On Oct. 1, 2020, the University of Maine hosted a webinar over Zoom to talk about the practical applications of artificial intelligence in today’s society. The webinar focused on artificial intelligence applications in the fields of industry, government, healthcare, and the environment. The seminar was the first in the UMaine AI lunch and learn series. Speaking during the program was Associate Dean of Engineering Mohamad Musavi, President and Co-founder of MedRhythms Inc. Owen McCarthy and Software Architect at Activas Diagnostics Somayeh Khosroazad.
Musavi opened up as the first speaker, and mainly spoke on the background of artificial intelligence and how it works. The main purpose of this technology is to act similarly to that of humans, to properly demonstrate how to learn tasks. The key is using artificial neural networks, which try to mimic similar functions of the human brain. The human brain is a network of 80-100 billion neurons, and computers try and replicate that with billions of transistors that can perform trillions of micro-operations in seconds. This wiring is the way to craft artificial neurons that mimic processing functions and adaptability. By imputing pieces of data, the program can learn over time, and acquires knowledge to better accomplish tasks. According to Musavi’s presentation, four primary functions of artificial intelligence in industry are classification, clustering, function and prediction. This can range from chromosome classification in the medical field to predicting brightness and opacity. Several industries can make use of neural networks to accomplish tasks in an automated and effective fashion.
“Whenever there is data, neural networks can provide insight into an event or process,” Musavi stated.
McCarthy took a deeper dive into the benefits of AI in regulated health care. While there are fears that AI has the potential to make physicians obsolete, McCarthy made the argument for why that isn’t the case. AI can serve the purpose of autonomous actions of treatment, dominantly in radiology as of right now. Programs are able to learn from imputed data and better diagnose cases for patients. AI has also been helping the United States Food and Drug Administration (FDA) as well.
“Products that are FDA approved are ones that have to do with safety and efficacy related to treating, diagnosing and curing a disease,” McCarthy said. She also spoke on how AI is working in the field of dermatology, being able to detect malignancies in skin abnormalities and deduce whether or not they are detrimental to one’s health.
Khosroazad focused her presentation on AI in healthcare as well. AI works well to categorize problems and better craft a plan to diagnose issues.
“We categorize AI problems, from a specific view to two groups, some specific problems in which a huge amount of data is collected and AI, supervised or unsupervised, tries to classify the data to multiple groups by finding by finding one or more thresholds or some limits, based on all statistical and problematic logics that are predefined for it. The AI then directly works with that data,” Khosroazad said.
The AI can then differentiate the data into either an expected outcome, and try to diagnose it, or can recognize abnormalities in the data, and bring it to the attention of physicians. AI is also being used in asleep signal analysis for early detection of Alzheimer’s disease and dementia. The AI records sleep signals and is able to recognize what those signals represent for deteriorating brain activity with the help of physicians.
Artificial intelligence has come a long way in the past decades, and will continue to take steps forward to better aid society in government, healthcare and the environment. The webinar will be made available to everyone on UMaine’s AI homepage.