• Dementia affects more than 55 million people worldwide, costing around $1.3 trillion per year.
• The most common form of dementia, Alzheimer’s disease, affects up to 70% of these people.
• New treatments are most effective if started early in the progression of the disease, but diagnosis is difficult at this stage as early symptoms are often dismissed as normal changes due to aging.
• Now, a new study has developed a deep learning framework that can identify the risk of progressing from mild cognitive impairment to Alzheimer’s.

In 2019, according to the Global Burden of Diseases, Injuries, and Risk Factors Study, an estimated 57.4 million people worldwide had dementia.

By 2050, the study predicts that the number will have risen to more than 150 million. And most of those people will have Alzheimer’s disease.

The cost of dementia is huge, placing enormous strains on care systems and families. Worldwide, annual costs are estimated to be $1.3 trillion per year.

Until recently, available treatments could alleviate symptoms, but none could slow or halt the progress of the disease.

Early diagnosis to ensure treatment efficiency

New monoclonal antibody treatments for Alzheimer’s disease, such as lecanemab, aducanumab, and donanemab, have been hailed as the first disease-modifying treatments.

They can clear the amyloid plaques that build up in the brain and are thought to be responsible for most symptoms of the disease.

However, these treatments are effective only if they are given early in the course of the disease. And therein lies the problem, as current diagnosis, according to the Alzheimer’s Association, relies largely on documenting mental decline, and the condition is rarely diagnosed before significant damage to the brain has occurred.

Biomarkers for Alzheimer’s disease, such as amyloid and tau proteins, can be detected in cerebrospinal fluid (CSF), but this invasive and expensive test is not entirely reliable.

Research has suggested that these biomarkers can also be detected in blood plasma but there is some way to go before these tests might be used in clinical practice.

Dr. Emer MacSweeney, CEO and consultant neuroradiologist at Re:Cognition Health, stressed the importance of early diagnosis, saying:

“With the recent and very long-awaited success in international clinical trials for new disease-modifying treatments for Alzheimer’s disease and FDA approval of aducanumab and lecanemab; there is an increasing imperative to develop inexpensive, ubiquitous, assessments to identify early individuals at risk of developing progressive cognitive decline, due to Alzheimer’s disease.”

Mild cognitive impairment and dementia risk

Many people experience mild cognitive impairment as they age, but the majority of these will not go on to develop Alzheimer’s disease.

One approach to diagnosis is to identify which individuals with mild cognitive impairment are most at risk of progressing to Alzheimer’s disease.

A new study has done just this — developing a deep learning framework that can stratify individuals with mild cognitive impairment based on their risk of advancing to Alzheimer’s disease. The research is published in iScience.

Dr. Percy Griffin, Alzheimer’s Association director of scientific engagement, not involved in this research, welcomed the study.

“If this work is validated in larger and more diverse cohorts, it will help clinicians with predicting the conversion from earlier to later stages of the disease. This is important because early detection and accurate diagnosis will enable people to take advantage of new and emerging treatments for Alzheimer’s earlier in the disease course,” he told Medical News Today.

A new way to detect subtle brain changes

The researchers who conducted the new study used data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) and the National Alzheimer’s Coordinating Center (NACC).

All the participants whose data were included in the study had mild cognitive impairment. The researchers used magnetic resonance images (MRI) and CSF biomarkers to diagnose mild cognitive impairment and Alzheimer’s disease, as well as postmortem data to confirm these diagnoses.

They separated individuals with mild cognitive impairment into groups based on their brain fluid amyloid levels. They then studied gray matter volume patterns within these groups to identify risk groups.

Radiologists analyzed the MRI scans to identify the presence and extent of atrophy in a number of regions. Brain atrophy is associated with the progression from mild cognitive impairment to Alzheimer’s disease.

The researchers then developed deep learning models to predict the progress from mild cognitive impairment to Alzheimer’s disease.

They then linked their model predictions with biological evidence, confirming Alzheimer’s diagnoses with post-mortem data.

Corresponding author Dr. Vijaya B. Kolachalama, an associate professor of medicine at Boston University Chobanian & Avedisian School of Medicine, explained

“We utilized survival-based deep neural networks in conjunction with minimally processed structural MRI, a widely available, non-invasive technique. Further, by employing state-of-the-art deep learning methods in conjunction with SHapley Additive exPlanations (SHAP), a method based on cooperative game theory and used to increase transparency and interpretability of machine learning models, we were able to identify regions particularly important for predicting increased progression risk.”

Dr. Griffin was encouraged by the study findings: “Alzheimer’s disease kills brain cells and changes the structure of the brain in several regions. In the early stages of the disease, these changes can be subtle and difficult to pick up.”

“Because the machine learning techniques used in this paper are better able to identify these subtle changes in affected brain regions, they may help improve the accuracy of prediction of conversion to later stages of the disease,” he added.

Promising development

“Early detection of those at risk is paramount and given the complexity of the brain and this disease, innovation using machine learning about of regions of the brain will likely be the best way to predict those at most risk,” said Dr. MacSweeney.

The authors suggest that their practical approach to forecasting individualized progression risk in persons with mild cognitive impairment might be useful in both clinical and research settings that have access to routinely collected structural neuroimaging data.

While welcoming the findings, Dr. Griffin added a note of caution: “The cohorts used to establish these models are not representative of our diverse communities that are affected by Alzheimer’s and other dementia. This means that racial and ethnic differences in the progression of the disease may not be captured in these models.”

“It is crucial to ensure that emerging technologies do not deepen existing disparities in healthcare. As such, these models need to be trained on larger, more diverse cohorts before they can be applied broadly,” he added.

However, Dr. MacSweeney was optimistic that this innovative approach might help earlier diagnoses.

“Unfortunately, there are millions of people who have this disease, and creating innovation at the intersection between pathology, neurology, and computer science is a very promising way forward to solve this gigantic problem,” she told MNT.

Source: Read Full Article