MedUni Vienna: Prostate cancer - New AI model can prevent unnecessary prostate removals
Surgical removal of the prostate to treat prostate cancer is currently determined on the basis of tissue sample analysis (Gleason grading). As this method is invasive and often unreliable, scientists around the world are looking for alternatives. A research team at MedUni Vienna has now developed a new method that can be used to identify those patients for whom surgical treatment is the best option. As a result, avoiding unnecessary surgery in patients with a lower risk of tumour spread is now possible. The study was recently published in the journal "Theranostics".
The research team led by Lukas Kenner (MedUni Vienna's Department of Pathology), Jing Ning and Clemens Spielvogel (MedUni Vienna's Department of Biomedical Imaging and Image-Guided Therapy) aimed to develop a new machine learning model for more precise tumor assessment. "We combined multi-omics technology with artificial intelligence applications," says Lukas Kenner, head of the study, emphasizing the unique approach. Multiomics is a method in medical research in which various "omics" data sources such as genetic information (genomics), imaging features (radiomics) and results from pathological examinations (pathomics) are integrated. This large amount of data fed into an AI model comes from 146 patients who underwent surgical removal of the prostate (radical prostatectomy) between May 2014 and April 2020.
Identifying high-risk patients
By combining multi-omics with machine learning, an AI model has been created that the researchers expect to achieve a great deal: "In our study, we were able to assess the changes in the prostate much more accurately and reliably than with the conventional biopsy method and the Gleason grading," reports Lukas Kenner. This makes it much easier to identify high-risk patients who would benefit from a radical prostatectomy and avoid unnecessary interventions in patients with a low risk of tumor spread.
Radical prostatectomy is an important element of prostate cancer treatment, but leads to urinary incontinence in around 30 percent of patients and erectile dysfunction in around 90 percent. Whether the procedure is indicated is decided based on the Gleason grading, a system for assessing the aggressiveness of prostate cancer. This score is usually determined by analyzing small tissue samples taken during a biopsy. However, the comparison of these values with the results of a complete tissue examination after removal of the prostate often shows discrepancies. "The results of our study underline the potential of machine learning and multi-omics to improve the diagnosis and personalized therapy of prostate cancer," says Lukas Kenner. Further studies to test the method are planned to advance its clinical application.
Publication: Theranostics
A novel assessment of whole-mount Gleason grading in prostate cancer to identify candidates for radical prostatectomy: a machine learning-based multiomics study;
Jing Ning, Clemens P. Spielvogel, David Haberl, Karolina Tractova1, Stefan Stoiber, Sazan Rasul, Vojtech Bystry, Gabriel Wasinger, Pascal Baltzer, Elisabeth Gurnhofer, Gerald Timelthaler, Michaela Schlederer, Laszlo Papp, Helga Schachner, Thomas Helbich, Markus Hartenbach, Bernhard Grubmüller, Shahrokh F Shariat , Marcus Hacker, Alexander Haug, Lukas Kenner
doi: 10.7150/thno.96921
https://www.thno.org/v14p4570.htm
Contact
Mag. Johannes Angerer
Medizinische Universität Wien
Leiter Kommunikation und Öffentlichkeitsarbeit
Telefon: 01/40160-11501
E-Mail: pr@meduniwien.ac.at
Website: https://www.meduniwien.ac.at/pr