The diagnosis of Alzheimer’s disease (AD) is complex and inaccurate. Scientists in Lithuania used artificial intelligence combined with Deep Learning to modify an existing scanning algorithm to identify and classify AD with nearly 100% accuracy.
https://www.ibm.com/cloud/learn/deep-learning. Deep learning is a term applied to learning applications that mimic human thought. It is a subset of machine learning – a rudimentary neural network that simulates the behavior of the human brain. Each layer of machine learning increases the accuracy of prediction.
Deep learning processes unstructured data and adjusts to self-improve predictability through gradient descent and backpropagation. Applying deep learning with machine learning and various artificial intelligence applications of the human brain, scientists discovered a way to predict mild cognitive impairment (MCI) that typically precedes AD. It captures the subtle signs before AD is manifested.
Alzheimer’s Disease Analysis
https://www.mdpi.com/2075-4418/11/6/1071/htm. Functional MRI (fMRI) scans are difficult to interpret. Applying deep learning, Lithuanian scientists at the Kaunas University of Technology were able to develop a computer algorithm that accurately assesses the different stages of MCI. As we age, it is common for mild cognitive impairment to evolve, but it is different than MCI associated as a precursor to AD. This algorithm differentiates the difference between normal aging and early signs of AD.
Deep Residual Learning
https://www.kaggle.com/pytorch/resnet18. ResNet 18 is a well-known computer algorithm. This application was finely tuned to recognize MCI in normal aging and patients susceptible to early-onset AD. One-hundred thirty-eight people were scanned. This group of participants were healthy or had progressive signs of AD.
The fMRI scans were studied, and the scientists determined that the algorithm was 99.99% accurate in predicting scans of the various participants. Thus, the accuracy was the apparent breakthrough from this study.
This was a small study (nearly 140 people) that approximated the real world. Nevertheless, the ability of the algorithm to accurately predict the development of AD is a positive sign that larger groups of people can be diagnosed more quickly with a higher degree of accuracy, even if it is less than 99%.
I expect that future clinical studies of larger populations will provide more accurate updates to AD prediction.
Live Longer & Enjoy Life! – Red O’Laughlin – RedOLaughlin.com