Researchers have developed two new methods to improve the accuracy of Alzheimer's disease detection using machine learning. The issue with current models is that they don't account for the variation in symptoms among individuals with the same diagnosis. This means that some people with AD may be more severely impaired than others, making it harder for models to distinguish between those with and without the disease. The new methods, Soft Target Distillation (SoTD) and Instance-level Re-balancing (InRe), estimate sample-specific probabilities of having AD and have been shown to improve detection performance on two datasets. --- Why it matters: These findings matter because they provide a better understanding of how to model within-class variation in speech-based Alzheimer's disease detection, which could lead to more accurate diagnosis and treatment. Improved models can help clinicians identify individuals with AD earlier and more accurately, potentially leading to better patient outcomes. Source: https://arxiv.org/abs/2409.16322