Mammography
Identifies breast cancer signs in screening and diagnostic exams
Unified Register of Russian Software
Included by Order No. 983 of the Ministry of Digital Development dated 20.09.2021
500+ medical institutions
Have implemented the “Third Opinion” AI service
SaMD
Class III risk category
AI Algorithm detects and describes
After AI processing
Physician receives a pre-filled report
Example Report:
Image Quality (PGMI Scale): G (Good)
Right Breast:
Radiological Density (ACR Scale): A
Skin: No abnormalities
Benign calcifications: Detected
Suspicious calcifications: Detected
1st cluster: 25 x 21 mm
2nd cluster: 14 x 10 mm
Calcification Distribution:
Within mass structures
Masses: Detected – 36 x 26 mm
Quadrant: Upper inner
Architectural distortion: Not detected
Axillary lymph nodes: Present, no abnormalities
Nipple retraction: Not detected
Density asymmetry: Not detected
Tissue edema: Not detected
Conclusion:
Right breast: BI-RADS 2 (Benign findings)
Image Quality (PGMI Scale): G (Good)
Right Breast:
Radiological Density (ACR Scale): A
Skin: No abnormalities
Benign calcifications: Detected
Suspicious calcifications: Detected
1st cluster: 25 x 21 mm
2nd cluster: 14 x 10 mm
Calcification Distribution:
Within mass structures
Masses: Detected – 36 x 26 mm
Quadrant: Upper inner
Architectural distortion: Not detected
Axillary lymph nodes: Present, no abnormalities
Nipple retraction: Not detected
Density asymmetry: Not detected
Tissue edema: Not detected
Conclusion:
Right breast: BI-RADS 2 (Benign findings)
Evaluates studies using standardized scales
Determines malignancy risk
Assesses breast tissue structure and density
BI-RADS
Assesses positioning quality
ACR
PGMI
Under 60 seconds
Processing time per study
Experiment Leader
Highest ROC AUC score in the Maturity Matrix
1 000 000+
Mammograms processed by AI in 2024
Use case scenarious
Retrospective analysis
Processes study archives and detects missed pathologies. Provides analytics at the facility or regional level
Analyzes the study, highlights pathological signs, and generates a pre-filled report
Normal case sorting
Distinguishes between "normal" and "pathology," prioritizing mammograms with abnormalities at the top of the list. Optimizes the specialist's workflow and helps focus on suspicious findings
Radiologic technologists assistant
Evaluates breast positioning and identifies errors during the examination. Allows the technician to retake a poor-quality image on the spot without requiring the patient to return
Radiologist assistant
Speeds Up Diagnosis by 57%
• Based on research by State Institution of Healthcare of the Moscow City " Scientific and Practical Clinical Center for Diagnostics and Telemedicine Technologies of the Moscow Department of Health".
Doctor
250
200
150
100
50
0
101 h
Third Opinion + Doctor
Time to describe 1,000 studies
240 h
Reduces reporting time by 2x when replacing the first read*
DICOM Image
AI processes the study
Integration Workflow
Returns
a secondary DICOM series with AI-visualization
a pre-filled report
a pre-filled report
Workstation
Device
Third Opinion
Возвращается дополнительная серия с ИИ-визуализацией и предварительно заполненный протокол
PACS
Transforming healthcare with partners
Let us help you integrate the AI service into your medical facility
FAQ
Yes. It holds a Class III medical device registration certificate, is listed in the Unified Software Register, and is included in the Roszdravnadzor recommended solutions list
It can be deployed on-premise or in the cloud, with automatic data transfer and result return — in full compliance with Federal Laws 152-FZ and 323-FZ
The server is provided by the client or the system integrator
We configure the transfer of studies from the PACS server to the AI server using standard PACS tools.
We work with all PACS that support DICOM v3.0.
We work with all PACS that support DICOM v3.0.
Compatible with any viewer or radiologist workstation that supports DICOM v3.0 and has access to the facility’s image storage/archive or a centralized subsystem
When deployed on-premise, data never leaves the facility — anonymization isn’t required.
When using a cloud setup, anonymization is performed by the client using specialized medical data anonymization software.
When using a cloud setup, anonymization is performed by the client using specialized medical data anonymization software.
• Technical meeting with PACS providers and project owners to design the integration architecture
• Clinical meeting with physicians and medical directors to define success metrics and assign roles
• Results are compiled into a formal pilot report based on agreed methodology
• Clinical meeting with physicians and medical directors to define success metrics and assign roles
• Results are compiled into a formal pilot report based on agreed methodology
Depends on system load and hardware, but typically up to 60 seconds after the image is received
• Secondary DICOM series (DICOM SC)
• Pre-filled structured report (DICOM SR), which can be edited, copied, or automatically sent to a medical information system
• Automatic submission of AI opinion to the Unified State Health Information System as a Standardized Electronic Medical Record with a physician’s signature, via integration with RIS and MIS
• Pre-filled structured report (DICOM SR), which can be edited, copied, or automatically sent to a medical information system
• Automatic submission of AI opinion to the Unified State Health Information System as a Standardized Electronic Medical Record with a physician’s signature, via integration with RIS and MIS
Yes. The service works within the facility’s internal network, receiving studies from PACS and returning results to PACS
Improve healthcare quality with AI
For career inquiries
Operations are carried out with grant support from the Skolkovo Foundation
For general inquiries and implementation
For marketing and PR inquiries
© 2025 Third Opinion Platform LLC, INN 7714420229, OGRN 5177746328106. All rights reserved. 121205, Moscow, Skolkovo Innovation Center, Nobel St., Building 7, Floor 2, Office 37, Workspace 2
