo it safely. The study, published in Nature Medicine, found a tenfold risk of cancer recurrence among patients with stage II, III or IV colorectal cancer who had small fragments of tumor DNA in their bloodstream after surgery, compared with patients who had no detectable ctDNA.

CT scan finds hypertension-linked hormone dysfunction: A CT scan that tracks excessive adrenal gland hormone secretion linked to high blood pressure has demonstrated efficacy as a diagnostic tool compared with adrenal vein sampling in a study published in Nature Medicine. A tracer for the scan is being tested, and researchers say it shows promise as a way to identify surgically treatable hypertension.

Wearable IDs posttraumatic neuropsychiatric sequelae biomarkers: A study in JAMA Psychiatry found that lower 24-hour activity variance and changes in six (6) rest-activity measures were associated with increased pain severity and pain changes over time, respectively. These were tracked using Verily Life Sciences’ Study Watch in patients who wore the device for 21 hours or more per day for eight (8) weeks following a traumatic event, such as a serious fall or car crash.

Deep learning predicts aggressive brain cancer progression: Researchers at the Universities of Waterloo and Toronto (Canada) are using data from patients with glioblastoma multiforme who have declined treatment to develop a deep learning model to identify patient-specific tumor characteristics linked with tumor progression. The model produced patient-specific progression estimates based on MRI data and was validated on synthetic tumors with known parameters, researchers reported in the Journal of Theoretical Biology.

Google has announced the launch of the Google Pixel Watch, which has been integrated with the health and fitness portfolio of Fitbit. The device uses machine learning to monitor heart rate once every second, checks breathing rate, provides sleeping insights and spots signs of atrial fibrillation (AFib).

Researchers are exploring strategies to identify when AI suggests a solution for clinical decision-making that is not effective in practice.

High confidence sensor accuracy extends surgeons’ navigation of devices used in minimally invasive procedures.

A small sponge called the SymphNode, which can carry cancer drugs and be implanted next to a tumor, has been developed by researchers at UCLA. The device has been tested in mice models with melanoma and breast cancer and achieved reduced tumor size in 100% of treated melanoma subjects and 80% breast cancer subjects.