Social context
The project is associated with a charitable foundation to help people with Spina Bifida, where early attention to possible signs of pathology, correct family routing and careful information support are important. For RESTART, this is an example of using AI not for the sake of demonstrating the technology, but for the sake of practical benefits for doctors, families, NGOs and the expert community.
The public AI verification project is available on the website spinabifidacheck.ru, and information about the fund is on helpspinabifida.ru.
Task
It was necessary to create a digital tool that helps analyze ultrasound images of the fetus and highlight areas that require specialist attention. An important limitation: the service should not look like an independent diagnostic system and should not replace a doctor.
The correct role of AI in the project is an additional support tool: image pre-processing, search for possible visual features, structuring the result and assistance in choosing the next step of the consultation.
What does the service do?
Uploading an image
The user or technician can submit an ultrasound image for initial AI verification within a specified scenario.
AI analysis
The computer vision model helps to look for visual signs that may be associated with Spina Bifida and highlight potentially significant areas.
Preliminary result
The service generates an information result for the doctor or user with an emphasis on the need for professional interpretation.
Routing
The user receives direction to profile information, fund, specialists or additional materials, if provided for in the scenario.
Technologies
| Layer | What is used |
|---|---|
| AI / ML | Computer vision, classification and detection of visual features, segmentation, highlighting of potentially significant areas. |
| ML pipeline | Image preparation, data quality control, validation, model versioning, human-in-the-loop approach. |
| Backend | API for loading an image, getting the result, logging events and integrating with the public interface of the project. |
| Safety | Data minimization, differentiation of rights, secure transfer, control of access to images and work with anonymized datasets where possible. |
Medical disclaimer
The AI service does not make a diagnosis and does not replace an ultrasound doctor, geneticist, neurosurgeon or other medical specialist. It helps to pay attention to possible signs, structure information and quickly make an informed decision on further routing. The final medical decision is always made by a qualified physician.
Architecture
Users
families and patientsmedical specialistsfund expertsNPO employeesChannels
public project websiteimage upload formchatbot / Telegramappeal to the fundAI environment
computer visionRAG assistantproven knowledge basehuman-in-the-loopSafety
data minimizationdepersonalizationaccess roleslogging and auditingThe diagram is impersonal and shows the basic architecture. The medical specialist and subject matter expert remain at the center of decision making; AI does not replace a doctor and does not form a final diagnosis.
Platform framework RESTART AI Platform
The project was developed using the RESTART AI Enterprise Platform: the RESTART platform AI environment for application products that require data sources, access roles, logging, human-in-the-loop, secure processing of sensitive information and managed AI scenarios.
In this project, the platform is used as the basis for computer vision, API integrations, data quality control, result routing and linking with a proven knowledge base. The same platform approach is used in other social and personal AI projects of RESTART, including the sports AI trainer 1trAIner.
Infrastructure for Social AI and MedTech
Social AI/MedTech projects require careful handling of images, sensitive data, roles, logs, and human review. AI Compute helps to transfer this approach to enterprise landscapes: calculations, storage, backup, access restrictions and safe operation rules are laid down before launch.
Let's discuss your environment
Describe the task, current systems, constraints, and expected results. We will offer a practical first step: diagnostics, pilot, audit, roadmap or project team.
