IBM researchers are using AI models to study diseases such as Alzheimer’s or the prediction and design of immune responses and the study shows that artificial intelligence can be really helpful for early diagnosis.
If there is one area where innovation and, specifically, technologies such as AI or Machine Learning are playing a leading role, it is undoubtedly the area of healthcare.
Improving treatments, accelerating the development of drugs, or predicting certain diseases are just some of the goals currently being pursued by major technology companies around the world.
IBM wanted to show how its researchers are using certain AI models for the early detection of Alzheimer’s disease; the study of the relationship between the sense of smell; the detection, correction, and use of unbiased data for diagnoses and treatments in dermatology; or the prediction and design of immune responses.
AI can be a key factor for early diagnosis
For example, Guillermo Cecchi, Manager of IBM Thomas J. Watson Research Center’s Computational Psychiatry and Neuroimaging Laboratory, explained how IBM Research and Pfizer have developed a new artificial intelligence (AI) model that uses short, non-invasive, standardized speech tests to help predict the eventual onset of Alzheimer’s disease in cognitively healthy people with an accuracy of 0.7 and an AUC of 0.74 (area under the curve).
These predictions were made against data samples from a group of cognitively healthy people who did or did not eventually develop the disease later in life, allowing the researchers to verify the prediction accuracy of the AI model. This is a significant increase over predictions based on clinical scales (59%) and random choice (50%).
IBM shows how AI can be used for health purposes
Another research explained by Celia Cintas, AI Science researcher at IBM Research Africa in the Nairobi Laboratory (Kenya) aims to end the biases of AI and Machine Learning models in the detection of dermatological diseases.
As Celia Cintas explained, in the research they discovered that the data used to train AI models are mostly made up of images of the Caucasian population, which can reduce the ability of these models to detect skin diseases in other patients and their corresponding negative impact on the quality of care and treatment that people receive.
For example, in populations of African descent, melanoma is commonly diagnosed at late stages. In addition, the paucity of images of the cutaneous manifestations of COVID-19 in Latin and African-descent patients is a problem, as it hinders clinical diagnosis.
