The possibilities of artificial intelligence in detecting prostate cancer according to magnetic resonance imaging: intermediate indicators of the effectiveness of the national second opinion system

Introduction: The use of deep learning in the detection of prostate cancer (PCa) based on multiparametric MRI (mpMRI) data is one of the most controversial areas of development of this technology in medicine. Despite high international interest in such systems, the number of thematic studies in the Russian Federation remains limited.

Objective: To describe the process of creating and validating a domestic medical decision support system (MDSS) designed to diagnose PCa based on mpMRI results using AI.

Materials and methods: Data on patients with a confirmed diagnosis who underwent radical prostatectomy were collected both retrospectively and prospectively. The patient sample was formed between January 10, 2023 and May 1, 2025. T2- weighted images (T2-WI) and measured diffusion coefficient (MDC) data were used as working sequences. Data annotation was performed manually by three specialists with more than 10 years of experience in independently performing and analyzing MR data of patients with suspected PCa. During the study, four different architectures were tested to evaluate their effectiveness and determine the most suitable one for the tasks at hand. The following metrics were used to measure accuracy: sensitivity, specificity, accuracy, area under the ROC curve (AUC), and Dice score (DSC).

Results: In accordance with the inclusion criteria, 441 studies performed on 19 different MRI scanners were collected. Validation of the most productive neural network architecture revealed the following indicators: accuracy — 73%, sensitivity — 71%, specificity — 71%, AUC — 0.70, DSC — 0.70.

Discussion: The work presented by us is aimed at familiarizing specialists with the intermediate results of the development of the first national MDSS, implemented according to a strategy of avoiding the shortcomings present in the works of other teams.

Conclusion: The results of testing the developed AI-based program for the detection and reconstruction of PCa on MRI images showed promising prospects for the domestic medical decision support system. To confirm the clinical applicability and effectiveness of the developed program, further multicenter studies are needed to assess its potential and feasibility for implementation in real clinical practice.

1. Kaprin A.D., Alekseev B.I., Matveev V.B. et al. Prostate cancer. Journal of Modern Oncology, 2021, Vol. 23, Nо. 2, pp. 211–247 (In Russ.). doi: 10.26442/18151434.2021.2.200959.

2. Rubtsova N.A., Мishchenkо А.V., Danilov V.V. et al. PI-RADS v2.1: moving towards clarity (comments on the updated version). Onkourologiya=Cancer Urology, 2020, Vol. 16, Nо. 2, pp. 15–28 (In Russ.). doi: 10.17650/1726-9776-2020-16-2-15-28.

3. Lee C.H., Vellayappan B., Tan C.H. Comparison of diagnostic performance and inter-reader agreement between PI-RADS v2.1 and PI-RADS v2: systematic review and meta-analysis // Br. J. Radiol. 2022. Vol. 95, Nо. 1131. P. 20210509. doi: 10.1259/bjr.20210509.

4. Jiang F., Jiang Y., Zhi H. et al. Artificial intelligence in healthcare: past, present and future // Stroke Vasc. Neurol. 2017. Vol. 2, Nо. 4. P. 230–243. doi: 10.1136/svn-2017-000101.

5. Talyshinskii A., Hameed B.M.Z., Ravinder P.P. et al. Catalyzing Precision Medicine: Artificial Intelligence Advancements in Prostate Cancer Diagnosis and Management // Cancers. 2024. Vol. 16, Nо. 10. P. 1809. doi: 10.3390/cancers16101809.

6. Reva S.A., Shaderkin I.A., Zyatchin I.V. et al. Artificial intelligence in cancer urology. Experimental and Clinical Urology, 2021, Vol. 14, Nо. 2, pp. 46– 51 (In Russ.). doi: 10.29188/2222-8543-2021-14-2-46-51.

7. Popkov V.M., Shatylko T.V., Korolev A.Ju. et al. Optimizing PSA Screening with Artificial Intelligence. Medical Bulletin of Bashkortostan, 2015, Vol. 10, Nо. 3, pp. 232–235 (In Russ.).

8. Talyshinskii A.E., Kamyshanskaya I.G., Mischenko A.V. et al. Application of artificial intelligence in the detection and stratification of prostate cancer: Literature review. Bulletin of St. Petersburg University. Medicine, 2023, Vol. 18, Nо. 2, pp. 150–166 (In Russ.). doi: 10.21638/spbu11.2023.204.

9. Alowais S.A., Alghamdi S.S., Alsuhebany N. et al. Revolutionizing healthcare: the role of artificial intelligence in clinical practice // BMC Med. Educ. 2023. Vol. 23, Nо 1. P. 689. doi: 10.1186/s12909-023-04698-z.

10. Kumar Y., Koul A., Singla R. et al. Artificial intelligence in disease diagnosis: a systematic literature review, synthesizing framework and future research agenda // J. Ambient Intell. Humaniz. Comput. 2023. Vol. 14, Nо. 7. P. 8459–8486. doi: 10.1007/s12652-021-03612-z.

11. Naik N., Talyshinskii A., Shetty D.K. et al. Smart Diagnosis of Urinary Tract Infections: is Artificial Intelligence the Fast-Lane Solution? // Curr. Urol. Rep. 2024. Vol. 25, Nо. 1. P. 37–47. doi: 10.1007/s11934-023-01192-3.

12. Mukherjee J., Sharma R., Dutta P. et al. Artificial intelligence in healthcare: a mastery // Biotechnol. Genet. Engl. Rev. 2024. Vol. 40, Nо. 3. P. 1659–1708. doi: 10.1080/02648725.2023.2196476.

13. Yu R., Ke-Wen J., Bao J. et al. PI-RADSAI: introducing a new human-in-the-loop AI model for prostate cancer diagnosis based on MRI // Br. J. Cancer. 2023. Vol. 128, Nо. 6. P. 1019–1029. doi: 0.1038/s41416-022-02137-2.

14. Castillo T.J.M., Starmans M.P.A., Arif M. et al. A Multi-Center, Multi-Vendor Study to Evaluate the Generalizability of a Radiomics Model for Classifying Prostate cancer: High Grade vs. Low Grade // Diagnostics. 2021. Vol. 11, Nо. 2. P. 369. doi: 10.3390/diagnostics11020369.

15. Hosseinzadeh M., Saha A., Brand P. et al. Deep learning-assisted prostate cancer detection on bi-parametric MRI: minimum training data size requirements and effect of prior knowledge // Eur. Radiol. 2022. Vol. 32, Nо. 4. P. 2224–2234. doi: 10.1007/s00330-021-08320-y.

16. Rajagopal A., Redekop E., Kemisetti A. et al. Federated Learning with Research Prototypes: Application to Multi-Center MRI-based Detection of Prostate Cancer with Diverse Histopathology // Acad. Radiol. 2023. Vol. 30, Nо. 4. P. 644–657. doi: 10.1016/j.acra.2023.02.012.

17. Alqahtani S., Wei C., Zhang Y. et al. Prediction of prostate cancer Gleason score upgrading from biopsy to radical prostatectomy using pre-biopsy multiparametric MRI PIRADS scoring system // Sci. Rep. 2020. Vol. 10, Nо. 1. P. 7722. doi: 10.1038/s41598-020-64693-y.

18. Siddiqui M.M., Rais-Bahrami S., Truong H. et al. Magnetic resonance imaging/ultrasound-fusion biopsy significantly upgrades prostate cancer versus systematic 12-core transrectal ultrasound biopsy // Eur. Urol. 2013. Vol. 64, Nо. 5. P. 713–719. doi: 10.1016/j.eururo.2013.05.059.

19. Ishioka J., Matsuoka Y., Uehara S. et al. Computer-aided diagnosis of prostate cancer on magnetic resonance imaging using a convolutional neural network algorithm // BJU Int. 2018. Vol. 122, Nо. 3. P. 411–417. doi: 10.1111/bju.14397.