November 8, 2011
Computer Algorithm Can Predict Outcomes Of Psychotic Illnesses
Computer algorithms that analyze magnetic resonance imaging (MRI) brain scans can help predict the duration and severity of a psychotic patient's illness, something that could help doctors make better treatment decisions, according to a new study published Monday in the journal Psychological Medicine.
Psychosis is a condition that affects people's minds, altering the way they think, feel and behave. It can be accompanied by hallucinations and delusions.
Many patients recover from psychosis with minimal symptoms, but for others, the psychosis can be persistent and can affect their ability to function well and lead a normal life.
Psychiatrists are currently without any clear method of assessing a person's risk of future episodes, or predicting how the disease will progress, which is important in making appropriate treatment decision.
While algorithms that quantify the risk of further episodes of disease are common in areas of medicine such as cardiovascular medicine and oncology, no accurate tests are available to psychiatrists.
Researchers have previously used MRI brains scans to predict outcomes in psychosis based on analyzing specific brain regions. However, changes in the brain associated with psychosis are often subtle and difficult to detect, and these approaches were of limited benefit for clinical practice.
However, in the current study, researchers from King's College London's Institute of Psychiatry and University College London report the successful use of computer algorithms to analyze MRI scans and predict a patient's outcome.
Lead researchers Dr. Paola Dazzan and Dr. Janaina Mourao-Miranda and colleagues worked with 100 patients, taking MRI brain scans when they presented to clinical services with a first psychotic episode. The researchers also scanned the brains of a control group of 91 healthy individuals.
The patients were followed up on around six years later and classified as having developed a continuous, episodic or intermediate illness course, depending on whether their symptoms remitted or not during this time.
The researchers then analyzed scans from twenty-eight subjects with a continuous course of illness, the same number from patients with an episodic course and again, the same number from healthy controls.
They used these scans as data to 'train' a software developed by a group led by Dr. Mourao-Miranda based on pattern recognition (a statistical approach that uses data from the whole brain rather than from a specific region) and to distinguish between the different severities of the illness.
The algorithm, applied to the scans collected at the first episode of psychosis, was able to differentiate between patients who then went on to develop continuous psychosis and those who went on to develop a more benign, episodic psychosis in seven out of ten cases.
"Although we have some way to go to improve the accuracy of these tests and validate the results on independent large samples, we have shown that in principle it should be possible to use brain scans to identify at the first episode of illness both patients who are likely to go on to have a continuous psychotic illness and those who will develop a less severe form of the illness," said Dr. Mourao-Miranda.
"This suggests that even by the time that they have their first episode of psychosis, significant changes have already occurred to their brains."
Dr. Dazzan said the study´s results offered hope that computers could someday help doctors determine the best course of treatment for their patients.
"This is the first step towards being able to use brain imaging to provide tangible benefit to patients affected by psychosis,"
"This could in future offer a fast and reliable way of predicting the outcome for an individual patient allowing us to optimize treatments for those most in need, while avoiding long-term exposure to antipsychotic medications in those with very mild forms.”
"Structural MRI scans can be obtained in as little as ten minutes and so this technique could be incorporated into routine clinical investigations. The information this provides could help inform the treatment options available to each patient and help us better manage their illness."
On the Net: