Continuous amyloid staging complements global amyloid burden in early Alzheimer's disease

Alzheimer's disease begins its work long before a person forgets a name or gets lost on a familiar street. Decades before these obvious signs appear, a sticky protein called amyloid starts to clump together in the brain. Scientists can see these clumps using a special camera called a PET scanner, which takes pictures of the brain's chemistry. For a long time, researchers have measured how much of this protein exists by averaging the amount across the entire brain, giving a single number that represents the total burden. This global measure is useful, but it has a blind spot. In the very earliest stages of the disease, the protein might only be gathering in a few small, specific neighborhoods of the brain. If you average the whole brain, those small, early changes can get lost in the noise, making it hard to tell exactly how far the disease has progressed or who is at risk next.

A team of researchers has developed a new way to look at these brain scans that captures this early, subtle activity. Instead of just counting the total amount of protein, they created a system that maps the specific pattern of where the protein is located. By training a computer to recognize the order in which these patterns appear over time, they built a tool that can place a person on a continuous scale of disease progression. This new method, which they call DAPS, does more than just confirm that amyloid is present; it reveals how the disease is moving through the brain, even when the total amount of protein seems unchanged. The researchers found that this new scale can predict who will lose their memory faster, even among people who have the same total amount of amyloid in their brains.

The study began with data from thousands of people who had undergone brain scans over several years. The researchers used a type of artificial intelligence that learns by itself, without being told who has the disease or who does not. They fed the computer thousands of brain images showing the distribution of amyloid protein across sixty-eight different regions of the cortex. The computer's job was to compress this complex map of sixty-eight regions into a single number that represented a stage of the disease. To make sure this number made sense over time, the researchers taught the computer a simple rule: if a person returns for a second scan years later, the disease stage number should either stay the same or go up, never down. This ensured that the tool respected the natural flow of time and the progression of the illness.

When the researchers tested this new tool, it proved to be much better at keeping the timeline straight than previous methods. Older approaches sometimes gave a person a lower stage number on a later scan than they had on an earlier one, which contradicts how the disease works. The new system avoided these errors almost entirely, creating a smooth, logical path that matched the real-world order of visits. More importantly, the researchers discovered that this new scale offered information that the old global count missed. They found a specific point on their scale where the relationship between the pattern of the disease and the total amount of protein changed. Before this point, the pattern of amyloid was shifting and becoming more organized in specific areas, even though the total amount of protein in the brain remained relatively steady. After this point, the total amount of protein and the pattern of the disease began to rise together.

This distinction turned out to be crucial for understanding risk. The researchers took two groups of people who had the exact same total amount of amyloid in their brains at the start of the study. One group was in the early stage of the new pattern-based scale, and the other was in the later stage. Even though their total protein counts were identical, the people in the later stage of the pattern scale began to lose their memory much faster than those in the early stage. This finding held true in two different large groups of people, including a separate group of participants who were all cognitively normal at the start. The tool successfully identified which individuals were on a steeper path toward decline, simply by looking at the shape of the amyloid distribution rather than just the total volume.

The researchers also looked closely at which parts of the brain were driving these early changes. They found that the regions most responsible for the shift in the early stage were the precuneus and the posterior cingulate, areas deep in the center and back of the brain. These are the same areas that other studies have identified as the first places where amyloid tends to accumulate. This suggests that the new tool is not just a mathematical trick, but a reflection of the actual biological process of the disease. By focusing on the spatial arrangement of the protein, the system can detect the disease's footprints before they become large enough to change the overall average.

The implications of this work are significant for how we might study and treat Alzheimer's in the future. Because the tool can identify people who are progressing faster even when their total amyloid levels look similar, it could help scientists select the right participants for clinical trials. Instead of grouping everyone with the same total protein count together, researchers could use this new scale to find those who are at the highest risk of decline. This could make it easier to see if a treatment is working, especially in the earliest phases of the disease when changes are subtle. The researchers note that while their tool was built using a specific type of brain scan, the logic behind it—looking at patterns rather than just totals—could apply to other ways of measuring the disease. The work suggests that to truly understand the beginning of Alzheimer's, we must look not just at how much is there, but exactly where it is going.