Artificial intelligence and blood-based biomarkers enable the objective diagnosis of early-onset psychosis in children—according to a team of researchers from the University of Zielona Góra. In cooperation with partners from the University of Oxford, Heidelberg University, and the Institute of Psychiatry and Neurology in Warsaw, the researchers conducted innovative studies whose results demonstrate that early psychosis in children and adolescents can be diagnosed objectively. This has been made possible through specially trained machine-learning algorithms that analyse blood biomarkers in combination with neuropsychometric tests.
“Psychiatric diagnostics in children constitute an enormous challenge. Clinical symptoms often overlap with difficulties typical of adolescence, and distinguishing pathological states from normal developmental experiences can be extremely difficult. Meanwhile, time is of crucial importance: the longer psychosis remains untreated, the poorer the prognosis,” explains Przemysław Zakowicz, MD, PhD, from the Department of Neuroengineering and Space Medicine, Institute of Medical Sciences, Collegium Medicum of the University of Zielona Góra.
Until now, psychiatry has lacked objective diagnostic tools comparable to those used, for example, in cardiology, where ECG examinations or blood tests provide unequivocal confirmation of a diagnosis. In such cases, symptoms combined with diagnostic test results offer physicians a comprehensive clinical picture, enabling them to make optimal decisions for the patient and, in some instances, save lives. By contrast, the diagnosis of psychotic disorders has largely relied on clinical interviews and observation, which do not always allow for a definitive diagnosis.
The KEPLER Study research programme, led by the scientific team of the Department of Neuroengineering and Space Medicine at the University of Zielona Góra, sheds new light on this long-standing problem. The international medical research team—comprising scientists from the University of Zielona Góra, the Institute of Psychiatry and Neurology in Warsaw, the University of Oxford, and Heidelberg University—has completed the first stage of the study. Using machine-learning methods, the researchers developed an algorithm capable of identifying psychotic disorders on the basis of blood test results combined with objective assessments of executive brain functions. “This represents a revolution in the diagnosis of mental disorders and a first step towards the broader implementation of such diagnostic methods,” the researchers unanimously conclude.
The study was conducted on a group of 45 patients with early-onset psychosis and a control group of 34 healthy children. Participants underwent a comprehensive assessment, including the analysis of serum protein biomarkers. The algorithm successfully identified and utilised these biomarkers, together with neuropsychometric test results, to establish an accurate diagnosis. This discovery is consistent with previous research demonstrating that impaired synaptic plasticity may underlie the pathophysiology of schizophrenia.
The ability to objectively diagnose psychotic disorders in children and adolescents is crucial for two primary reasons. Clinicians require diagnostic certainty when identifying conditions that significantly affect the future course of a young person’s life. At the same time, researchers developing new therapies need reliable tools to identify disorders in patients and to measure therapeutic outcomes. “With such methods, the risk of diagnostic error is reduced to a minimum, and we gain the opportunity to detect the disease at a very early stage, thereby avoiding complications. The biomarkers we analysed are related to proteins involved in synaptic plasticity, that is, the brain’s capacity for regeneration. It appears that, in psychosis, the brain may inadequately cope with the reconstruction of neural connections, resulting in difficulties with decision-making and orientation in the often challenging reality surrounding the patient,” summarises Dr Zakowicz, adding: “Thanks to international collaboration, we are increasingly closer to solving one of the greatest challenges in modern medicine.”
The results of the study are presented in the article “Machine learning helps predict early onset psychosis with serum protein biomarkers, neuropsychometry, and clinicodemographic data”, published in one of the world’s most prestigious scientific journals, Nature Scientific Reports.
