Based on artificial intelligence and automatic microscopy, the serological testing model developed at the Szeged Biological Research Centre is highly accurate, fast and cost-effective.
The Biomag Research Group of the Szeged Biological Research Centre (SZBK) at the Eötvös Loránd Research Network, led by Péter Horváth and working in collaboration with the Institute of Microbiology and Immunology of the University of Szeged, two research groups of the University of Helsinki and a spinoff company, Single-Cell Technologies, has developed an entirely new serological test for SARS-CoV2 – which has never been used anywhere, including Hungary.
The method, based on artificial intelligence and automated microscopy, identifies those who have already recovered with high accuracy and provides reliable feedback on the level of protection. It may also be suitable for the identification of newly infected people. The method has been validated for over 1,000 cases and accuracy has been measured at close to 100%. The method has very high throughput and is already suitable for 5-10,000 tests per day, so in the case of a second or further wave of infection, it could be successfully used for mass testing, i.e. to identify those who have already contracted the disease as well as newly infected individuals.
In addition to a turnaround time of 6 to 8 hours, repeatability and cost-effectiveness, another important feature of the method is high sensitivity, which allows the detection of infection even with mild immunity. The method does not show false positive results in healthy samples, and another great advantage is that it can be easily adapted to the proteins of any virus, so it can be quickly applied to epidemics caused by other viruses.
The theoretical basis of the model is the detection of antibodies – immunoglobulins – produced by the human body as these can already be detected in the blood a few days after infection and then for months afterwards. During the test, the blood sample is added to specially modified cells and then the cells are studied with a high-sensitivity, high-throughput automated microscope. Finally, the presence or absence of the antibody in each cell is determined by the artificial intelligence-based method.
The analysis of images captured with an automated microscope using artificial intelligence is one of the main profiles of Péter Horváth’s research group. For this, so-called “deep learning” algorithms are used, which belong to a new branch of artificial intelligence. A similar algorithm was also used in their new testing model developed for SARS-CoV2 to automatically and reliably evaluate images. Interestingly, these deep learning algorithms are also used for many functions of self-driving cars – e.g. zebra crossing pedestrian detection and overtaking – and for facial recognition in social media.