4 Distinct Subtypes of Alzheimer’s Disease Discovered by Artificial Intelligence

For a long time, dementia researchers and experts have wondered if Alzheimer’s is actually more than one disease and whether there’s such a thing as “typical” Alzheimer’s disease. Now, thanks to artificial intelligence, we may have an answer.

Alzheimer’s disease often presents itself in different ways for different patients, and it can progress at different speeds as well. This has led researchers to suggest what has always been diagnosed as “Alzheimer’s” may actually be a few different diseases with some similarities. No consensus had really emerged until a group of scientists at Lund University in Sweden decided to use artificial intelligence to investigate the situation further.

The study, led by Oskar Hansson and published in Nature Medicine, used SuStaIn artificial intelligence technology to analyze the largest set of tau PET data to date—thousands of scans from 1,667 people with Alzheimer’s disease.

The machine-learning technology separated participants into four distinct subtypes, each with a different progression of neurofibrillary tangle pathology over time. Each subtype of Alzheimer’s also has its own clinical profile and speed of progression.

“Our data suggest the existence of multiple common AD subtypes, challenging the notion that there is such a pathological entity that can be described as ‘typical’ AD,” write the authors.

Roughly 12 percent of the scans could not be categorized into any particular subtype. Those scans were from people who had very low tau levels and were likely cognitively normal. The rest of the scans, however, where there was more tau buildup, fit cleanly into one of four.

“Once people become cognitively impaired and deposit more tau, it is easier to subtype them,” says Jacob Vogel, the first author of the study who works in Alan Evans’s lab at McGill University, Montreal.

Of the scans that could be categorized, 33 percent of people were grouped into subtype 1, which is the closest to what researchers consider “typical” Alzheimer’s. S1 Alzheimer’s has a predominantly limbic pattern of tangle deposition with Braak-like spacial progression. People with the S1 subtype tend to score better on global cognition tests than other Alzheimer’s patients but worse on memory. They were found to be more likely to carry the APOE4 allele and to have fewer tangles than the other subtypes.

18 percent of people fell into the second category, S2, where tangles mainly emerge in the parietal lobe and tend to spare the temporal lobe. S2 type people are less likely to carry the APOE4 allele, but they tend to be younger with more tangles.

Roughly 30 percent of people are predicted to be in subtype 3, and their tangles develop mostly in the posterior occipital lobe and then expand toward the front of the brain. S3 is the hardest to pin down, as it doesn’t have any specific traits that help differentiate it from the other subtypes, but patients with S3 have the slowest rate of disease progression of the four subtypes.

About 19 percent of people fall into subtype 4. Their tangles appear in the tempo-parietal cortices on the left side of the brain before spreading to the parietal and frontal cortices. Like S2 patients, S4s have more overall tau pathology, particularly on the left side of their brains. Their memory functions were found to be mostly preserved, but their cognition and language skills were lacking compared to other groups. Their condition worsened the fastest. Women were less likely to be S4 than men.

The researchers found that even atypical forms of Alzheimer’s disease, such as posterior cortical atrophy and logopenic primary progressive aphasia, fit nicely into their model. They believe these cases are extreme forms of posterior (S3) and lateral temporal (S4) subtypes.

Interestingly, the researchers also learned that the severity of the disease was negatively correlated with age. This data tells them that when the disease strikes earlier, it tends to be more aggressive.

“This is the kind of paper that can change a field,” says Bernard Hanseeuw, Massachusetts General Hospital. “We know that not all AD patients are equal and that they don’t have exactly the same set of symptoms or trajectories, but the power of their model is that they can explain both the symptoms and the progression of the disease.”

However, the researchers say it would still be tricky at this point for a clinician to recognize any of these four subtypes based on their patient’s behavior.

“S2 is the more cortical- and limbic-sparing, and people in this subtype tend to be younger, more dysexecutive, not as amnestic, and quite less likely to be ApoE4-positive,” says Hansson. “So if you see all that in the clinic you might think ‘S2,’ but for the others, it would be difficult to predict.”

Much more work is needed to find out how clinically significant the four subtypes are and how they can best be diagnosed in the real world. However, the team is hopeful that this new data can change our approach to treating the disease in the future.

“From a purely clinical standpoint, they question the existence of ‘typical’ AD, reinforce the need for careful clinical phenotyping, and suggest that subtyping may in due course allow for more individualized approaches to patient care and prognostication,” says Jonathan Schott of the University College London.

The study may also change the way future research on Alzheimer’s disease is conducted and how we view past research as we move forward. We can only wait and hope it’s the beginning of a big treatment breakthrough.