A deep learning tool to predict cardiovascular risk

A team of researchers from Brigham and Women’s Hospital’s Artificial Intelligence in Medicine Program and Massachusetts General Hospital’s Cardiovascular Imaging Research Center have developed an AI deep learning system that can automatically measure coronary artery calcium from computed tomography (CT) scans and predict cardiovascular events, such as heart attacks. The team trained the deep learning system on data from the Framingham Heart Study. Dedicated calcium scoring CT scans, which all participants received, were manually scored by human readers and used to train the deep learning system. The AI-based system was then applied to three additional study cohorts, comprising heavy smokers having lung cancer screening, patients with stable chest pain having cardiac CT and patients with acute chest pain having cardiac CT. In all, the deep learning system was validated in more than 20,000 patients. A test-retest analysis was conducted separately on the manual and deep learning risk scores of a subset of 252 image pairs from FTS-CT1. “Each image pair was taken consecutively within the same setup and within 1-min to 1-h difference,” the authors wrote. “The results showed a great stability between the automatically calculated calcium scores for each image per pair achieving an intra-class correlation (ICC) of 0.993 (P < 0.001), compared to the ICC of manual calculated calcium scores of 0.997 (P < 0.001).” The automated scores also independently predicted who would subsequently have a major adverse cardiovascular event. The study was published online Jan. 29, 2021, in Nature Communications.

Smartphone-based COVID-19 test

Convenience is important when you’re trying to expand testing and put the brakes on a pandemic. To that end, researchers at the University of Arizona are developing a low-cost, smartphone-enabled COVID-19 test that analyzes saliva samples and kicks back results in results in about 10 minutes. The team, led by professor Jeong-Yeol Yoon, is adapting an inexpensive approach first devised to detect norovirus that uses a smartphone microscope. The aim is to employ the method using a saline swish-gargle test developed at the university. In their latest research using water samples, the team demonstrated the “extremely sensitive mobile detection of norovirus from water samples using a custom-built smartphone-based fluorescence microscope and a paper microfluidic chip.” They analyzed the smartphone images using intensity- and size-based thresholding to eliminate background noise and autofluorescence, as well as to isolate immunoagglutinated particles. The pixel counts of the particles corresponded to the norovirus concentration of the specimen. “Although the method described is for detection of norovirus, the same protocol could be adapted for detection of other pathogens by using different antibodies,” the researchers said. Their findings appeared in the Jan. 29, 2021, online issue of Nature Protocols.

PCOS passed on in the epigenes

Researchers at the University of Strasbourg and INSERM have discovered that polycystic ovarian syndrome (PCOS) was inherited via epigenetic alterations in mouse models. PCOS is a common disorder, which causes both reduced fertility and metabolic impairments. It is also common, strongly heritable and understudied. Exposure to androgens late in pregnancy causes PCOS-like symptoms in mice, and the condition is then heritable, suggesting an epigenetic mechanism. The team compared methylation patterns of ovarian cells in control mice and third-generation PCOS mice. They found that multiple PCOS-associated genes were hypomethylated in the PCOS animals, and treatment with a methyl donor could ameliorate some PCOS symptoms. In blood samples from women with PCOS, they showed the same hypomethylation patterns seen in the mouse model. “These findings show that the transmission of PCOS traits to future generations occurs via an altered landscape of DNA methylation and propose methylome markers as a possible diagnostic landmark for the condition, while also identifying potential candidates for epigenetic based therapy,” the authors wrote. Their work appeared in the Feb. 3, 2021, online issue of Cell Metabolism.