Artificial Intelligence in Medicine: Predicting Alzheimer’s Disease Based on Accelerated Brain Aging

Dementia is not fun. The symptoms and conditions that signal mental deterioration are a slow and inexorable crumbling process that eventually incapacitates a person. This process places burdens on the individual patient, his or her loved ones, and even society as people are living longer.

According to the Alzheimer’s Foundation, “Dementia is a general term for loss of memory, language, problem-solving and other thinking abilities that are severe enough to interfere with daily life.” While Alzheimer’s is the thought to be the most common cause of severe dementia, the risk of memory loss and cognitive dysfunction increases with aging. The World Health Organization (WHO) reports that 50 million people worldwide suffer from dementia, and there are nearly 10 million new cases each year. In fact, WHO says, “The impact of dementia on carers, family and society at large can be physical, psychological, social and economic.”

Presently, there is no known cure for dementia. However, detecting it early offers the best hope for optimal management of troubling behavioral and psychological symptoms as well as accompanying physical illness. Developing a care team as soon as possible helps alleviate the load borne by family and friends; early diagnosis also provides data to foster creation of large-scale social programs, especially in less advantaged communities where dementia appears to exist at higher rates.

Can Artificial Intelligence predict dementia and Alzheimer’s?

A recent paper by Huang, et al (2021)^[i] explored the possibility of using brain imaging (T1-weighted MRI) to identify aging-related differences between patients with mild memory-loss cognitive impairment vs. healthy controls (no cognitive impairment). The advantage of using non-invasive MRI is the ability to quantify image features in order to train Artificial Intelligence (AI) to predict a trajectory of abnormal brain aging. Thus, image-based biomarkers for risk of dementia and Alzheimer’s would be available to clinicians prior to a patient developing more challenging symptoms.

The concept of applying AI to raw imaging data in order to derive biomarkers for brain age prediction is not new. In 2017, Cole, et al.^[ii] wrote that “…age predictions can be accurately generated on raw T1-MRI data, substantially reducing computation time for novel data, bringing the process closer to giving real-time information on brain health in clinical settings.” This means not only obtaining accurate biomarkers but also speed and efficiency to reduce calculation time and workload for radiologists and neurologists.

What is predicted age difference (PAD)?

As people grow older, age-related changes occur in the brain. There is a normal range of mental and cognitive slowdown that “golden agers” joke about, things like “Where did I leave my car keys?” and “I used to be good at remembering names but now I can hardly remember my own!” The age at which this becomes noticeable can vary widely among individuals. Over a decade ago, Dosenbach, et al. (2010)^[iii] demonstrated a machine learning model trained on neuroimaging of normal aging brains could accurately predict brain age in healthy individuals. However, the age-related probability of dementia and neurodegenerative disease is no joking matter. Thus, if a 55-year-old individual undergoes a T1-weighted MRI and the result looks like the brain of a 63-year-old, it’s a significant deviation from what the brain is expected to look like at the person’s chronological age of 55; there is a predicted age difference (PAD) of 8 years. The higher the PAD number, the greater the projected trajectory toward severe dementia/Alzheimer’s disease.

For the Huang study, the team trained a machine learning model on 975 controls from two imaging datasets. The individual images were correlated with clinical records of cognitive impairment, genetic risk factors, pathologic (tissue, blood) markers of Alzheimer’s disease, and clinical progression in patients with mild memory impairment. Finally, their model was tested on 270 healthy controls and 185 patients with mild cognitive impairment to predict the “actual age” of their brains.

As might be expected, those with existing cognitive dysfunction had higher PADs than the healthy controls, and this was “significantly associated with individual cognitive impairment in several cognitive domains” for the patients with impairment. The authors also noted carriers of specific pathology biomarkers associated with Alzheimer’s disease had higher PADs than noncarriers. They concluded that AI-predicted PAD is “a sensitive imaging marker related to individual cognitive differences in patients with aMCI [amnestic cognitive impairment].”

Once again, Artificial Intelligence—in this case, machine learning trained on T1-weighted MRI brain scans—shows its potential to aid not just individuals and families, but global populations by predicting declining cognitive function and the increasing burden of Alzheimer’s disease.

NOTE: This content is solely for purposes of information and does not substitute for diagnostic or medical advice. Talk to your doctor if you are experiencing pelvic pain, or have any other health concerns or questions of a personal medical nature.

[i] Huang W, Li X, Li H, Wang W et al. Accelerated brain aging in amnestic mild cognitive impairment: relationships with individual cognitive decline, risk factors for Alzheimer disease and clinical progression. Radiology: Artificial Intelligence. Published Online: Jun 23 2021 https://doi.org/10.1148/ryai.2021200171
[ii] Cole JH, Poudel RPK, Tsagkrasoulis D., Caan MWA. et al. Predicting brain age with deep learning from raw imaging data results in a reliable and heritable biomarker, NeuroImage (2017), doi: 10.1016/j.neuroimage.2017.07.059.
[iii] Dosenbach NUF, Nardos B, Cohen AL, Fair DA et al. Prediction of Individual Brain Maturity Using fMRI. Science. 2010;329:1358-1361.