Using AI to battle Alzheimer's

Among those taking on the challenge are researchers Teresa Wu, Yi Su and Jay Shah. 

Wu is a professor and Shah is a graduate research associate in the School of Computing and Augmented Intelligence, one of the seven Ira A. Fulton Schools of Engineering at Arizona State University. Su is the director of computational image analysis at the Banner Alzheimer’s Institute, or BAI, in Phoenix, a part of Banner Health.

Wu is also co-director of ASU-Mayo Center for Innovative Imaging, and Su is one of the center’s affiliate faculty members.

Wu and Su are the co-senior authors and Shah the lead author of a recently published paper in Alzheimer’s & Dementia, the journal of the Alzheimer’s Association, “Deep residual inception encoder-decoder network for amyloid PET harmonization.”

The research paper describes the intricate details of leveraging deep learning techniques and amyloid imaging with the use of positron emission tomography, or PET, tracers to provide a more accurate interpretation and better quantitative analysis as a basis for Alzheimer’s detection and diagnostics.

Amyloid plaques, as described by news-medical.net, are aggregates of abnormally configured proteins that form in spaces between nerve cells. The proteins are thought to play a central role in the development of Alzheimer’s disease. The amyloid plaques first develop in the areas of the brain associated with memory and other cognitive functions.

The brain contains tens of billions of neurons that process and transmit information by electrical and chemical signals, sending messages between different parts of the brain, and from the brain to muscles and organs. Alzheimer’s disease disrupts communications between neurons, causing loss of body function and cell death.

In their work, Wu, Su, Shah and others on their research teams from ASU and BAI are exploring a more efficient way to analyze images produced by amyloid PET tracers, the in vivo technology used to quantify beta-amyloid deposition in the brain, which is a precursor for Alzheimer’s disease.

Multiple PET tracers are used in different research studies, and the differences between them pose a challenge for developing a consensus on accurate quantitative comparisons for Alzheimer’s diagnosis, the researchers say.  So, their work is instead aimed at “harmonizing” these PET tracers through deep learning algorithms that can be used by medical experts to quantify amyloid depositions more effectively.

“By harmonizing, we are talking about making the PET images that are obtained by using different types of tracers look more similar, and also making the quantitative measures obtained from these images more comparable,” Wu says. “The goal is to allow the different amyloid PET tracers to be used interchangeably after applying our harmonization technique, so that clinicians and researchers can make the same interpretation, no matter which tracer is used.’’

The researchers have filed for a provisional patent for the harmonization technique.

“I think this work will accelerate research so that people can better leverage the data that have already been collected or from ongoing studies using different PET tracers,” says Su, whose expertise includes neuroimaging and Alzheimer’s disease.

“It will help us better understand how Alzheimer’s disease is developed, how it progresses over time, and enable data-driven approaches that facilitate Alzheimer’s disease research,” he says. “The ultimate goal is to develop effective treatments that might slow down disease progression, prevent or even cure the disease.”

The project combines Wu’s expertise in health informatics with Su’s experience in medical imaging.

Shah, a computer scientist, is adding to the mix his knowledge of machine learning and deep learning techniques, both of which intersect with artificial intelligence concepts and technologies.

“I want to improve on the modeling used to help make advances in clinical research using the combined capabilities of both machine learning and deep learning,” Shah says.

By doing that, the researchers hope to take steps toward expanding methods they have specifically designed to reveal the signs and progression of Alzheimer’s disease.

The project seems likely to lead to other patents for new techniques and methodologies the team is developing, as well as potential commercialization of the specific tools the researchers might create to perform those methods and techniques.

Progress on that front would open possibilities for support of their work from the federal Small Business Innovation Research and Small Business Technology Transfer programs, Wu says.

The research currently has funding from the Arizona Department of Health Services and from several National Institutes of Health, or NIH, grants. With advances toward their goal, Wu, Su and Shah hope to attract additional NIH support.

Wu emphasizes that whatever transpires on the commercialization and funding fronts, she and her colleagues want to ensure that the tools and knowledge they produce will be made available for use in further studies by medical researchers and to educate students.

“We want to put the results of what we are discovering into the hands of experts in many areas of science and engineering research and development,” Wu says. “That’s what will be needed to win the battle against Alzheimer’s.”