A Revolutionary Development in Screening for Cervical Cancer

A new solution jointly developed by the University of Debrecen, Delta Services and Sightspot Network will make the diagnosis of cervical cancer faster and more reliable, allowing doctors to start treatment at an early stage. The new solution is an image-based diagnostic system, which is able to identify cancer cells based on deep learning technology.

As we learned at the event held at the Faculty of Informatics on Friday to introduce the diagnostic system, the new solution developed by the consortium may revolutionize cervical screening. Research into artificial intelligence plays a central role in the scientific activities of the faculty, and the development of this new diagnostic system is perfectly in line with this strategy. Currently, patients have to wait for 2-4 weeks for the screening results. In the new system, the AI-supported software applications used in the screening of samples allow quicker evaluation of the results.

András Hajdu, dean of the Faculty of Informatics of UD emphasized that the prototype was ready and could be deployed in practice.

- Our goal is to see the new diagnostic system introduced in clinical practice. To this end, we still need international tests and evaluation. Due to artificial intelligence, results can be obtained quickly and cost-efficiently, while early diagnosis allows surgeons to perform significantly more life-saving interventions. During the project great emphasis was placed on manual work, i.e. supervised machine learning. The database that will be created during the operation of the system may be used in other areas as well, in addition to cytology – added dean Hajdu.

László Vidra, managing director of Delta Systems Kft. highlighted that the solution would significantly reduce the load on the healthcare system, because it would be able to identify abnormal cells in the digitalised smear without human intervention.

- In terms of competitiveness, digitalisation is crucial. A further benefit of the solution is that it may reduce the effects of workforce shortage. We may function as an accelerator in a high-priority area, as the development of digitalisation plays an important role in the national economy as well. In this respect, we used deep learning technology with the help of artificial intelligence. The cooperation between the university, the industry and the system integrator that will help us introduce the system to the market has been exemplary, which also indicates the emergence of a progressive ecosystem – pointed out Mr. Vidra.

Cytopathologist Ilona Kovács, head of the Department of Pathology of the Clinical Centre of UD noted that the new software application would also be able to perform quality control regarding the identification of abnormal cells.

- A team of specialists was involved in the digitalisation process, the evaluation of images, and the selection of samples in the teaching phase. In order to support machine learning and testing, they processed more than 264,000 cell images from 10,000 smears based on data received from the IT working group. Providing that the current strategy in cervical screening changes, this new solution may meet our expectations – explained department head Kovács.

According to Balázs Harangi, associate professor of the Faculty of Informatics of UD, the accuracy of the system  is currently 93% at cellular level.

- The role of cytologists in collecting samples and digitalisation was essential. Our primary task was to extract cells from the samples. For this purpose, we used state-of-the-art image-processing methods. Images with 100,000x200,000 pixel resolution were made, 7GB in size each. We would like to perform a test that proves that the system works and can be used in practice with confidence – added associate professor Harangi.

The project was carried out within the framework of the call entitled “R&D Competitiveness and Excellence Cooperations”, which was announced as part of the Széchenyi 2020 programme.

Press Centre - BZ