Automated radiology workstation: comparing two approaches for selecting radiodiagnostic solutions (technical report)

Cover Image


Cite item

Full Text

Abstract

BACKGROUND: Patient examinations generate large amounts of data on various diseases, which are virtually impossible to process manually. Currently, automated radiology workstations incorporate clinical decision–support systems that facilitate image analysis. At present, two options are used: built-in vendor-dependent and vendor-independent software solutions. Both have broad functionality, but they also have their advantages and drawbacks. Therefore, the choice of the best clinical decision–support system depends on the healthcare organization.

AIM: This study aimed to compare two approaches for selecting radiodiagnostic solutions for automated radiology workstations in Moscow.

METHODS: The study conducted a two-stage, cross-sectional survey between September 2023 and March 2024. Data on vendor-independent software were obtained from participants of the Experiment on the use of innovative computer vision technologies for analysis of medical images in the Moscow healthcare system. Conversely, data on vendor-dependent solutions was collected from the manufacturers’ websites. Furthermore, a survey of 40 radiologists was performed to evaluate the relevance of software functions.

RESULTS: At the time of the survey, the vendor-independent software demonstrated slightly more functions than the vendor-dependent solutions. The greatest differences were observed in the functionality of computed tomography and magnetic resonance imaging modalities, whereas the functionality for X-ray imaging and mammography was nearly identical. The survey of radiologists showed that 6 of 17 functions were unique to built-in vendor-dependent software. However, <40% of the radiologists actually needed these functions, whereas the shared functions were relevant for >50% of the respondents.

CONCLUSION: Vendor-independent and built-in vendor-dependent software solutions share only half of their functions. Thus, in choosing between these two options, healthcare facilities should consider their specialization and their radiologists’ requests. Moreover, approximately two-thirds of the built-in vendor-dependent software functions used by radiologists in Moscow can be implemented using vendor-independent solutions. Therefore, the choice of software should be based on the facility’s technical capacity and economic feasibility.

Full Text

BACKGROUND

Information technology has become an integral part of current medical practices. The high patient load is putting significant strain on the healthcare system as a whole, as well as on individual healthcare providers. Moreover, the amount of data on different conditions has grown so large that it cannot be processed manually [1, 2]. Therefore, automated radiology workstations (ARWs) are increasingly being used by radiologists to support medical decision-making by analyzing medical images [3–5].

There are two common ways to configure an ARW:

  • Using a ready-made embedded solution from a radiological equipment manufacturer (a vendor-specific software solution); or
  • Using independent set of ARW components (a vendor-neutral software solution).1
  • Both options can:
  • Provide an independent conclusion: A radiologist can only access the software results once the scan interpretation is complete;
  • Support simultaneous interpretation: A radiologist can access the software results while viewing images obtained from diagnostic equipment;
  • Sort tests based on their interpretation priority: A radiologist first receives scan results marked as abnormal or urgent by the software; and
  • Exclude tests: Scan results marked by thesoftware asnon-abnormal are transferred to aradiologist as

Both vendor-specific and vendor-neutral solutions offer a wide range of features and are constantly improving through the use of artificial intelligence (AI) technologies [6]. This inevitably leads healthcare decision-makers such as ministers, chief medical officers, and their deputies, to answer a question whether vendor-specific or vendor-neutral solution is more suitable for a given healthcare provider.

AIM

To compare two strategies for selecting radiology software solutions for ARWs based in Moscow.

METHODS

This was a two-stage, cross-sectional survey.

The first stage evaluated the features of vendor-specific and vendor-neutral solutions available in Moscow. Information on vendor-neutral solutions was obtained from an AI Service Catalog used in the Experiment on the Use of Innovative Computer Vision Technologies for Medical Image Analysis and Further Use in the Moscow Healthcare System.2 Information on vendor-specific solutions was gathered from the websites of equipment manufacturers.34567891011121314151617

The second stage involved an opinion survey of 40 radiologists at the Moscow reference center. We developed a special questionnaire titled The Practical Value of the Key Features of Vendor-Specific Software Solutions (Supplement 1). This questionnaire was developed for this study independently and was not formally validated. The tool contained 17 closed-ended questions, each designed to evaluate a specific feature based on real-world experiences of the respondents. The practical value of the features was evaluated using a three-point scale: high, average, or low. The response option “Didn't use/don't know” is included with each question to avoid confounding factors.

The work was conducted from September 2023 to March 2024.

Ethics Approval

This study was based on the data from the approved Experiment on the Use of Innovative Computer Vision Technologies for Medical Image Analysis and Further Use in the Moscow Healthcare System (extract from Minutes No. 2 of the Independent Ethics Committee of the Moscow Regional Branch of the Russian Society of Radiographers and Radiologists dated February 20, 2020), registered at ClinicalTrials (NCT04489992).

RESULTS

Comparison of Feature Sets of Vendor-Specific and Vendor-Neutral Solutions

At the time of this survey, vendor-neutral solutions had slightly more features than vendor-specific solutions (Supplement 2).

Notably, both vendor-specific and vendor-neutral solutions had almost the same numbers of features for radiography. A similar situation was observed with mammography.

Vendor-neutral solutions had more available features than vendor-specific solutions for CT and MRI modalities.

Figure 1 shows the distribution of features between vendor-specific and vendor-neutral solutions. Approximately half of the features were implemented in both types of solutions.

 

Fig. 1. The number of features depending on their implementation in vendor-specific and vendor-neutral solutions.

 

Practical Value of Key Features of Vendor-Specific Solutions

Six of the 17 evaluated features did not have equivalent features in vendor-neutral solutions (Fig. 2).

 

Fig. 2. The practical value of key features of vendor-specific software solutions based on a survey of radiologists: results are presented as a percentage of the total number of respondents.

 

Six vendor-specific software features with no vendor-neutral equivalents (35.3%) were used by fewer than 40% of radiologists. Vascular analysis and brain perfusion assessment demonstrated the highest practical value, whereas colon cancer detection was rated the lowest.

Six of 11 vendor-specific software features with vendor-neutral equivalents (64.7%) were used by more than 50% of radiologists. Of this feature set, pulmonary aeration assessment was the least significant feature.

DISCUSSION

Although both vendor-specific and vendor-neutral solutions had some similar features, they also had unique ones.

Comparison of Features of Vendor-Specific and Vendor-Neutral Solutions

As mentioned above, the features of vendor-specific and vendor-neutral radiography solutions had nearly identical features such as detection of neoplasms, infectious and inflammatory changes, and injuries.

All mammography solutions focused on detecting signs of breast tumors.

The greatest differences in features were found between vendor-specific and vendor-neutral solutions for CT and MRI.

The vendor-specific solutions had a limited number of features for detecting lung conditions in chest CT scans (they did not detect tuberculosis, sarcoidosis, or bronchiectasis). However, vendor-specific solutions had unique features for examining the cardiovascular system, such as vessel analysis and 3D modeling, automated clot detection, and vessel and heart spectral analysis. The reverse situation was observed with vendor-neutral solutions.

Vendor-specific solutions were able to detect polyps and colon cancer in abdominal CT scans. Although vendor-neutral solutions did not have such features, they could detect signs of urolithiasis, and adrenal and kidney neoplasms.

Although vendor-specific solutions evaluated perfusion in brain MRI scans, they were unable to detect multiple sclerosis and neoplasms, which could be detected using vendor-neutral solutions.

Additionally, the vendor-neutral solutions offered unique features for detecting focal structural changes in vertebral bones, as well as protrusions, herniated discs, and spinal stenosis.

The greatest differences in features were found between vendor-specific and vendor-neutral solutions for CT and MRI.

Comparison of Advantages and Disadvantages of Vendor-Specific and Vendor-Neutral Solutions

Vendor-specific solutions are provided by manufacturers of diagnostic equipment and workstations. This has certain advantages:

  • Full compatibility of all components;
  • Service support;
  • No need to maintain a large team of IT systems support;
  • Unique, high-demand features for radiologists that are not available in vendor-neutral solutions.
  • Thedisadvantages ofvendor-specific solutions are asfollows:
  • High cost;
  • Incompatibility with workstations from third-party manufacturers;
  • Limited feature sets of individual modules.

These solutions are reported to have a high rate of false positives and to perform inconsistently in clinical practice [7–12]. This may be explained by the fact that unlike most vendor-neutral solutions, which are based on deep learning techniques, many vendor-specific solutions rely on signs that are selected manually or semi-automatically by experts, and do not account for the full range of potential findings [13, 14]. Nevertheless, some vendor-specific solutions demonstrate high diagnostic accuracy. In recent years, vendor-specific solutions based on deep learning have emerged that are equally accurate as vendor-neutral solutions. However, they are not routinely validated. Another potential disadvantage of vendor-specific solutions is their inability to identify the major types of abnormalities in a given anatomical region using a specific imaging technique, e.g., chest or abdomen CT scans.

Vendor-neutral solutions offer several advantages, such as regular quality improvements that are quickly made available to end users, and a wide variety of unique features. A growing number of vendor-neutral solutions can detect all major types of abnormalities in a given anatomical region using a specific imaging technique. Another advantage of vendor-neutral solutions is their flexible payment system, which means that the cost of use depends on the number of scans processed.18

The need for additional infrastructure is considered a disadvantage of vendor-neutral solutions. For example, small healthcare providers without stable, high-speed communication channels are unable to benefit from cloud solutions. Local installation of vendor-neutral solutions is possible in theory, but in practice, their performance may be substantially lower than developers claim [15–18]. Since 2020, Russia has been actively developing infrastructure to ensure the seamless implementation of diagnostic radiology solutions across all regions. A system of evidence-based technological and clinical monitoring ensures the consistently high performance level of these solutions [19–22].19

Additionally, the development potential of both vendor-specific and vendor-neutral solutions should be discussed. Many vendor-specific solutions rely on algorithms based on a limited set of signs that are selected manually or semi-automatically by experts. To achieve higher diagnostic accuracy, the underlying algorithms should be improved or changed, which seems very labor-intensive. Therefore, developers have recently been using deep learning for both vendor-specific and vendor-neutral solutions because it enables a relatively quick increase in diagnostic accuracy through retraining models with new datasets [23–25]. However, fast retraining requires adequate compute capacity.

Key Selection Criteria for Vendor-Specific and Vendor-Neutral Solutions

There is a clear need for a system that supports radiologists' medical decision-making processes at their ARWs, because such a system improves diagnostic quality and reduces the time needed to interpret scans [21, 26].

We believe that the type of healthcare provider who implements the solution will largely determine the choice of solution. Healthcare decision-makers such as ministers, chief medical officers, and their deputies, need to consider the following types of providers when making decisions.

Outpatient clinics

One of the primary functions of outpatient healthcare providers is to prevent diseases. Therefore, they perform a large number of preventive screenings, including fluorography, radiography, and mammography. As shown above, these modalities have a similar set of features in both vendor-specific and vendor-neutral solutions. Therefore, cost per scan will be the selection criterion.

Additionally, outpatient clinics are able to perform CT and MRI scans. Because these scans are used to identify conditions in the chest, abdomen, spine, and other anatomical regions, vendor-neutral solutions are preferable because they have the necessary features.

Reference centers (remote center of competence for diagnostic radiology)

According to Order No. 560н (560n) of the Ministry of Health of the Russian Federation dated June 9, 2020, “On Approval of the Rules for Conducting Radiological Examinations,”20 reference centers shall (including, but not limited to):

  • Interpret radiological results;
  • Organize and conduct a double review of mass screening results, using automated systems if applicable;
  • Investigate causes of inconsistencies in radiological results and develop and implement quality assurance measures.

Therefore, the criteria for selecting systems that support medical decision-making may be similar to the criteria described for outpatient clinics. However, it is important to consider the unique workflow of each reference center.

For example, in Moscow, radiological examinations are performed in outpatient clinics and the results are transferred to a unified radiological service. The radiologists at the Moscow Reference Center use this service to view and interpret results. This diagnostic radiology workflow does not allow for the use of vendor-specific solutions because ARWs are not directly connected to diagnostic equipment.

Therefore, specialists cannot use the unique features of vendor-specific solutions. However, the survey showed that fewer than 40% of radiologists found these features somewhat valuable.

Inpatient clinics

These healthcare providers typically specialize in one or more areas of treatment. If there are no other options, then the solution with the necessary functions should be selected. For example, vendor-specific solutions designed for cardiovascular scans are preferable at a cardiovascular center. The use of vendor-neutral solutions with the right features is more justified in degenerative spinal condition centers.

The choice between two types of solutions depends on the level of infrastructure development and economic feasibility, provided that the necessary tools are included.

Both vendor-specific and vendor-neutral solutions are still relevant for interpreting radiological scans. Which option is preferred depends on the characteristics of the healthcare provider.

Work Limitations

The analysis presented was based on the Moscow healthcare system. The characteristics of local healthcare systems should be considered when choosing between vendor-specific and vendor-neutral solutions for analyzing radiological scans in other regions of Russia and other countries.

In the survey, radiologists evaluated only the solutions available at their workplace. The accuracy of the practical value measurements may have been affected because the author-developed questionnaire was not validated.

The study was conducted from September 2023 to March 2024. During this time, the market for these products changed quickly, and developers introduced new features. For this reason, we recommend a market analysis before purchasing any solution.

The list of features was based on key diagnostic requirements [27] and information from websites of manufacturers.

CONCLUSION

Only half of the features in vendor-specific and vendor-neutral solutions available in Moscow overlap.

The unique features of vendor-specific and vendor-neutral solutions determine which one is chosen, depending on the needs of the radiologists at a given clinic. Both types of solutions can be used simultaneously if necessary. Nearly two-thirds (64.7%) of the features used by Moscow radiologists in vendor-specific solutions can be implemented with vendor-neutral solutions. Economic feasibility should be considered in such conditions.

ADDITIONAL INFORMATION

Supplement 1: Questionnaire for assessing the practical significance of the main functions of vendor-dependent software. doi: 10.17816/DD631519-4379064

Supplement 2: Functionality of embedded vendor-dependent and vendor-independent solutions. doi: 10.17816/DD631519-4379067

Author contributions: Yu.A. Vasilev: conceptualization; E.A. Slavushcheva: data curation, writing—original draft, review & editing; M.A. Zelenova: data curation, survey creation and implementation; I.M. Shulkin, K.M. Arzamasov: conceptualization, writing—review & editing. All the authors approved the final version of the manuscript for publication and agreed to be accountable for all aspects of the work, ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Acknowledgments: The authors would like to thank R.N. Akhmetov for consultation on the current demands of healthcare providers regarding clinical decision–support systems, and S.V. Mikhailin and D.V. Kozlov for analyzing the market for automated radiology workstations.

Ethics approval: This study was based on the data from the approved Experiment on the use of innovative computer vision technologies for analysis of medical images in the Moscow healthcare system (extract from Minutes No. 2 of the Independent Ethics Committee of the Moscow Regional Branch of the Russian Society of Radiographers and Radiologists dated February 20, 2020), registered at ClinicalTrials (NCT04489992).

Funding sources: This article was part of the research and development project “Development of a platform for improving the quality of AI services for medical diagnostics” (EGISU No. 123031400006-0) under Moscow City Health Department Order No. 1196 On the Approval of State Assignments Funded from the Budget of the City of Moscow for State Budgetary (Autonomous) Institutions Subordinate to the Moscow City Health Department for 2023 and the Planned Period of 2024–2025, dated December 21, 2022.

Disclosure of interests: The authors have no relationships, activities, or interests for the last three years related to for-profit or not-for-profit third parties whose interests may be affected by the content of the article.

Statement of originality: No previously published material (text, images, or data) was used in this study or article.

Data availability statement: All data obtained in this study are available in the article and its supplementary material. In particular, in Supplements 1 and 2.

Generative AI: No generative artificial intelligence technologies were used to prepare this article.

Provenance and peer-review: This article was submitted unsolicited and reviewed following the standard procedure. The peer review process involved two members of the Editorial Board.

 

1 Vendor-specific solutions are those offered by suppliers of diagnostic equipment, such as X-ray, mammography, CT, and MRI. Vendor-neutral solutions are offered by third-party companies that do not manufacture diagnostic equipment.

2 AI Service Catalog [Internet]. In: Center for Diagnostics and Telemedicine. Available at: https://mosmed.ai/service_catalog/ Accessed on October 24, 2024.

3 Medical Imaging Software: Catalog [Internet]. Medical Expo. Available at: https://www.medicalexpo.ru/cat/radiologia/ Accessed on August 01, 2023.

4 CAD4TB: AI-Powered Tuberculosis Detection Software [Internet]. Available at: https://delft.care/cad4tb/ Accessed on August 01, 2024.

5 Philips IntelliSpace Portal 12 [Internet]. Available at: https://www.philips.com.om/healthcare/product/ Accessed on October 30, 2024.

6 Gehealthcare [Internet]. Available at: https://www.gehealthcare.com/products/healthcare-it Accessed on October 30, 2024.

7 Siemens-healthineers [Internet]. Available at: https://www.siemens-healthineers.com/digital-health-solutions/ Accessed on October 30, 2024.

8 Canon [Internet]. Available at: https://us.medical.canon/specialties/medical-oncology-solutions/ Accessed on October 30, 2024.

9 Hologic [Internet]. Available at: https://www.hologic.com/hologic-products/breast-health-solutions/ Accessed on October 30, 2024.

10 Aidoc [Internet]. Available at: https://www.aidoc.com/solutions/#aipowered/ Accessed on October 30, 2024.

11 RADinfo SYSTEMS [Internet]. Available at: https://radinfosystems.com/products.html Accessed on October 30, 2024.

12 Sectra Medical [Internet]. Available at: https://medical.sectra.com/product/sectra-radiology-pacs-ris/ Accessed on October 30, 2024.

13 Visage Imaging [Internet]. Available at: https://visageimaging.com/ Accessed on October 30, 2024.

14 TeraRecon [Internet]. Available at: https://www.terarecon.com/ Accessed on October 30, 2024.

15 Agfa HealthCare [Internet]. Available at: https://www.agfa.com/he/russia/ru/internet/main/ Accessed on October 30, 2024.

16 Synapse [Internet]. Available at: https://synapse-emea.fujifilm.com/discovering-synapse-workflow.html Accessed on October 30, 2024.

17 Merge PACS — American Medical Imaging [Internet] Available at: https://ami-ii.com/products/merge-pacs/ Accessed on October 30, 2024.

18 Order No. 134 of the Moscow Healthcare Department dated February 24, 2022 “On Approval of the Procedure and Conditions for conducting an Experiment on the Use of Innovative Computer Vision Technologies for Medical Image Analysis and Further Use in the Moscow Healthcare System.” Available at: https://mosmed.ai/ai/docs/ Accessed on February 10, 2024.

19 Decree No. 490 of the President of the Russian Federation dated October 10, 2019 “On the Development of Artificial Intelligence in the Russian Federation” (as amended and supplemented). Available at: https://base.garant.ru/72838946/ Accessed on February 10, 2024.

20 Order No. 560н (560n) of the Ministry of Health of the Russian Federation dated June 9, 2020, “On Approval of the Rules for Conducting Radiological Examinations.” Available at: https://base.garant.ru/74632238/ Accessed on February 10, 2024.

×

About the authors

Yuriy A. Vasilev

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies; National Medical and Surgical Center named after N.I. Pirogov

Email: VasilevYA1@zdrav.mos.ru
ORCID iD: 0000-0002-5283-5961
SPIN-code: 4458-5608

MD, Cand. Sci. (Medicine)

Russian Federation, Moscow; Moscow

Ekaterina A. Slavushcheva

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies

Email: SlavuschevaEA1@zdrav.mos.ru
ORCID iD: 0009-0009-1315-0829

MD

Russian Federation, Moscow

Maria A. Zelenova

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies

Email: ZelenovaMA@zdrav.mos.ru
ORCID iD: 0000-0001-7458-5396
SPIN-code: 3823-6872

Cand. Sci. (Biology)

Russian Federation, Moscow

Igor M. Shulkin

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies

Email: ShulkinIM@zdrav.mos.ru
ORCID iD: 0000-0002-7613-5273
SPIN-code: 5266-0618

MD, Cand. Sci. (Medicine)

Russian Federation, Moscow

Kirill M. Arzamasov

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies; MIREA — Russian Technological University

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

MD, Dr. Sci. (Medicine)

Russian Federation, Moscow; Moscow

References

  1. Shulkin IM, Vladzimirsky AV. Data-based Management in Imaging: Evaluation of the Performance of a Unified Radiological Information Service Model. Manager Zdravoochranenia. 2022;(7):68–80. doi: 10.21045/1811-0185-2022-7-68-80 EDN: IYPKHK
  2. Finakov AS. Information Technologies in Medicine. Scientific Journal of Academy. 2022;(3):63–67. EDN: WSNKRW
  3. Morozov SP, Vladzymyrskyy AV, Ledikhova NV, et al. Moscow Experiment on Computer Vision in Radiology: Involvement and Participation of Radiologists. Medical Doctor and IT. 2020;(4):14–23. doi: 10.37690/1811-0193-2020-4-14-23 EDN: VEWGXO
  4. Guo Z, Xie J, Wan Y, et al. A Review of the Current State of the Computer-Aided Diagnosis (CAD) Systems for Breast Cancer Diagnosis. Open Life Sciences. 2022;17(1):1600–1611. doi: 10.1515/biol-2022-0517 EDN: RSGRVD
  5. Fujita H. AI-based Computer-Aided Diagnosis (AI-CAD): The Latest Review to Read First. Radiological Physics and Technology. 2020;13(1):6–19. doi: 10.1007/s12194-019-00552-4 EDN: JHXXPS
  6. Mayo RC, Kent D, Sen LC, et al. Reduction of False-Positive Markings on Mammograms: a Retrospective Comparison Study Using an Artificial Intelligence-Based CAD. Journal of Digital Imaging. 2019;32(4):618–624. doi: 10.1007/s10278-018-0168-6 EDN: BHBLVP
  7. Kohli A, Jha S. Why CAD Failed in Mammography. Journal of the American College of Radiology. 2018;15(3):535–537. doi: 10.1016/j.jacr.2017.12.029
  8. Lehman CD, Wellman RD, Buist DSM, et al. Diagnostic Accuracy of Digital Screening Mammography With and Without Computer-Aided Detection. JAMA Internal Medicine. 2015;175(11):1828. doi: 10.1001/jamainternmed.2015.5231
  9. Fenton JJ, Taplin SH, Carney PA, et al. Influence of Computer-Aided Detection on Performance of Screening Mammography. New England Journal of Medicine. 2007;356(14):1399–1409. doi: 10.1056/NEJMoa066099
  10. Gilbert FJ, Astley SM, Gillan MGC, et al. Single Reading with Computer-Aided Detection for Screening Mammography. New England Journal of Medicine. 2008;359(16):1675–1684. doi: 10.1056/NEJMoa0803545
  11. Tchou PM, Haygood TM, Atkinson EN, et al. Interpretation Time of Computer-aided Detection at Screening Mammography. Radiology. 2010;257(1):40–46. doi: 10.1148/radiol.10092170
  12. Philpotts LE. Can Computer-aided Detection Be Detrimental to Mammographic Interpretation? Radiology. 2009;253(1):17–22. doi: 10.1148/radiol.2531090689
  13. Khanna NN, Maindarkar MA, Viswanathan V, et al. Economics of Artificial Intelligence in Healthcare: Diagnosis vs. Treatment. Healthcare. 2022;10(12):2493. doi: 10.3390/healthcare10122493 EDN: UDQHMZ
  14. Xing X, Zhao X, Wei H, Li Y. Diagnostic Accuracy of Different Computer-Aided Diagnostic Systems for Prostate Cancer Based on Magnetic Resonance Imaging. Medicine. 2021;100(3):e23817. doi: 10.1097/md.0000000000023817 EDN: EUAPKQ
  15. Hadjiiski L, Cha K, Chan HP, et al. AAPM Task Group Report 273: Recommendations on Best Practices for AI and Machine Learning for Computer-Aided Diagnosis in Medical Imaging. Medical Physics. 2023;50(2):e1–e24. doi: 10.1002/mp.16188 EDN: RFUOFA
  16. Roberts M, Driggs D, Thorpe M, et al; AIX-COVNET. Common Pitfalls and Recommendations for Using Machine Learning to Detect and Prognosticate for COVID-19 Using Chest Radiographs and CT Scans. Nature Machine Intelligence. 2021;3(3):199–217. doi: 10.1038/s42256-021-00307-0 EDN: YFUJIA
  17. Gu Y, Chi J, Liu J, et al. A Survey of Computer-Aided Diagnosis of Lung Nodules from CT Scans Using Deep Learning. Computers in Biology and Medicine. 2021;137:104806. doi: 10.1016/j.compbiomed.2021.104806 EDN: LPGPQM
  18. Aggarwal R, Sounderajah V, Martin G, et al. Diagnostic Accuracy of Deep Learning in Medical Imaging: A Systematic Review and Meta-analysis. NPJ Digital Medicine. 2021;4(1):65–65. doi: 10.1038/s41746-021-00438-z EDN: RYOQDM
  19. Vladzimirskyy AV, Vasilev YuA, Arzamasov KM, et al. Computer Vision in Radiation Diagnostics: The First Stage of the Moscow Experiment. Moscow: Izdatel'skie resheniya; 2023. ISBN: 978-5-0059-3043-9 (In Russ.) EDN: FOYLXK
  20. Morozov SP, Vladzymyrskyy AV, Shulkin IM, et al. Feasibility of Using Artificial Intelligence in Radiology (First Year of Moscow Experiment on Computer Vision). Medical Doctor and IT. 2022;(1):12–29. doi: 10.25881/18110193_2022_1_12 EDN: UCHWWP
  21. Vasilev YuA, Vladzymyrskyy AV, Bondarchuk DV, et al. Importance of Artificial Intelligence Technologies to Prevent Defects in Radiologist’s Practice. Medical Doctor and IT. 2023;(2):16–27. doi: 10.25881/18110193_2023_2_16 EDN: SYZAOQ
  22. Vasilev YuA, Vladzimirskyy 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
  23. Kim DW, Jang HY, Kim KW, et al. Design Characteristics of Studies Reporting the Performance of Artificial Intelligence Algorithms for Diagnostic Analysis of Medical Images: Results from Recently Published Papers. Korean Journal of Radiology. 2019;20(3):405–410. doi: 10.3348/kjr.2019.0025
  24. Nam JG, Hwang EJ, Kim J, et al. AI Improves Nodule Detection on Chest Radiographs in a Health Screening Population: A Randomized Controlled Trial. Radiology. 2023;307(2):e221894. doi: 10.1148/radiol.221894 EDN: ZXSHYF
  25. Yu X, Wang J, Hong QQ, et al. Transfer Learning for Medical Images Analyses: A survey. Neurocomputing. 2022;489:230–254. doi: 10.1016/j.neucom.2021.08.159 EDN: NQEUQB
  26. Vladzymyrskyy AV, Kudryavtsev ND, Kozhikhina DD, et al. Effectiveness of Using Artificial Intelligence Technologies for Dual Descriptions of the Results of Preventive Lung Examinations. Russian Journal of Preventive Medicine. 2022;25(7):7–15. doi: 10.17116/profmed2022250717 EDN: JNUMFN
  27. Morozov SP, Abuladze LR, Andreychenko AE, et al. Basic Recommendations for the Operation of Artificial Intelligence Services for Radiation Diagnostics. Moscow: Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies; 2022. (In Russ.) EDN: EHQKPW

Supplementary files

Supplementary Files
Action
1. JATS XML
Download
2. Supplement 1: Questionnaire for assessing the practical significance of the main functions of vendor-dependent software.
Download (28KB)
Indexing metadata
3. Supplement 2: Functionality of embedded vendor‑dependent and vendor-independent solutions.
Download (37KB)
Indexing metadata
4. Fig. 1. The number of features depending on their implementation in vendor-specific and vendor-neutral solutions.

Download (97KB)
Indexing metadata
5. Fig. 2. The practical value of key features of vendor-specific software solutions based on a survey of radiologists: results are presented as a percentage of the total number of respondents.

Download (322KB)
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 - 79539 от 09 ноября 2020 г.