Predicting Progression of Alzheimer’s
Alzheimer’s disease, the sixth leading cause of death in the US, kills more senior citizens than breast cancer and prostate cancer combined. The disease affects millions of people annually, says the Alzheimer’s Association. If diagnosed early, as much as $7.9 trillion in medical and care costs could be saved, according to an article by Kyle Wiggers in Venture Beat.
Researchers at Unlearn.AI think they have a solution, the article says. The startup company designs software tools for clinical research and anticipates a major role for artificial intelligence in personalizing diagnosis and treatment of Alzheimer’s. Their paper (“Using deep learning for comprehensive, personalized forecasting of Alzheimer’s Disease progression”) on the preprint server Arvix.org describes a system for predicting disease progression. In other words, the system projects the symptoms that individual patients will have at any time in the future. The researchers published a training video on the website Unlearn.AI.
As they explained, “Two patients with the same disease may present with different symptoms, progress at different rates, and respond differently to the same therapy. Understanding how to predict and manage differences between patients is the primary goal of precision medicine. Computational models of disease progression developed using machine learning approaches provide an attractive tool to combat such patient heterogeneity.”
While presenting an interesting possibility for precision medicine in the management of Alzheimer’s disease, AI-enabled systems for tracking cognitive decline have been suggested before. Neurologists at Montreal’s McGill University created a “PET scan-ingesting algorithm that identified which patients ended up with dementia with 84 percent accuracy,” and scientists at North Carolina’s Duke University and Croatia’s Rudjer Boskovic Institute “used machine learning to pick up changes in brain tissue loss over time,” Wiggers related.
The innovative element in Unlearn.AI’s system is its unsupervised learning approach. This system utilizes data that is unclassified or unlabeled. Additionally, it can compute predictions and confidence intervals for numerous patient characteristics at the same time.
The researchers developed the system in two stages. Initially, they modeled clinical data using a Boltzmann Encoded Adversarial Machine (BEAM), a form of neural network highly effective for “classification and feature modeling tasks.” They trained and tested BEAM on the Coalition Against Major Diseases (CAMD) Online Data Repository for Alzheimer’s Disease, that includes 1,908 patients evaluated during a timeframe of 18 months including 42 variables, such as the individual components of ADAS-Cog (a frequently used cognitive subscale) and Mini-Mental State Examination (a questionnaire that measures cognitive impairment in clinical and research venues).
In the second phase, the researchers utilized the trained model to develop “virtual patients” and their related cognitive exam scores, laboratory tests, and clinical data. They ran simulations for individual patients to figure out their disease progression in such areas as orientation, word recall, and naming, which were then utilized to figure out the overall ADAS-Cog score.
The Unlearn.AI researchers claim that the unsupervised model could make accurate ADAS-Cog predictions to at least 18 months in the future. They suggest that the system can be adapted to preject the outcomes for patients who have other degenerative diseases.
The team concluded, “The approach to simulating disease progression that we describe here can be easily extended to other diseases. Widespread application of deep generative models to clinical data could produce synthetic datasets with lower privacy concerns than real medical data, or could be used to run simulated clinical trials to optimize study design. In certain disease areas, tools that use simulations to forecast risks for specific individuals could help doctors choose the right treatments for their patients.”