AI and Supercomputers Help Doctors Diagnose And Treat Mental Illness
The mental disease affects about a billion people worldwide. The world’s second most powerful supercomputer and artificial intelligence are being used by clinicians at Cincinnati Children’s Hospital, however, to aid in early detection. It’s possible, they argue, that this may be the deciding factor.
A recent TechFirst podcast interview with Dr John Pestian said that if we can discover this early, “we can treat and relieve about half of the mental disease that passes into adulthood,” Pestian said. Thus, earlier detection and therapy are crucial.
The Summit supercomputer is owned and operated by Oak Ridge National Laboratory in Tennessee; the project is a cooperation between the Cincinnati Children’ hospital, University of Cincinnati, University of Colorado, and Oak Ridge National Laboratory. To achieve exaflop performance (one quintillion floating-point operations per second) and 200 petaflops, IBM built Summit, the world’s first supercomputer.
Why do we need such equipment?
AI models for assessing mental health risk factors are trained on the supercomputer 17,000 times quicker than your PC. It would take a decade to execute Pestian’s prior algorithm for preventing veteran suicides on a typical computer.
To avoid veteran suicide, Pestian is a member of the US military’s Million Veteran Program, which uses natural language processing. The Cincinnati initiative for children’s mental health is conducted by 25 scientists from nine different institutions, and he is one of them. According to the researchers, attempting to run the model on a local computer would take nearly a decade. A typical university cluster machine, which isn’t almost a supercomputer in terms of performance. “The supercomputers allow us to reduce it to five hours, and we need it to train the model,” says Pestian.
Pestian argues that the complexity of mental disease necessitates the employment of high-end technology. It’s important to consider all of the many aspects of one’s life, such as one’s physical surroundings as well as one’s mental state. Developing an AI model that can make sense of and accurately anticipate all of these variables is a difficult task.
Natural language processing is a key part of the difficulty since it must be able to take what children say, comprehend it, and utilise it as input for risk assessment models. It is also crucial to note that to keep current, the project software constantly updates itself by going through Medline and PubMed papers to discover research results that could be useful in upgrading their models and highlighting them for a human user to see.
Researchers argue that a decade ago, this was unimaginable. A $10 million donation from the Cincinnati Children’s Research Foundation has made this feasible. According to the foundation’s estimates, mental, behavioural, or developmental disorders affect 17% of all children in the United States between the ages of two and eight. That works out to roughly five students each class.
For those who have been diagnosed, that’s only the beginning. The project’s main purpose is to render “near real-time” models for the prior detection of children at threat of depression and suicidal thoughts. Unusual and more than a bit unnerving is one of the primary resources. Medical and mental health practitioners can communicate with patients, but they must know what they’re searching for. Also, a tool like this may assist in spotting warning indications of suicide ideation.
According to Dr Pestian, “They’re all created off of this vast collection of… a few thousand suicide notes that I gathered and then developed natural language models off of that.” Notes written by those who took their own lives are also included in this collection. Other researchers based their inquiries on Pestian’s notes, such as “Do you have secrets?” and “Are you angry?” As a result of the data collected, an AI model will predict whether or not a child is more likely to suffer from depression or other mental health issues. Other factors are at play, such as the way individuals voice, the number of pauses they take between phrases, and the looks on their faces while interacting.
Pestian adds that although artificial intelligence is crucial, a person is needed and will be in the loop from the beginning. Human intervention will be introduced to decision assistance soon, adds the author. You don’t have to let the computer decide if you’re going to fall into depression if the computer says, “Oh, it seems like you’ll be headed into a depression.” As a result, we must ensure that we support choices while allowing for human involvement. It doesn’t matter where the at-risk assessment conclusions are made, Pestian points out. Psychiatrists, psychologists, and school counsellors all play a crucial role.