- Deep Learning Could Improve Cohort Selection for Clinical Trials
- AI Supports Clinical Trial for Rare Genetic Disease Treatment
For the BEACH study, the tool will be used to automatically identify patients with suspected ICH who meet the trial inclusion volumetric assessment criteria, defined as ICH volume between 10mL and 60mL. The tool will notify research team members of trial candidates via a HIPAA-compliant web- and phone-based app.
“By incorporating Viz RECRUIT software into the BEACH trial, we’re enabling more patients with intracerebral hemorrhage to get access to novel treatments like MW189,” said Daniel Hanley, MD, co-principal investigator of the BEACH trial and a professor of neurology at Johns Hopkins University, in the press release. “Success for the development of novel treatments like this is dependent upon increased enrollment in clinical trials, which in turn opens the door for larger trials of MW189 in acute [central nervous system] injury and age-related dementias.”
There is currently no approved treatment for ICH, and the condition often leads to high rates of long-term disability and mortality, making expedited enrollment in the BEACH study key to improving patient outcomes, the press release states.
“Clinical trial enrollment is often a bottleneck when it comes to developing novel therapies… Together, with Johns Hopkins, we have an exciting opportunity to increase access to the BEACH trial and help to enhance neurologic recovery and outcomes for patients,” said Jayme Strauss, RN, chief clinical officer at Viz.ai, in the press release.
This effort to leverage artificial intelligence (AI) to improve clinical trial enrollment is part of a larger trend to use technology to improve clinical trials.
Researchers in 2019 revealed that deep-learning models could identify key features for cohort selection in clinical trials, improving selection while significantly reducing associated time and costs.
Cohort definition and selection are extremely time-consuming tasks because of the large number of patient records that have to be manually reviewed by researchers, the research team explained. These difficulties can be compounded by variations in how the information is recorded, medical coding mistakes, and sparse or missing data.
To address this, the researchers tested different deep-learning models for cohort selection, including a simple convolutional neural network (CNN), a deep CNN, a recurrent neural network (RNN), and a hybrid model combining both CNN and RNN.
The research team found that all the models achieved high performance, indicating deep learning’s potential to streamline clinical trial cohort selection.
By 