News

Artificial intelligence could provide a definitive diagnosis of schizophrenia, a condition that currently cannot be diagnosed through tests. At Lublin University of Technology, researchers are working on a method that will allow doctors to diagnose the condition... by looking into the patients' eyes.

Schizophrenia is a mental disorder classified among psychoses. The World Health Organization estimates that approximately 21 million people worldwide are affected by this disease. In Poland, the number could be around one million. The exact cause of the disorder has not yet been determined, and there are no definitive diagnostic methods. The diagnosis is primarily based on conversations between the doctor, the patient, and their family, as well as observations. There are no laboratory tests to confirm or rule out schizophrenia.

However, there is hope that this could change. A team of scientists led by Dr. Paweł Karczmarek, a professor at Lublin University of Technology and head of the Department of Computational Intelligence, in collaboration with researchers from the Medical University of Lublin, the Catholic University of Lublin, and the University of Rochester (USA), is investigating whether it is possible to diagnose the disease through retinal scanning.

"The retina is closely connected to the central nervous system, and one could say that it is a simplified model of what happens in the brain," says Dr. Paweł Karczmarek.

An optical tomography device, commonly used in cardiology, dermatology, and ophthalmology, is employed for scanning. This device allows for the examination of the retina's structure layer by layer, and its images have the same high resolution as histological images.

"This type of examination can be completed in a much shorter time (around one minute) compared to an MRI scan, which reduces both the time and the cost of the tests," adds Dr. Paweł Karczmarek.

"Data from the retina can serve as a good basis for classifying patients as either healthy or sick," says the researcher from Lublin University of Technology. "Previous studies confirm the existence of thinner retinal layers in people with schizophrenia. However, these could also be caused by other diseases, which is why such an exceptionally precise analysis is needed."

Dr. Paweł Karczmarek's team conducted an analysis of 59 schizophrenia patients and 61 healthy individuals. The results were published in December last year in Scientific Reports. The authors from Lublin University of Technology, besides Dr. Paweł Karczmarek, include Dr. Kamil Jonak, Professor at the university, Dr. Małgorzata Plechawska-Wójcik, and Dr. Adam Kiersztyn.

"We interpreted the optical tomography images using advanced artificial intelligence algorithms. We applied various analysis models that increased the precision of detecting differences between healthy individuals and schizophrenia patients," says the scientist. "It turned out that combining retinal imaging with deep learning techniques based on neural networks can significantly improve the diagnosis of schizophrenia."

This important discovery proves that thinner retinal layers can be biomarkers for schizophrenia. This will allow for faster and more accurate diagnosis of the disease, leading to earlier treatment and an improvement in the quality of life for patients.

"It is worth emphasizing that new classification algorithms were used in the work, which aggregate results obtained from multiple classifiers, such as neural networks. The techniques applied, inspired by numerical quadrature extensions of the Choquet integral, have roots in fuzzy logic. They can increase the accuracy (effectiveness) of classification by several percentage points compared to the best individual classifiers. This is particularly important considering that the datasets used to train the models are still relatively small and contain few samples, mainly due to the difficulty in obtaining them, especially for relatively rare diseases," adds the researcher from Lublin.

"Our ambition is to operate at the intersection of basic and applied research. The former is achieved by creating new machine learning models. However, their true validation occurs when applied to real-world problems, such as research on schizophrenia."

The collaboration between the researchers from Lublin University of Technology and those from the Medical University of Lublin and the University of Rochester continues. A grant proposal for joint research has been submitted in the USA. Further studies are being prepared on retinal analysis using machine learning and artificial intelligence.

"Let us emphasize that the researchers from Lublin University of Technology are also actively involved in applied research in diagnostics and monitoring other diseases, such as the progression of retinal pigment degeneration. One example is the ongoing grant 'Lublin Digital Union – Using Digital Solutions and Artificial Intelligence in Medicine – Research Project.'"

"Modern imaging technologies supported by artificial intelligence could revolutionize diagnostics not only in ophthalmology but also in neuropsychiatry, enabling more precise monitoring of diseases such as schizophrenia," concludes Dr. Paweł Karczmarek, Professor at the university.