Artificial Intelligence can help predict who will develop dementia, a new study finds

 

Photo by Lukas Budimaier

If you could find out years ahead that you were likely to develop Alzheimer’s, would you want to know?

Researchers from McGill University argue that patients and their families could better plan and manage care given this extra time. So the team has developed new artificial intelligence software that uses positron emission tomography (PET) scans to predict whether at-risk patients will develop Alzheimer’s within two years.

They retrospectively studied 273 individuals with mild cognitive impairment who participated in the Alzheimer’s Disease Neuroimaging Initiative, a global research study that collects imaging, genetics, cognitive, cerebrospinal fluid and blood data to help define the progression of Alzheimer’s disease.

Patients with mild cognitive impairment have noticeable problems with memory and thinking tasks that are not severe enough to interfere with daily life. Scientists know these patients have abnormal amounts of tau and beta-amyloid proteins in specific brain regions involved in memory, and this protein accumulation occurs years before the patients have dementia symptoms.

However, not everyone with mild cognitive impairment will go on to develop dementia, and the McGill researchers aimed to predict which ones will.

First, the team trained their artificial intelligence software to identify patients who would develop Alzheimer’s, by identifying key features in the amyloid PET scans of the ADNI participants. Next, they assessed the performance of the trained AI using an independent set of ADNI amyloid PET scans. It predicted Alzheimer’s progression with an accuracy of 84 percent before symptom onset, as reported in a recent paper in Neurobiology of Aging.

The researchers hope their new AI tool will help improve patient care, as well as accelerate research to find a treatment for Alzheimer’s disease by identifying which patients to select for clinical trials.

“By using this tool, clinical trials could focus only on individuals with a higher likelihood of progressing to dementia within the time frame of the study. This will greatly reduce the cost and time necessary to conduct these studies,” said Serge Gauthier, MD, a senior author and professor of neurology and neurosurgery and of psychiatry at McGill, in a recent news release.

The new AI tool is now available to scientists and students, but the McGill researchers need to conduct further testing before it will be approved and available to clinicians.

This is a reposting of my Scope blog story, courtesy of Stanford School of Medicine.

New Imaging Method to Detect Heart Attack Risk

Image courtesy of NIH / Wikimedia Commons
Image courtesy of NIH / Wikimedia Commons

785,000 people have an initial heart attack and another 470,000 people have a recurrent heart attack every year in the United States, according to the American Heart Association. This means that an American has a heart attack every 34 seconds and one dies from heart disease every minute. A new imaging technique may help identify who is at high risk.

The primary cause of heart attacks is clogged arteries. Arteries are blood vessels that carry oxygen-rich blood throughout the body. Blood flows easily in healthy arteries with smooth walls. But blood flow is reduced or blocked completely in clogged arteries, when a substance called plaque builds up on the inner walls of the arteries.

Artery-clogging plaque is made up of fat, calcium, cholesterol and other substances found in the blood. Over time, this plaque can harden and rupture. If it breaks apart, a blood clot can form on its surface and completely block the artery, preventing blood from reaching the heart muscle and causing a heart attack. If the blood flow isn’t quickly restored, the portion of the heart fed by the artery begins to die.

Coronary angiography is the “gold standard” way to identify these plaque blockages in the heart, but it’s an invasive surgical procedure. During a coronary angiography, a thin flexible tube called a catheter is put into a blood vessel in your arm, groin or neck and threaded into your coronary arteries. Then a special die is released through the tube, making your coronary arteries visible on X-rays pictures taken as the die flows through them.

New study results, recently published in The Lancet medical journal, show that these high-risk plaque blockages can also be identified using a non-invasive imaging technique. The study was carried out by Dr. Nik Joshi and his research team from the University of Edinburgh, the Royal Infirmary of Edinburgh and the University of Cambridge.

The study involved 40 people who had recently suffered a heart attack and 40 additional people who had stable chest pain (angina). The patients were given a standard coronary angiography and a non-invasive imaging PET-CT scan.

A PET-CT scan measures metabolic activity using positron emission tomography (PET) and anatomical structure using X-ray computed tomography (CT). A trace amount of radioactive drug is injected into the patient’s vein and used to produce 3D images. Joshi and his research team used a radioactive drug called sodium fluoride (NaF).

The study aimed to show how well a PET-CT scan using sodium fluoride detected plaques that had already ruptured or were at high risk of rupturing. The coronary angiography was used as a gold standard to identify the culprit plaque deposits that blocked the arteries.

The researchers measured the sodium fluoride distribution to determine if the artery-clogging plaques took up a significant amount of the drug. In 93% (37/40) of the people who had had a heart attack, significant sodium fluoride uptake was seen in the plaque responsible for the heart attack. The average drug uptake in these culprit plaque deposits was 34% higher than anywhere else in the heart.

In 45% (18/40) of the people with stable chest pain, culprit plaque deposits also took up significant amounts of the sodium fluoride drug. For both sets of patients, the culprit plaque deposits identified by PET-CT imaging were confirmed by histology or intravascular ultrasound to have high-risk characteristics such as calcification and a dead tissue core.

Further research studies with a broad range of patients are now needed before PET-CT sodium fluoride imaging is accepted as a standard clinical technique. These studies are likely to take several years to complete. If they confirm the initial promising results, the technique could then move immediately into clinics since it is already approved and commonly used for other applications.

“If the results are confirmatory then this technique has the potential to fundamentally alter the way we treat coronary artery disease,” concluded the investigators. “It could, for example, permit the identification of the vulnerable patient with single or multiple high-risk or silently ruptured plaques, providing an opportunity to treat and modify their risk to prevent future adverse cardiovascular events.”

This is a repost of my KQED Science blog.

PET Imaging — Not for Cats or Dogs

PET ring drawingAs a medical imaging researcher, I notice when medical imaging technologies are mentioned by popular news media or medical-themed television shows. Lately I’ve been seeing PET imaging mentioned more frequently, including on TV shows like House and Grey’s Anatomy. This probably just reflects the fact that dramatically increasing numbers of PET scans are being performed in real life in clinics and hospitals. So what is PET imaging? Funny that you ask, because I just happen to do research in this field.

In this context, PET stands for Positron Emission Tomography. During a PET scan, a trace amount of biologically-active, radioactive drug is injected into the patient’s vein. The drug localizes somewhere in the patient, depending on the metabolic properties of the selected drug. The drug then emits a positron (anti-particle of the electron), and the positron annihilates with an electron in the patient’s body. The resulting energy forms gamma ray pairs that pass through the patient and are detected by the PET scanner. These detected gamma ray signals are used to create a 3-D volumetric image or picture of the drug’s concentration in the body.

PET imaging technology is unique because it images a patient’s metabolism, whereas most other medical imaging techniques measure anatomical structure. For example, X-ray CT or MRI scans can be used to identify a tumor because they show the patient’s anatomy in detail. However, PET imaging can identify if the tumor is benign or cancerous, by measuring whether or not the tumor takes up the radioactive drug. In reality, you’d really like to know both though — detailed anatomical structure and metabolic function. Recent work has demonstrated the increased clinical diagnostic value of fusing imaging technologies based on function (e.g., PET, SPECT or functional MRI) with those based on structure (e.g., CT, MRI, or ultrasound). As a result, PET and CT scanners are now typically combined into a single gantry system, so that images can be taken from both devices sequentially during a single procedure.

Since PET measures metabolism instead of anatomical structure, it is mostly used to image organs whose size or shape does not indicate whether they are functioning properly, such as the brain or heart. It is also used to diagnose diseases that exhibit an abnormal metabolism, such as cancer.

Stay tuned this week when I discuss some Alzheimer’s research that utilizes PET imaging.