Dr. Vijaya Kolachalama: Six-Month Update
Dr. Vijaya Kolachalama and each member of his team earned Toffler Scholar Awards in 2021 at Boston University. They are studying how artificial intelligence (AI) can be used to provide accurate diagnosis of Alzheimer’s disease and other cognitive disorders. Earlier this summer, we published a piece about Dr. Kolachalama and the groundbreaking research that he and his team are doing.
Over the past six months, Dr. Kolachalama and his team have made promising strides in advancing their research, validating their models, and achieving their penultimate goal: creating AI models that can be used in primary care settings, both independently and alongside physicians, to diagnose or detect Alzheimer’s disease and other cognitive disorders.
Timely and Accurate Diagnosis Is Critical
Historically, the biggest challenge faced with Alzheimer’s disease is that, by the time the patient is diagnosed, it’s too late. More often than not, a patient’s first visit is to their primary care provider. This usually involves some preliminary tests, a waiting period to assess results, and then one or more follow-up visits.
If the primary care physician deems it necessary, they will refer the patient to a specialist. Depending on that specialist’s availability, it may take months before the patient can secure an appointment. From there, the waiting, referrals, tests, and months continue to build. The more time that passes, the less likely a timely diagnosis becomes.
Assistive technologies like the AI models that Dr. Kolachalama and his team are developing could prevent, if not entirely eliminate, this issue. And the benefits of these tools do not end there.
Bridging a Growing Gap of Specialized Expertise
In addition to making accurate diagnosis of Alzheimer’s disease possible, these AI models would also help bridge an increasing and alarming gap of specialized expertise: Patient demand for neurologists is steadily outpacing supply. By 2025, it’s predicted that the shortfall of neurologists will exceed patient needs by 19 percent (Neurology, June 15, 2021; 96 ).
If general practitioners had access to assistive tools and technology that could perform, in Dr. Kolachalama’s words, “at the neurologist level,” the impact on the field of cognitive disease intervention, management, and treatment would be transformational and profound. Fortunately, the team’s AI models are already delivering on their promise.
Validation testing has shown that the models are successfully “diagnosing” Alzheimer’s disease. In fact, when compared to diagnoses generated by neurologists, the models are slightly more accurate. Brain, a journal of neurology, published a detailed article about this research in May of 2020 (Volume 143, Issue 6, June 2020).
Now that the models are validated, Dr. Kolachalama and his team are focusing on extending their work to demonstrate efficacy at scale.
Expanding Their Knowledge Base
The team’s previous work was based on data from 1,000 cases. The cases included people with normal cognition, some in various stages of Alzheimer’s, and 11 who had died of the disease. Thanks to their progress and success to date, Dr. Kolachalama and his team now have access to data from over 8,000 cases, including approximately 150 that are post-mortem.
They also have access to more neurologists and neuroradiologists. In their first round of research and validation testing, they consulted with 11 neurologists representing various specialties. They are now partnering with about 27 clinicians — 7 neuroradiologists and about 20 neurologists — representing locations ranging from Nebraska to China. This expanded breadth of both data and expertise has enabled Dr. Kolachalama and his team to generate even more robust and demonstrative results.
Testing AI Models in a Clinical Setting
The team has now completed about 80% of the work needed to validate their AI models at scale. Once they are finished, they will be able to tell a story in terms that science, technology, and medicine understand: one with a plot that unfolds in data-backed statistics, figures, and percentages that confirm the AI models’ viability.
When this step is complete, Dr. Kolachalama and his team can move onto the next critical phase in their research: performing studies to test their AI model within a clinical setting. In this scenario, the model and a neurologist will be working in tandem and in real time.
The two will collect a patient’s data, process it, compare it to the data they have each reviewed to date, and generate a diagnosis. The neurologist can then review the results (i.e., the diagnosis) generated by the model and see how closely it matches their own. Combining the results of these studies with those generated by their previous research, Dr. Kolachalama and his team can continue to refine their models.
Success at this next stage of the research would be more than promising. It would provide demonstrable proof that the models can accurately diagnose Alzheimer’s disease and other cognitive disorders at their earliest possible stages. Most importantly, it would establish proof of principle, allowing them to pursue FDA approval.
Karen Toffler Charitable Trust Investment
The Investment from the Karen Toffler Charitable Trust will enable Dr. Kolachalama and his team to continue to research the application and use of AI models to provide an accurate diagnosis of Alzheimer’s disease. If proven effective, these AI models could be used to detect and diagnose any number of diseases and disorders, both independent of and in tandem with medical specialists. Because AI and machine learning are comparably new technologies in biomedicine, researchers and developers do not have access to adequate funding sources. Dr. Kolachalama hopes he and his team’s work will have a transformative impact on disease detection and diagnosis while also mitigating the growing global shortage of medical specialists and physicians.