Researchers recently created an AI model that could accurately diagnose Alzheimer's Disease with over 99 per cent accuracy. Deep learning techniques allowed them to train their model using functional MRI pictures from 138 subjects.
OpenAI's large languages model GPT-3 was used in this study to discover signs of Alzheimer’s Disease in speech. They were able differentiate between transcripts of speech taken from healthy volunteers or Alzheimer's patients using this method.
Background
AI is capable of making predictions from data, and then applying these predictions to real-world circumstances. These models are useful in many industries including image analysis, medical diagnosis, resource optimization, and image analysis.Early detection can offer patients treatment options and help slow down the progression of Alzheimer’s disease. Sometimes, brain imaging or comprehensive cognitive evaluations are required to diagnose this debilitating disorder.
Researchers are currently developing an AI model for scanning brain images for signs and symptoms of Alzheimer's. The AI model could detect it early, giving patients the opportunity to make lifestyle changes that might delay or prevent its onset.
The research team used fMRI images gathered from 138 subjects at varying ages, cognitive statuses to train their algorithm. When it was compared against the work by experts neurologists to assess its performance, the model identified features associated Alzheimer's Disease and mild cognitive Impairment. They were surprised at the AI model's performance, which was slightly higher than that of experts neurologists.
Methods
Many people with mild cognitive impairment are at higher risk of developing Alzheimer’s disease later. This could be due either to genetics or lifestyle choices as well as exposure to specific environmental pollutants, such tobacco and alcohol.Recent research has revealed that early stages of brain degeneration can be detected by AI models. They are subtle enough for even the most skilled specialists to miss. AI models are able to detect these signs in seconds. This may allow for quicker diagnosis and better treatment options. The researchers hope this will lead to better treatment options and earlier diagnoses.
The researchers developed their model by using supervised computer learning. This uses subject matter experts to review and classify new data points. Hualou Liang of Drexel University Philadelphia says supervised models often perform well because they are trained with existing data.
These are the results
Researchers used brain MRIs to build an AI model that could detect Alzheimer's better than a popular natural word processing program. Their work was published in PLOS ONE journal.Researchers scanned 11,103 brain MRIs for Alzheimer's patients from 2,348 patients. They also scanned 26,892 scans for patients without the disease. They also tested it with MRIs from other systems, including MGH and Brigham and Women?s Hospital.
Comparing it to another AI system with similar data, the AI was 80% more accurate in identifying Alzheimer's disease early on. Additionally, it was able to detect MRIs on scanners different than those in the training dataset - a crucial finding because Alzheimer's symptoms can sometimes present without apparent symptoms.
Conclusions
AI is increasingly being used to streamline healthcare tasks including medical diagnosis, image analyses and resource optimization. AI that can explain healthcare decision-making, called Explainable AI, allows for standardized and augmented treatment plans. This helps to reduce healthcare expenses while speeding up treatment decisions.AI models can learn to show representative data traits using many methods. Starting with simple features, such as colors and lines before moving onto more complex structural elements. These characteristics are given different levels of importance depending on which model they are.
This study employs a neural network to analyze PET scans of Alzheimer's patient and healthy controls. Combining the data from both groups, this algorithm can predict if an individual will eventually get Alzheimer's disease. It can also predict if they will have mild cognitive impairment (MCI).