Automated morphometry of the prostate gland by the results of magnetic resonance imaging

Cover Page


Cite item

Full Text

Abstract

BACKGROUND: Within the framework of the experiment on using the innovative technologies in the field of computer vision for analyzing the medical images and on further usage of these technologies in the healthcare system of the City of Moscow, the research was carried out using the equipment based on the artificial intelligence (AI-service) for the purpose of automatization of the morphometry of the prostate gland using the magnetic resonance imaging (MRI), for the issue is topical due to the high incidence of urological diseases among men. Unlike the 11 previous systems, oriented at the retrospective analysis, this solution helps the radiologists in shortening the time of describing the examination results and in increasing their accuracy. AIM: to evaluate the quality and the validity of automatic morphometry of the prostate gland by the MRI results using the technologies of artificial intelligence in the settings of practical healthcare. METHODS: A prospective diagnostic research in accordance with the methodology of reporting results of scientific research involving the STARD 2015 diagnostic tests was conducted during the period from April until October of 2024. A total of 560 MRI results were used and compared to the data from the morphometric AI-service. RESULTS: An evaluation of the accuracy of using the AI-service for the morphometry of the prostate gland was carried out. A total of 7 clinical monitoring procedures were conducted using 560 MRI datasets with the complete conformity reported in 71.6%. The rate of false-negative cases was 3.9%, technical defects were found in 3.8% of the cases. The integral clinical evaluation has achieved the range of 88.0–97.0%, confirming the high diagnostic quality. The predominant errors were the ones related to the contouring of the gland (52%) and incorrect measurements (13%), often related to the prolapsing of the prostate gland apex. CONCLUSION: The automatization of routine measurements greatly contributes to the standardizing the processes of describing the results obtained by radio-diagnostic methods. This aspect is of special importance from the point of view of providing the continuity of medical aid in case of patients presenting to various medical organizations. The artificial intelligence technologies for the automatization of the prostate gland measurements have demonstrated high clinical value in 92.0%, which indicates their accuracy and quality. These data can be used for developing new MRI-based automated morphometry products.

Full Text

BACKGROUND

In recent years, an unswerving growth is observed in the interest to using the artificial intelligence technologies (AI) in healthcare in general and in the radiodiagnostics in particular [1–6]. The automated analysis of the results of diagnostic examinations is considered a potentially effective tool for increasing the productiveness and the quality of operations performed by radiologists, for optimizing the processes in the radiodiagnostics departments along with solving the issues of staff shortage. Constant growth is observed in the number of scientific publications on the AI topic, however, the number of scientific research products does not mean their quality, both in terms of the publications themselves and in terms of the proposed AI-based decisions. The vast majority of the published developments still includes experimental prototypes, while their underlying mathematical models were taught and tested using limited data samples. The retrospective assessment is predominant in assessing the accuracy of AI-technologies in laboratory settings, while the quantity of clinical research works with proper AI quality and applicability are vanishingly small. In the 21st century, it is difficult to believe in the situation, when a novel drug substance is tested only in a laboratory, not being tested in a full-scale clinical research. In terms of the AI- technologies, exactly this pattern is observed, which is completely unacceptable for modern medical sciences.

In Russia, from the year of 2020, an experiment is carried out on using the innovative technologies in computer vision for analyzing medical images and further use of these technologies in the healthcare system of the City of Moscow (hereinafter — the Moscow experiment; mosmed.ai) [7]. Currently, this is the world’s largest scientific prospective multicenter research on the applicability, safety and quality of AI. Within the frameworks of the Moscow experiment, a two-staged research is conducted on using the AI-based software (the so-called AI-services), developed for solving strictly specific diagnostic tasks. The first retrospective stage involves quite a standard testing using the reference data sets, however, the second (the main) and the prospective stages included the evaluation of the quality and the stability of AI-services when operating with the real flow of examination results in the settings of practical healthcare [8–10]. Due to the long-term (lasting months and years) use of AI-technologies in real-time practice, a possibility arouses for studying their effects on the working processes in radiodiagnostics, moreover — not in general, but in a context of solving a strictly specific working operation and clinical task. One of such quite specific tasks is the morphometry of the dimensions and the volume of the prostate gland when interpreting and describing the results of magnetic resonance imaging (MRI) of the lesser pelvis. It should be noted that, within the structure of the urological diseases in men, oncological and inflammatory diseases of the prostate gland prevail, which makes their screening and diagnostics an exceptionally topical issue [11, 12].

Morphometry represents a routine procedure, which takes the working time of the radiologist and creates potential risks due to the subjectiveness of measurements. Its is evident that, for increasing the productivity and the quality of work, this procedure can be easily automated. For exactly the automatization of measurements has shown a significant shortening of the duration of describing the results of radiology examinations along with the parallel increase in the productivity of the operations performed by the radiologist [13]. It is worth noting that previously, a set of data was compiled — the “MosMedData MRI of the lesser pelvis with morphometry parameters of the prostate gland” for the purpose of calibration testing, representing a structurized set of two hundred de-personalized results of multi-parametric magnetic-resonance imaging examinations data among adult men with the presence of morphometry marks (vertical, anterior-posterior and frontal dimensions of the prostate gland in millimeters relative to axis of the organ)1. The images are provided in the DICOM 3.0 format, with the marking provided as the XLSX file.

With a background of the colossal interest in using AI in radiodiagnostics, the challenges of automated analysis of the results of the lesser pelvis MRI are practically not researched. In particular, not so long ago, approximately 11 AI-based developments were published for analyzing the results of the prostate gland MRI. None of them is intended for direct helping the physician by means of the automatization of the routine mechanical procedure of measuring the target organ, though the developers are trying, mainly with no success, to solve the hardest challenges of differential diagnostics. Another substantial defect is that all these developments were tested only retrospectively or in the contest mode (competing with each other), i.e. the verification and evaluation of accuracy in the real clinical settings were not implemented [14, 15].

Research aim — to evaluate the quality and the validity of automatic morphometry of the prostate gland based on the magnetic resonance imaging results using the artificial intelligence technologies in the settings of practical healthcare.

METHODS

Research design

This was a prospective diagnostic research conducted in accordance with the methodology of reporting the results of scientific research with using the STARD 2015 diagnostic tests2.

Conformity criteria

Inclusion criteria: male patients older than 18 years; MRI scanning performed in the out-patient settings following the standard protocol; scanning protocol: loc (locators), Т2-WI (Т2-weighted images), Т1-WI (Т1-weighted images), DWI (diffusion-weighted images) with fat tissue suppression and with computing the ADC (apparent diffusion coefficient) charts along with dynamic contrast enhancement (DCE); the presence of the results of automated analysis (operation of the AI-service); the presence of informed voluntary consent for conducting the research.

Exclusion criteria: motor artifacts, artifacts from foreign objects at the investigated level; technical defects of MRI-scanning; technical defects in the results of AI-service operation.

Research facilities

The examination was carried out within the premises of the State Budgetary Healthcare Institution “Scientific-Practical Clinical Center for Diagnostics and Telemedicine of the Moscow Healthcare Department” (SBHI SPCC D&T, MHD). The research was carried out with using the results of radio-diagnostic examinations, conducted at the medical organizations of the Healthcare Department of the City of Moscow, providing medical aid to adult population in the out-patient settings (municipal polyclinics). The examination results were stored at the centralized archive of medical images of the City of Moscow (the Unified Radiology Information Service of the Unified Medical Information-Analytical System of the City of Moscow, URIS of UMIAS), while their description and compilation of protocols were conducted by radiology physicians of the reference-center within the premises of the SBHI SPCC D&T (Moscow Healthcare Department).

In accordance with the Decree issued by the Moscow Government on November 21, 2019 No.1543-PP3, SBHI SPCC D&T (MHD) is an operator of the experiment on using the innovative technologies in the field of computer vision for analyzing the medical images and further using these technologies in the Healthcare system of the City of Moscow (the Moscow experiment). The staff of the institution were performing the independent testing and quality control of the operation of the software products based on AI technologies. The developers and the right holders of the said software products are the third parties — the companies and enterprises of various form of incorporation, not affiliated by the SPCC D&T (MHD).

Research duration

The research work was arranged during the period from 01.04.2024 until 31.10.2024. Within the stated period of time, the tested software product based on AI-technologies was analyzing the results of MRI-scanning of the target area in accordance with the Moscow experiment procedures. A monthly sample was compiled for monitoring the operational quality (see below for detailed description of this procedure). The summarization and the analysis of the results were conducted during the period from 01.01.2025 until 01.03.2025.

Research description

An evaluation was carried out of the applicability of AI technologies for the automatization of the prostate gland measurements in the settings of practical healthcare.

The Index-test (the test method) — is a software product based on AI technologies (AI-service) for the recognition and analysis of MRI-scans of the prostate gland, integrated into the URIS UMIAS in accordance with the procedures of the Moscow experiment.

The functions of the AI-service are the following: measuring the vertical, the anterior-posterior (sagittal) and the frontal (transverse) dimensions of the prostate gland in millimeters relative to the axis of the organ (urethra); calculation of the prostate gland volume [16]. The research includes the IMV PIRADS AI-service (“Imvision” LLC, Russia) — the only participant of the Moscow experiment in the stated field (the limitations of this research shall be provided in the corresponding section at the end of the article).

Reference test: clinical monitoring of the quality of AI-service operation in accordance with the original methods, developed and validated within the settings of the Moscow experiment [14, 17, 18].

Methods of arranging the clinical monitoring of the quality of AI operation. Compilation of the sample from the whole volume of MRI-scans analyzed by the AI-service for the report period. The sample is to be compiled randomly, while its size is pre-justified and equals 80 scans every month [1]. The sample is reviewed by two radiologists with a work experience of not less than 5 years.

Each expert evaluates the results of automated analysis of this examination by two criteria: the correctness of detecting and labeling the location of the pathological signs (labeling by AI-service); the correctness of AI-service interpretation of the results of radiology examination (conclusion from the AI-service).

Taking into consideration the morphometric characteristics of the tested AI-service, the first criterion was considered the correctness of segmentation, while the second was the measuring of the volume and size of the prostate gland. For each examination result, the expert was setting the variant of assessment: full conformity (1 point), incorrect assessment (0.5 points), false-positive result (0.25 points), false-negative result (0 points). The assessment variant was defined for each of the abovementioned criteria separately, then all the obtained points were summed; the maximum possible value of the sum for this sample was taken as 100.0%, after which the specific weight was calculated for the added sum of points; eventually, the level of clinical assessment was obtained, varying within a range from 0.0 to 100.0%.

In this context, the following approaches were used: the false-positive result means erroneous measuring of the dimensions or the volume of the prostate gland, resulting in the definition of the case as the pathological with initially normal status of the target organ; false-negative result — erroneous measurement of the dimensions or the volume of the prostate gland, resulting in the definition of the case as normal with the actual presence of pathological changes. Erroneous measurements can be due to incorrect segmentation or due to the defects of mathematical calculation, related to the classification of the results of automated analysis. An incorrect assessment means the presence of discrepancies in the measurements conducted by the AI-service and by the radiologist, however, such discrepancies are not resulting in the occurrence of false-positive or false-negative result.

The sample from the second stage included the results of clinical monitoring sessions, conducted within 7 months (each month, a new sample consisting of 80 results of the prostate gland MRI, processed by the morphometry AI-service, was compiled for monitoring purposes). Respectively, the whole research included 560 cases.

Statistical analysis

The research had not established comparison groups, due to which, the sample size was not calculated. At the same time, the sample size for the regular monitoring of the quality of AI operation is 80 scans each month. The approaches for its determination were published previously [19].

The MedCalc v. 23.1.1 statistical software (MedCalc Software Ltd, Belgium) was used during the research. No special statistical criteria were used, only the descriptive statistics procedures were applied.

RESULTS

Research sample (participants)

In the settings of practical healthcare, an evaluation was carried out of the diagnostic accuracy of the AI-service for the morphometry of the prostate gland. Using this software product in accordance with methodology of the Moscow experiment, a total of seven clinical monitoring sessions were carried out (one monitoring each month, sample size — 80 cases). As appropriate, samples of examination results were compiled, allowing for evaluating the diagnostic accuracy of the operation of AI technologies in the prospective design, i.e. in the settings of practical healthcare. For the monitoring purposes, a total of 560 prostate gland MRI results with the results of automated analysis were selected randomly.

Primary findings

The full conformity of the results of AI-service operation was obtained in 71.6% (n=401) of the cases (Table 1). Quite high was the percentage of cases with partial agreement with the expert physician — 20.5% (n=115). The specific weight of false-negative cases was 3.8% (n=22). Notably, almost the same level was shown for the percentage of technical defects occurring during the processing of MRI-scanning results — 3.8%, i.e. in a total of 21 cases, the AI-service was shown as technically unreliable.

 

Table 1

The parameters of clinical monitoring the of AI-service operation in the morphometry of the prostate gland based on the results of magnetic resonance imaging

Number

(n=560)

Evaluation

Full conformity

Incorrect assessment

FP

FN

Defect

Abs.

401

115

1

22

21

%

71.6

20.5

0.2

3.9

3.8

Note. FP — false-positive result; FN — false-negative result.

 

The clinical assessment was conducted each month for the sample of 80 examination results, while its values varied within a range from 88.0 to 97.0%. For the total sample of 560 scanning results, i.e. for the whole period of using the AI-service in the settings of practical healthcare, the clinical assessment parameter value was 92.0%.

Separate analysis is required for the episodes of incorrect operation of the AI-service. In total, 138 results of automated analysis were to one degree or another assessed as low-quality (Fig. 1). Within the structure of the erroneous or incorrect operations of the AI-service, detected during the monitoring, the incorrect delineation of the prostate gland was prevailing — 52.0%; in 14.0% of the incorrect cases, the segmentation of one of the projections was missing. What calls attention to itself is the high percentage of incorrect measurements (including the incorrect arithmetic calculation of the volume) with a background of completely adequate segmentation of the target organ — 13.0%.

 

Fig. 1. The structure of incorrect or erroneous operation of the AI-service during the morphometry of the prostate gland on the results of magnetic resonance imaging, %.

 

The most typical segmentation errors were related to the presence of prolapsing apex of the prostate gland in the urinary bladder, to the contouring of the gland with capturing the dilated venous plexuses or the seminal vesicles. The examples of automated analysis performed by the AI-service for the morphometry of the prostate gland by the MRI results are provided in Fig. 2 and 3.

 

Fig. 2. The results of magnetic resonance imaging of the lesser pelvis organs in males aged 65 (а) and 67 (b) years, analyzed by the AI-service: segmentation and morphometry of the prostate gland were done correctly. 1 — sagittal projection; 2 — axial projection.

 

Fig. 3. The results of magnetic resonance imaging of the lesser pelvis organs in a male aged 86 years, analyzed using the AI-service: an example of incorrect segmentation of the prostate gland (the contour of the prostate gland in the sagittal projection is not marked in full range).

 

The obtained data show quite high diagnostic quality of the tested tool. Hence, the AI technology (computer vision) can be considered applicable for the automatization of the prostate gland measurements.

DISCUSSION

The problematics of using AI technologies for analyzing the results of the MRI of the lesser pelvis (in particular, of the prostate gland) in scientific literature is enlightened to a certain degree unidirectionally [20–22]. General compilation is provided for the possible tasks for development in the corresponding topical area, the number of which includes the segmentation and the search for pathological foci, the classification and increasing the quality of multi-parametric imaging, the detection and the differential diagnostics of malignant neoplasms, the classification of risk degree by the Gleason score4 [23–25]. In the real development, the prevailing are the oncological diagnostics, the use of radiomics methods, as well as the multimodal approach expressed as the combined use of the MRI results and the pathomorphological examination of the prostate gland [26]. Specifically, the radiomics models have shown a high accuracy of differential diagnostics for the foci in the prostate gland. The fact that increases the value of such publications is that the developed models also pass the external validation, i.e. the independent verification via a set of new data. The meta-analysis summarizing the data from 43 articles (9983 patients) has allowed for obtaining the mean accuracy values (area under the characteristic curve) of the radiomic models — 0.91–0.93 [27], at the same time, there are still disputable issues of the precision of the results obtained when using such models, their applicability not in the laboratory, but in clinical settings. With this background, the tasks of the prostate gland morphometry have undeservedly little attention. Meanwhile, performing the routine measuring procedures by the radiology physician “manually” leads to wasting time and can always result in errors in terms of accuracy and precision [28–32].

In general, the scientific literature shows a sufficiently high level of diagnostic precision of the AI (based on the typical neural network architectures) when analyzing the MRI of the prostate gland. Thus, upon the automated segmentation of the prostate gland, the Dice coefficient ranges within 0.86–0.9 [33, 34], upon the classification of the pathological manifestations, the area under the characteristic curve also reaches 0.84–0.91 [35, 36]. The accuracy of detecting the pathological foci is lower and falls within a range of 0.64–0.81 [37]. During the comparative research, it was found that AI has surpassed the international group of radiologists (consisting of 62 specialists) in terms of detecting the clinically significant prostate cancer and its classification using the PI-RADS scale (Prostate Imaging Reporting and Data System) [38]. The significant downside of these research works is their experimental pattern. All the AI-related research on diagnostics of abnormalities in the prostate gland were conducted in laboratory settings, using the reference data sets (including the quite vast comparison of accuracy between the AI and the group of 62 physicians). Moreover, the independent analysis of the quality of such articles has shown that 47.0% of them were lacking the full and correct description of the reference data set, i.e. of the main AI accuracy measuring tool. Even more negative is the fact that up to 92.0% articles contained data on manipulating the statistical analysis data for the purpose of concealing the low accuracy of the model [7].

With this background, our research was carried out in the prospective design, allowing for determining the accuracy and quality of AI operation in the settings of practical healthcare. For the first time ever, quite a vast set of material was used to obtain the parameters of accuracy and stability of AI-service operation for the morphometry of the prostate gland, distinct by its scientific novelty. The obtained clinical measurement value of 92.0% indicates the sufficiently high accuracy of the tested AI-service operating with the real flow of radiodiagnostics data. At the same time, the analysis of the structure of the defects has allowed for objectively detecting problems with the segmentation of the target organ. Based on the experience of the Moscow Experiment, incorrect segmentation is a typical error of the AI-services, especially often occurring at the early stages of their development (for this specific clinical task) [7]. The studied AI-service was introduced into the experiment relatively recently, which is why the said type of defect is quite explainable. The results of the clinical monitoring can be used both for the elimination of specific defect and for the general improvement of the AI-service along with its preparation for clinical testing for the purpose of receiving the status of the medical device.

Research limitations

The research work included only one software product based on AI technologies. This situation is due to the absence of other developments capable of solving the tasks of automated morphometry of the prostate gland. As of the date of 01.03.2025, in the Russian Federation, there are no AI-based devices for solving this challenge; the research includes only one relevant participant of the Moscow experiment. Evidently, certain measures are required for stimulating the development of other solutions, including the ones having the functional capabilities of supporting the diagnostic decisions based on the radiomics.

CONCLUSION

The automatization of routine measurements greatly contributes to standardizing the processes of describing the results of radio-diagnostic examinations. This aspect has a special importance from the point of view of providing the continuity of medical aid in cases of the patient presenting to various medical organizations.

The AI technologies are applicable for the automatization of the prostate gland measurements when describing the results of MRI-scanning of the lesser pelvis organs. The experience of using the corresponding AI-service in the settings of practical healthcare has shown the clinical measurement value of 92.0%, which allows for characterizing the accuracy and the quality of its operation in a flow of MRI-scanning results as high.

The obtained data can be used as the methodical material for developing other products for the automated morphometry of the prostate gland based on the results of MRI-scanning of the lesser pelvis.

ADDITIONAL INFORMATION

Author contributions. N.M. Nasibian: literature review, data collection and processing, analysis of results, manuscript preparation; A.V. Vladzymyrskyy: study conception, final editing, manuscript approval; K.M. Arzamasov: final editing, manuscript approval. Thereby, all authors provided approval of the version to be published and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Ethics approval. This study was carried out in accordance with the framework established by the Experiment on the Use of Innovative Technologies in Computer Vision for Medical Image Analysis, with subsequent applications in the healthcare system of Moscow. The research received approval from the ethical committee, as documented in protocol No. 2 of IEC of MRB of the RSR dated February 20, 2020. Additionally, the study is registered on ClinicalTrials under the identifier NCT04489992. All patients whose images were included in the study signed an informed voluntary consent upon admission to the hospital to use the results of the examination and treatment for scientific purposes.

Funding sources. This paper was prepared by a team of authors as a part of the research and development project “Development of a platform to generate datasets containing diagnostic imaging studies” (EGISU No 123031500003-8) in accordance with Order No. 1258 dated December 22, 2023: “On approval of state assignments funded from the Moscow city budget to state budgetary (autonomous) institutions subordinated to the Moscow Healthcare Department for 2024 and the planning period of 2025 and 2026”.

Disclosure of interests. The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Statement of originality. The authors did not use previously published information (text, illustrations, data) in conducting the research and creating this paper.

Data availability statement. The editorial policy regarding data sharing is not applicable to this work, data can be published as open access.

Generative AI. Generative AI technologies were not used for this article creation.

Provenance and peer-review. This paper was submitted to the journal on an initiative basis and reviewed according to the usual procedure. Two external reviewers and the scientific editor of the publication participated in the review.

 

1 Certificate of data base registration — RU 2025620045/09.01.2025. Application No. 2024626323 dated 20.12.2024. Vasilyev Yu.А., Nasibyan N.М., Vladzimirskiy А.V. et. al. MosMedData: MRI of lesser pelvis with the morphometry parameters of the prostate gland. EDN: IXRMQR

2 Certificate of registration of PC software — RU 2025610804/14.01.2025. Application No. 2024691653 dated 20.12.2024. Vasilyev Yu.А., Vladzimirskiy А.V., Omelyanskaya О.V. et al. Data set preparation platform. EDN: TZQQHN

3 Decree issued by the Moscow Government on November 21, 2019 No.1543-PP «On the conduction of the experiment on using the innovative technologies in the field of computer vision for analyzing the medical images and their further use in the Healthcare System of the City of Moscow». Access mode: https://www.garant.ru/products/ipo/prime/doc/73059396/

4 Certificate of state registration of database No. 2024620575 dated 06.02.2024. Application No. 2024620252/26.01.2024. Vasilyev Yu.А., Blokhin I.А., Geleje P.B. et al. A set of biparametric MRI data of the prostate gland with histological verification. EDN: XEAAGM. Access mode: https://telemedai.ru/nauka/nauchnaya-infrastruktura/nauchnaya-deyatelnost/intellektualnaya-sobstvennost

×

About the authors

Nelli M. Nasibian

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies

Author for correspondence.
Email: nelli-nasibyan94@yandex.ru
ORCID iD: 0009-0004-4620-6204
SPIN-code: 4936-2738
Russian Federation, 24 Petrovka st, bldg 1, Moscow, 127051

Anton V. Vladzymyrskyy

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies

Email: VladzimirskijAV@zdrav.mos.ru
ORCID iD: 0000-0002-2990-7736
SPIN-code: 3602-7120

MD, PhD

Russian Federation, 24 Petrovka st, bldg 1, Moscow, 127051

Kirill M. Arzamasov

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies

Email: ArzamasovKM@zdrav.mos.ru
ORCID iD: 0000-0001-7786-0349
SPIN-code: 3160-8062

MD, PhD

Russian Federation, 24 Petrovka st, bldg 1, Moscow, 127051

References

  1. Васильев Ю.А., Владзимирский А.В., Омелянская О.В., и др. Обзор метаанализов о применении искусственного интеллекта в лучевой диагностике // Медицинская визуализация. 2024. Т. 28, № 3. С. 22–41. [Vasiliev YuA, Vladzymyrskyy AV, Omelyanskaya OV, et al. Review of meta-analyses on the use of artificial intelligence in radiology. Medical visualization. 2024;28(3):22–41]. doi: 10.24835/1607-0763-1425 EDN: QYASNZ
  2. Kelly BS, Judge C, Bollard SM, et al. Radiology artificial intelligence: a systematic review and evaluation of methods (RAISE). Eur Radiol. 2022;32(11):7998–8007. doi: 10.1007/s00330-022-08784-6
  3. Hosny A, Parmar C, Quackenbush J, et al. Artificial intelligence in radiology. Nat Rev Canc. 2018;18(8):500–510. doi: 10.1038/s41568-018-0016-5
  4. Katzman BD, van der Pol CB, Soyer P, Patlas MN. Artificial intelligence in emergency radiology: A review of applications and possibilities. Diagn Interv Imaging. 2023;104(1):6–10. doi: 10.1016/j.diii.2022.07.005
  5. Ваньков В.В., Артемова О.Р., Карпов О.Э., и др. Итоги внедрения искусственного интеллекта в здравоохранении России // Врач и информационные технологии. 2024. № 3. С. 32–43. [Vankov VV, Artemova OR, Karpov OE, et al. Results of the implementation of artificial intelligence in the Russian healthcare. Medical doctor and information technology. 2024;(3):32–43]. doi: 10.25881/18110193_2024_3_32 EDN: TIASHB
  6. Гусев А.В., Артемова О.Р., Васильев Ю.А., Владзимирский А.В. Внедрение медицинских изделий с технологиями искусственного интеллекта в здравоохранении России: итоги 2023 г. // Национальное здравоохранение. 2024. Т. 5, № 2. С. 17–24. [Gusev AV, Artemova OR, Vasiliev YuA, Vladzymyrskyy AV. Integration of ai-based software as a medical device into Russian healthcare system: Results of 2023. National Health Care (Russia). 2024;5(2):17–24]. doi: 10.47093/2713 EDN:
  7. Владзимирский А.В., Васильев Ю.А., Арзамасов К.М., и др. Компьютерное зрение в лучевой диагностике: первый этап Московского эксперимента. 2-е изд. Москва: Издательские решения, 2023. 388 с. [Vladzimirsky AV, Vasiliev YA, Arzamasov KM, et al. Computer vision in radiation diagnostics: The first stage of the Moscow experiment. 2nd ed. Moscow: Izdatel’skie resheniya; 2023. 388 р. (In Russ.)]. EDN: FOYLXK
  8. Аполихин О.И., Сивков А.В., Комарова В.А., Никушина А.А. Болезни предстательной железы в Российской Федерации: статистические данные 2008-2017 гг. // Экспериментальная и клиническая урология. 2019. № 2. С. 4–13. [Apolikhin OI, Sivkov AV, Komarova VA, Nikushina AA. Prostate diseases in the Russian Federation: Statistical data for 2008-2017. Experimental & clinical urology. 2019;(2):4–13]. doi: 10.29188/2222-8543-2019-11-2-4-12 EDN: DHXMJP
  9. Chen J, He L, Ni Y, et al. Prevalence and associated risk factors of prostate cancer among a large Chinese population. Sci Rep. 2024;14(1):26338. doi: 10.1038/s41598-024-77863-z
  10. Tan EH, Burn E, Barclay NL, et al.; OPTIMA Consortium. Incidence, prevalence, and survival of prostate cancer in the UK. JAMA Network Open. 2024;7(9):e2434622. doi: 10.1001/jamanetworkopen.2024.34622
  11. Fernandes MC, Yildirim O, Woo S, et al. The role of MRI in prostate cancer: Current and future directions. Magma (New York). 2022;35(4):503–521. doi: 10.1007/s10334-022-01006-6
  12. Stempel CV, Dickinson L, Pendsé D. MRI in the management of prostate cancer. seminars in ultrasound, CT, and MR. 2020;41(4):366–372. doi: 10.1053/j.sult.2020.04.003
  13. Sunoqrot MR, Saha A, Hosseinzadeh M, et al. Artificial intelligence for prostate MRI: Open datasets, available applications, and grand challenges. Eur Radiol Exp. 2022;6(1):35. doi: 10.1186/s41747-022
  14. Васильев Ю.А., Владзимирский А.В., Омелянская О.В., и др. Методология тестирования и мониторинга программного обеспечения на основе технологий искусственного интеллекта для медицинской диагностики // Digital Diagnostics. 2023. Т. 4, № 3. С. 252–267. [Vasiliev YuA, Vlazimirsky AV, Omelyanskaya OV, et al. Methodology for testing and monitoring artificial intelligence-based software for medical diagnostics. Digital Diagnostics. 2023;4(3):252–267]. doi: 10.17816/DD321971 EDN: UEDORU
  15. Четвериков С.Ф., Арзамасов К.М., Андрейченко А.Е., и др. Подходы к формированию выборки для контроля качества работы систем искусственного интеллекта в медико-биологических исследованиях // Современные технологии в медицине. 2023. Т. 15. № 2. С. 19–27. [Chetverikov SF, Arzamasov KM, Andreichenko AE, et al. Approaches to sampling for quality control of artificial intelligence in biomedical research. Modern technologies in medicine. 2023;15(2):19–27]. doi: 10.17691/stm2023.15.2.02 EDN: FUKXYC
  16. Cohen JF, Korevaar DA, Altman DG, et al. STARD 2015 guidelines for reporting diagnostic accuracy studies: explanation and elaboration. BMJ Open. 2016;6(11):e012799. doi: 10.1136/bmjopen-2016-012799
  17. Maki JH, Patel NU, Ulrich EJ, et al. Part I. Prostate cancer detection, artificial intelligence for prostate cancer and how we measure diagnostic performance: A comprehensive review. Curr Probl Diagn Radiol. 2024;53(5):606–613. doi: 10.1067/j.cpradiol.2024.04.002
  18. Bozgo V, Roest C, van Oort I, et al. Prostate MRI and artificial intelligence during active surveillance: Should we jump on the bandwagon? Eur Radiol. 2024;34(12):7698–7704. doi: 10.1007/s00330-024
  19. Бобровская Т.М., Васильев Ю.А., Никитин Н.Ю., и др. Объем выборки для оценки диагностической точности программного обеспечения на основе технологий искусственного интеллекта в лучевой диагностике // Сибирский журнал клинической и экспериментальной медицины. 2024. Т. 39, № 3. С. 188–198. [Bobrovskaya TM, Vasilev YuA, Nikitin NYu, et al. Sample size for assessing a diagnostic accuracy of ai-based software in radiology. Siberian Journal of Clinical and Experimental Medicine. 2024;39(3):188–198]. doi: 10.29001/2073-8552-2024-39-3-188-198 EDN: BPCPLL
  20. Belue MJ, Turkbey B. Tasks for artificial intelligence in prostate MRI. Eur Radiol Exp. 2022;6(1):33. doi: 10.1186/s41747-022-00287-9
  21. Harmon SA, Tuncer S, Sanford T, et al. Artificial intelligence at the intersection of pathology and radiology in prostate cancer. Diagnost Intervent Radiol (Ankara, Turkey). 2019;25(3):183–188. doi: 10.5152/dir.2019.19125
  22. Рева С.А., Шадеркин И.А., Зятчин И.В., Петров С.Б. Искусственный интеллект в онкоурологии // Экспериментальная и клиническая урология. 2021. Т. 14, № 2. С. 46–51. [Reva SA, Shaderkin IA, Zyatchin IV, Petrov SB. Artificial intelligence in cancer urology literature review. Experimental & clinical urology. 2021;14(2):46–51]. doi: 10.29188/2222-8543-2021-14-2-46-51 EDN: BHBSNP
  23. Абоян И.А., Редькин В.А., Назарук М.Г. и др. Искусственный интеллект в диагностике рака предстательной железы с помощью магнитно-резонансной томографии. Новый подход // Онкоурология. 2024. Т. 20, № 2. С. 35–43. [Aboyan IA, Redkin VA, Nazaruk MG, et al. Artificial intelligence in diagnosis of prostate cancer using magnetic resonance imaging. New approach. Cancer urology. 2024;20(2):35–43]. doi: 10.17650/1726-9776-2024-20-2-35-43 EDN: MBGNUY
  24. Алифов Д.Г., Звезда С.А., Кельн А.А., и др. Лучевая диагностика рака простаты на основе искусственного интеллекта и радиомного машинного обучения // Университетская медицина Урала. 2021. Т. 7, № 4. С. 48–50. [Alifov DG, Zvezda SA, Cologne AA, et al. Radiation diagnostics of prostate cancer based on artificial intelligence and radome machine learning. Universitetskaya meditsina Urala. 2021;7(4):48–50]. EDN: CUBNFQ
  25. Kaneko M, Magoulianitis V, Ramacciotti LS, et al. The novel green learning artificial intelligence for prostate cancer imaging: A balanced alternative to deep learning and radiomics. Urol Clin North Am. 2024;51(1):1–13. doi: 10.1016/j.ucl.2023.08.001
  26. Lu Y, Yuan R, Su Y, et al. Biparametric MRI-based radiomics for noninvastive discrimination of benign prostatic hyperplasia nodules (BPH) and prostate cancer nodules: A bio-centric retrospective cohort study. Sci Rep. 2025;15(1):654. doi: 10.1038/s41598-024-84908-w
  27. Lomer NB, Ashoobi MA, Ahmadzadeh AM, et al. MRI-based radiomics for predicting prostate cancer grade groups: A systematic review and metaanalysis of diagnostic test accuracy studies. Acad Radiol. 2024:S1076-6332(24)009541. doi: 10.1016/j.acra.2024.12.006
  28. Учеваткин А.А., Юдин А.Л., Афанасьева Н.И., Юматова Е.А. Оттенки серого: как и почему мы ошибаемся // Медицинская визуализация. 2020. Т. 24, № 3. С. 123–145. [Uchevatkin AA, Yudin AL, Afanas’yeva NI, Yumatova EA. Shades of grey: How and why we make mistakes. Medical visualization. 2020;24(3):123–145]. doi: 10.24835/1607-0763-2020-3-123-145 EDN: XVKCLV
  29. Van der Loo I, Bucho TM, Hanley JA, et al. Measurement variability of radiologists when measuring brain tumors. Eur J Radiol. 2024;183:111874. doi: 10.1016/j.ejrad.2024.111874
  30. Sanford TH, Zhang L, Harmon SA, et al. Data augmentation and transfer learning to improve generalizability of an automated prostate segmentation model. AJR. 2020;215(6):1403–1410. doi: 10.2214/AJR.19.22347
  31. Ushinsky A, Bardis M, Glavis-Bloom J, et al. A 3D-2D hybrid U-net convolutional neural network approach to prostate organ segmentation of multiparametric MRI. AJR. 2021;216(1):111–116. doi: 10.2214/AJR.19.22168
  32. Wang B, Lei Y, Tian S, et al. Deeply supervised 3D fully convolutional networks with group dilated convolution for automatic MRI prostate segmentation. Med Phys. 2019;46(4):1707–1718. doi: 10.1002/mp.13416
  33. Le MH, Chen J, Wang L, et al. Automated diagnosis of prostate cancer in multi-parametric MRI based on multimodal convolutional neural networks. Phys Med Biol. 2017;62(16):6497–6514. doi: 10.1088/1361
  34. Liu S, Zheng H, Feng Y, Li W. Prostate cancer diagnosis using deep learning with 3D multiparametric MRI. In: Medical imaging: Computer-aided diagnosis. Computer Vision and Pattern Recognition (cs.CV); Machine Learning (stat.ML); 2017. P. 1013428. doi: 10.1117/12.2277121
  35. Cao R, Mohammadian Bajgiran A, Afshari Mirak S, et al. Joint prostate cancer detection and gleason score prediction in mp-MRI via FocalNet. IEEE Trans Med Imaging. 2019;38(11):2496–2506. doi: 10.1109/TMI.2019.2901928
  36. Ishioka J, Matsuoka Y, Uehara S, et al. Computeraided diagnosis of prostate cancer on magnetic resonance imaging using a convolutional neural network algorithm. BJU Int. 2018;122(3):411–417. doi: 10.1111/bju.14397
  37. Saha A, Bosma JS, Twilt JJ, et al.; PI-CAI consortium. Artificial intelligence and radiologists in prostate cancer detection on MRI (PI-CAI): An international, paired, non-inferiority, confirmatory study. Lancet Oncol. 2024;25(7):879–887. doi: 10.1016/S1470
  38. Belue MJ, Harmon SA, Lay NS, et al. The low rate of adherence to checklist for artificial intelligence in medical imaging criteria among published prostate MRI artificial intelligence algorithms. J Am Coll Radiol. 2023;20(2):134–145. doi: 10.1016/j.jacr.2022.05.022

Supplementary files

Supplementary Files
Action
1. JATS XML
Download
2. Fig. 2. Results of magnetic resonance imaging of the pelvic organs in men aged 65 (a) and 67 (b) years, analyzed by the AI service: segmentation and morphometry of the prostate gland are performed correctly. 1 — sagittal projection; 2 — axial projection.

Download (2MB)
Indexing metadata
3. Fig. 1. The structure of incorrect or erroneous operation of the AI-service during the morphometry of the prostate gland on the results of magnetic resonance imaging, %.

Download (1MB)
Indexing metadata
4. Fig. 2. The results of magnetic resonance imaging of the lesser pelvis organs in males aged 65 (а) and 67 (b) years, analyzed by the AI-service: segmentation and morphometry of the prostate gland were done correctly. 1 — sagittal projection; 2 — axial projection.

Download (2MB)
Indexing metadata
5. Fig. 3. The results of magnetic resonance imaging of the lesser pelvis organs in a male aged 86 years, analyzed using the AI-service: an example of incorrect segmentation of the prostate gland (the contour of the prostate gland in the sagittal projection is not marked in full range).

Download (1MB)
Indexing metadata

Copyright (c) 2025 Eco-Vector

Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
СМИ зарегистрировано Федеральной службой по надзору в сфере связи, информационных технологий и массовых коммуникаций (Роскомнадзор).
Регистрационный номер и дата принятия решения о регистрации СМИ: серия ПИ № ФС 77 - 38032 от 11 ноября 2009 года.