Machine learning and AI are finding ever more ways to integrate into daily life. The concept is used for a broad range of applications, from finance to space exploration to healthcare. In the latter field, one novel approach is attempting to use AI to assist with identifying mental illness.
Indian researchers may have developed a means by which to determine schizophrenia in a patient with a high degree of accuracy, by using AI to analyse trends in brain activity.
Psychiatry, and specifically the diagnosis of mental illness, is a complex procedure. The brain is a far more complex organ than, for example, the heart. The average human brain consists of 100 billion neurons. Their interaction determines a person’s thought process and personality, with mental illnesses being caused by numerous factors within these interactions.
Typically a psychologist will perform a diagnosis of a patient based on their behaviours. Due to this, diagnosis is highly personal and not always entirely accurate. Schizophrenia, for example, has a range of overlapping symptoms with other psychiatric conditions, leaving it difficult to diagnose. AI, however, may hold the key to streamlining the process of diagnosing the disorder, and even increasing accuracy.
Researchers at the National Institute of Mental Health and Neurosciences (NIMHANS) used functional magnetic resonance imaging (fMRI) to map the brain activity of 93 healthy individuals and 81 previously diagnosed schizophrenia patients. The fMRI analysed frequency of brain waves, correlation between brain activity of closely placed regions, and connectivity between different brain regions — all factors shown in previous studies to differ in schizophrenia patients.
The analysis uncovered 84 points of data by which to compare the fMRI scans of healthy individuals and schizophrenia patients. Using these points the team created an AI model which can analyse fMRI scans and determine schizophrenia within a patient with an accuracy of 87 percent. The model has been named “EMPaSchiz” or ‘Ensemble algorithm with Multiple Parcellations for Schizophrenia prediction’.
They noted that the model needs more work to establish a more user-friendly variant that could be used in a hospital setting. However, the model does represent a means for a hospital to diagnose schizophrenia in a patient without first having to go through lengthy sessions with a psychiatrist.
In India, access to trained psychiatric personnel is limited. Mental healthcare is available to only a small minority and programmes such as this may at the very least begin to fill the cracks. Issues may arise when providing treatment and medical intervention may be possible if supplies are available, but without addressing the shortage of psychiatric staff treatments such as therapy are all but beyond reach for the vast majority of mental health patients in India.