Experiment on the Use of Innovative Computer Vision Technologies for Analysis of Medical Images in the Moscow Healthcare System

It is planned to integrate various services based on computer vision technologies for analysis of the certain type of x-ray study into Moscow Unified Radiological Information Service (hereinafter referred to as URIS).

As a result of using computer vision-based services, it is expected:

1. Reducing the number of false negative and false positive diagnoses;
2. Reducing the time between conducting a study and obtaining a report by the referring physician;
3. Increasing the average number of radiology reports provided by a radiologist per shift.

Study Overview

• Status: Recruiting
• Conditions: COVID-19; Fractures, Bone; Lung Cancer; Chest--Diseases; Brain Disease; Mammary Cancer; AI (Artificial Intelligence); Abdomen Disease; X-Rays; Lesion

Detailed Description

Recently a growth in the number of radiology studies across multiple modalities has been observed alongside the modest increase in staffing levels. This carries higher risks of increased workload and efficiency losses. The integration of computer vision-based services into URIS will improve the radiologists' productivity and job performance.

Existing prerequisites for conducting the study:

1. Increasing the number of preventive and diagnostic radiological studies entails the growing workload for radiologists and increased risk of interpretation errors, which in turn leads to the decrease in quality of medical care.
2. When a radiologist opens a worklist of studies, in the absence of special notes, he/she writes a report in the random order, not being able to select from the list the studies that require the most attention and prompt response (studies with pathological findings), which increases the time of diagnosis.
3. The absence of the structured pre-filled template of report leads to the increase in time for preparing reports.
4. A radiologist has to spend considerable time evaluating the dynamics of pathological changes, which also increases the time to prepare a report as well as the risk of error.
5. Interpretation of preventive studies requires double reading, which is implemented inefficiently due to the staff shortage.

Study objectives:

1. Study the diagnostic accuracy of the Services in accordance with the methodological guidelines No. 43 "Clinical trials of software based on intelligent technologies (diagnostic radiology)" (recommended by the Expert Council on Science of the Moscow Healthcare Department, Protocol No. 8 of June 25, 2019).
2. Audit the studies conducted with Services application in order to determine the number of interpretation errors, and compare it with the audit result without their application (hypothesis 1).
3. Conduct timekeeping to estimate time for preparing a report and the total number of evaluated studies with and without using the Services (hypothesis 2,3).
4. Conduct a survey of radiologists who use the Services in their work, in order to determine their opinion about the implementation of innovative technologies in the diagnostic process.

METHODOLOGY

1. The Experiment is carried out by the Moscow Healthcare Department in accordance with Regulation No. 43 of January 24, 2020 "On approval of the procedure and conditions for conducting the Experiment on the use of innovative computer vision technologies for analysis of medical images and further application in Moscow healthcare system".
2. The experiment is conducted on the next types of studies:

1)Detection of CT signs consistent with COVID-19 (coronavirus) lung involvement (Chest CT); 2) Emphysema extent (Chest CT); 3) Detection of CT signs consistent with malignant neoplasm in the lungs (Chest CT); 4) Detection of LDCT signs consistent with malignant neoplasm in the lungs (Chest LDCT); 5) Detection and localization of compression vertebral fractures with a degree of vertebral body deformity of over 25% according to the Genant semi-quantitative scale, grades 2-3 (Chest CT); 6) Detection of free pleural fluid (effusion) (Chest CT); 7) Detection of enlarged intrathoracic lymph nodes (lymphadenopathy) (Chest CT); 8) Detection of bronchiectasis (Chest CT); 9) Detection of CT signs consistent with pulmonary tuberculosis (Chest CT); 10) Coronary calcium score (Chest CT/ LDCT); 11) Paricardial fat volume (Chest CT); 12) Dilation of ascending and descending thoracic aortas (Chest CT/ LDCT); 13) Dilation of the pulmonary trunk (Chest CT/ LDCT); 14) Detection of sarcoidosis (Chest CT); 15) Detection of signs consistent with the impairment of lung airness (Chest CT); 16) Detection of signs consistent with the focal lesions in the chest bones (Chest CT); 17) Detection of CT signs consistent with rib fracture (Chest CT); 18) Detection of signs of urolithiasis (Abdominal CT); 19) Detection of signs consistent with the focal lesions in the skeleton bones (Abdominal CT); 20) Detection of liver lesions (Abdominal CT); 21) Detection of CT signs consistent with gallbladder stones (Abdominal CT); 22) Detection of CT signs consistent with renal lesions (Abdominal CT); 23) Measuring the abdominal aorta dilation (Abdominal CT); 24) Detection of adrenal lesions (Abdominal/Chest CT); 25) Detection and localization of compression vertebral fractures with a degree of vertebral body deformity of over 25% according to the Genant semi-quantitative scale, grades 2-3 (Abdominal CT); 26) Automation of routine liver measurements (dimensions, liver density, choledochus diameter, portal vein diameter) (Abdominal CT); 27) Automation of routine kidney measurements (kidney size, pelvicalyceal system size) (Abdominal CT); 28) Automation of routine measurements of spleen and pancreas (size, density of the spleen and pancreas) (Abdominal CT); 29) Detection of acute ischemic stroke and its ASPECTS score (Head CT); 30) Detection of hemorrhage and its automatic volume calculation in ml or cm³ (Head CT); 31) Automation of routine measurements (ventriculometry, displacement of median structures, measurement of the craniovertebral junction) (Head CT); 32) Detection and localization of (at least 7) signs consistent with the priority disease (Chest XR); 33) Detection of signs (at least one) consistent with bone fracture (MSS XR); 34) Detection of radiologic signs (at least one) consistent with arthrosis of the joints (MSS XR); 35) Detection of radiological signs (at least one) consistent with deforming arthrosis of the hip (MSS XR); 36) Detection of radiological signs (at least one) consistent with the fracture of the shoulder joint bones (MSS XR); 37) Detection of radiological signs (at least one) consistent with the fracture of the hip joint bones (MSS XR) 38) Detection of radiological signs (at least one) consistent with the fracture of the ankle joint bones (MSS XR).

39) Detection of reduced pneumatization / opacity of the paranasal sinuses (Head XR) 40) Detection of signs (at least one) consistent with transverse flat foot (MSS XR) 41) Detection of signs (at least one) consistent with the longitudinal flat foot in the lateral plane (MSS XR); 42) Detection of the signs of osteoporosis: detection and localization of compression vertebral fractures with a degree of height loss of over 25% as well as the radio density measurements of vertebral bodies (Spine XR); 43) Detection of signs consistent with osteochondrosis in the frontal and/or sagittal plane (Spine XR); 44) Detection of signs consistent with scoliosis in the frontal plane (Spine XR); 45) Detection of signs consistent with spondylolisthesis in the sagittal plane (Spine XR); 46) Detection and localization of findings consistent with breast cancer (MMG); 47) Detection of multiple sclerosis (Brain MRI); 48) Detection and localization of intracranial neoplasms (extracerebral, intracerebral) (Brain MRI); 49) Automation of routine measurements (ventriculometry, displacement of median structures, measurement of the craniovertebral junction, changes in white matter, intracranial measurements) (Brain MRI); 50) Detection of signs consistent with the focal lesions in the cervical spinal cord (Cervical spine MRI); 51) Detection and localization of MRI signs (at least one) consistent with degenerative changes in the cervical discs on sagittal and axial T2-WI (Cervical spine MRI); 52) Detection and localization of MRI signs (at least one) consistent with degenerative changes in the thoracic discs on sagittal and axial T2-WI (Thoracic spine MRI); 53) Detection of signs consistent with the focal lesions in the thoracic spinal cord (Thoracic spine MRI); 54) Detection and localization of MRI signs (at least one) consistent with degenerative changes in the lumbosacral discs on sagittal and axial T2-WI (Lumbosacral spine MRI); 55) Detection of signs consistent with the focal lesions in the lumbosacral spinal cord (Lumbosacral spine MRI); 56) Detecting signs consistent with the areas of cartilage breakdown (chondromalacia) along the articular surfaces of the knee and the patellofemoral joint (Knee joint MRI); 57) Automated routine measurements of the prostate gland (dimensions) (Lesser pelvis MRI); 58) Automated routine measurements of the uterus (corpus and cervix: position, dimensions, displacements) (Lesser pelvis MRI).

3. For each Service during the Experiment, a certain number of studies is provided for processing based on their type:

1. CT/LDCT - 30 250 studies;
2. XR - 55 000 studies;
3. MMG - 48 500 studies;
4. MRI - 22 500 studies.

4. A methodology for including services in the Experiment has been developed. For each Service, the participation process in the Experiment consists of the following stages:

1. selection;
2. the preparatory stage;
3. the main stage;
4. the final stage.

During the Experiment, a radiologist will routinely be able to:

• work on a sorted list of patients (triage);
• work with images processed by the Service;
• work with a pre-filled template of the radiological report on each study;
• evaluate the work of the Service according to the developed questionnaire. During the Experiment, a patient will receive the individual plan of the follow-up support. It includes preventive examinations or observation as well as treatment by a specialist.

Systematization and final analysis of the Experiment results is carried out within three months from the completion date of the last Service participation in the Experiment.

Based on the results of the Experiment, recommendations can be prepared on the possibility to register certain services as a medical device (software).

• Study Type: Observational
• Enrollment (Estimated): 133000

Contacts and Locations

• Study Contact: Name: Kirill Arzamasov, PhD; Phone Number: +79152514838; Email: ArzamasovKM@zdrav.mos.ru
• Study Contact Backup: Name: Olga Omelyanskaya; Phone Number: +79263948149; Email: OmelyanskayaOV@zdrav.mos.ru

Study Locations

• Russian Federation
  • Moscow, Russian Federation
    • Recruiting
    • Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies of the Moscow Health Care Department
    • Contact: Kirill Arzamasov, PhD; Phone Number: +79152514838; Email: ArzamasovKM@zdrav.mos.ru
    • Contact: PhD

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: 18 years and older (Adult, Older Adult)
• Accepts Healthy Volunteers: Yes

• Sampling Method: Non-Probability Sample

Study Population

patients over the age of 18 attending outpatient clinics

Description

Inclusion Criteria:

• Age (over 18 years)
• Gender (male and female)
• Referral for the study
• Signed informed consent to participate in the Experiment
• Chest computed tomography and Low-dose computed tomography for lung cancer detection or mammography for breast cancer detection or chest X-ray for lung pathology detection

Exclusion Criteria:

• Another type of study (including a different modality and anatomical area)

Study Plan

How is the study designed?

Number of groups / cohorts

2

Cohorts and Interventions

Group / Cohort
Standard radiology studies with AI; The experiment is conducted on 10 types of studies with AI: Chest CT/ LDCT with different pathologies;; Abdominal CT with different pathologies;; Head CT with different pathologies;; MSS XR with different fractures; Spine XR with different pathologies;; MMG;; Brain MRI with different pathologies;; Cervical spine MRI, Lumbosacral spine MR and Thoracic spine MRI with spine pathologies; Knee joint MRI; Lesser pelvis MRI.
Standard radiology studies without AI; The experiment is conducted on 10 types of studies without AI: Chest CT/ LDCT with different pathologies;; Abdominal CT with different pathologies;; Head CT with different pathologies;; MSS XR with different fractures; Spine XR with different pathologies;; MMG;; Brain MRI with different pathologies;; Cervical spine MRI, Lumbosacral spine MR and Thoracic spine MRI with spine pathologies; Knee joint MRI; Lesser pelvis MRI.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Number of errors	Change of at least 30% in the number of errors in interpretation of the studies with using computer vision-based services compared to the number of errors in interpretation without their application.	Upon completion, up to 4 years

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Report turnaround time	Change of at least 30% of the time from the study completion to report finalization by a radiologist.	Upon completion, up to 3 year
Number of reports	Change of at least 30% in the number of radiology reports provided by a radiologist per shift.	Upon completion, up to 4 years
Change in the errors of services per the feedback form	Change of at least 30% in computer vision-based services errors as per integrated feedback form for radiologists in the PACS. Types of errors: Technological defect (absent AI-generated series, partially generated AI series, DICOM SR and images mismatch, multiple conflicting results); Major discrepancy (findings outside the region of interest, irrelevant findings); Inaccurate diagnosis; Inaccurate lesion localisation; Inaccurate lesion classification; Other (free-text field)	Upon completion, up to 4 years

Collaborators and Investigators

• Sponsor: Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies of the Moscow Health Care Department
• Investigators: Study Director: Anton Vladzymyrskyy, Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies of the Moscow Health Care Department

Publications and helpful links

Helpful Links

• Experiment on the use of innovative computer vision technologies for analysis of medical images in the Moscow healthcare system

Study record dates

Study Major Dates

• Study Start (Actual): February 21, 2020
• Primary Completion (Estimated): January 1, 2024
• Study Completion (Estimated): January 1, 2024

Study Registration Dates

• First Submitted: July 17, 2020
• First Submitted That Met QC Criteria: July 24, 2020
• First Posted (Actual): July 28, 2020

Study Record Updates

• Last Update Posted (Actual): May 26, 2023
• Last Update Submitted That Met QC Criteria: May 25, 2023
• Last Verified: May 1, 2023

More Information

Terms related to this study

• Keywords: Computed tomography; COVID-19; Brain Disease; mammography; Fractures, Bone; Low-dose CT; AI (Artificial Intelligence); X-ray chest; Chest--Diseases; Abdomen Disease
• Additional Relevant MeSH Terms: Central Nervous System Diseases; Nervous System Diseases; Skin Diseases; Coronavirus Infections; Coronaviridae Infections; Nidovirales Infections; RNA Virus Infections; Virus Diseases; Infections; Respiratory Tract Infections; Respiratory Tract Diseases; Neoplasms; Pneumonia, Viral; Pneumonia; Lung Diseases; Neoplasms by Site; Wounds and Injuries; Breast Diseases; COVID-19; Breast Neoplasms; Brain Diseases; Fractures, Bone
• Other Study ID Numbers: 2020-3

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: NO

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No