This new AI can identify early Alzheimer’s disease markers with more than 99% accuracy. By evaluating brain scans of older adults, the algorithm can detect subtle changes that often take place before diagnosis, allowing doctors to offer early treatment to high-risk people.
Thus, AI successfully recognizes signs of mild cognitive impairment that generally do not produce noticeable symptoms, and is associated with changes in certain regions of the brain that can be detected on functional magnetic resonance imaging (fMRI) scans. . However, doctors don’t always identify them by looking directly at these scanners.
The researchers repurposed an existing neural network called ResNet18 and created an AI model capable of identifying these details with greater reliability. The AI was trained with 51,443 brain scans of 138 people . Then another 27,310 images were used to validate the algorithm, which was able to identify early cognitive decline with 99.99% accuracy and late MCI with 99.95% accuracy.
“Modern signal processing allows you to delegate image processing to the machine, which can complete it quickly and accurately enough,” explained study author Rytis Maskeliūnas in his study published in the journal Diagnostics. “Of course, we dare not suggest that a medical professional should trust an algorithm one hundred percent.”
“The proposed model performed better than other known models in terms of precision, sensitivity and specificity,” the authors write, adding that their system is “more reliable and accurate” than existing diagnostic tools for future risk of Alzheimer’s.
Referencia: Analysis of Features of Alzheimer’s Disease: Detection of Early Stage from Functional Brain Changes in Magnetic Resonance Images Using a Finetuned ResNet18 Network by Department of Multimedia Engineering, Kaunas University of Technology, 44249 Kaunas, LithuaniaDepartment of Applied Informatics, Vytautas Magnus University, 44248 Kaunas, Lithuania Academic Editor: Markos G. Tsipouras
Diagnostics 2021, 11(6), 1071; https://doi.org/10.3390/diagnostics11061071
