Researchers have discovered that artificial intelligence (AI) can now detect Alzheimer’s using brain imaging techniques. Researchers created a unique algorithm that gained positive results during a study to spot the disease about six years before the actual diagnosis.
The study published in Radiology states the studies results. The study is a forward step in the treatment of incurable Alzheimer's disease. This research will allow doctors to intervene in the development of disease earlier.
Early diagnoses of Alzheimer's has many challenges. Many symptoms of the disease remain undiscovered in early diagnosis.
Positron emission tomography (PET) is a major tool to scan the Alzheimer's disease before it gets severe. It measures the levels of specific molecules such as glucose in the brain. Level of glucose is observed as it is the main cause of fuel for brain cells which keeps them active.
Alzheimer's disease occurs when the brain cells die which happens as a result of lowered glucose level in brain tissues. However, Alzheimer's disease develops too slow and is difficult to spot easily.
To tackle the issue, researchers used a deep machine learning algorithm on PET scans data provided by Alzheimer’s Disease Neuroimaging Initiative (ADNI). The PET scans were from patients who had been diagnosed with either Alzheimer’s disease, mild cognitive impairment, or those who do not have the disease.
The algorithm was tested on 2 novel datasets to evaluate the performance. The algorithm performed well to detect Alzheimer’s disease in 92 percent patients of the first test set and 98 percent of the second test set. Amazingly, the algorithm was able to predict on an average 75.8 months before the patient had received their final diagnosis.
The research team now wants to deploy the algorithm on a larger dataset to ensure optimal performance on a wide variety of patients.
Dr. Jae Ho Sohn and Dr. Benjamin Franc from UCSF are co-authors of this study.