AI-powered medical appointment and pre-diagnosis kiosk system.
A research and development project conducted at Tyumen Industrial University (TIU) — designed to revolutionize patient–clinic interaction through neural network–driven triage, intelligent scheduling, and offline-first kiosk infrastructure.
⚠️ Important Notice
The full source code of this project is the intellectual property of Tyumen Industrial University (TIU) and is not publicly available.
This repository is a technical showcase and portfolio document describing the project's goals, architecture, methodology, and my individual contribution.
No proprietary code, datasets, or trained models are published here.
- Project Overview
- Problem Statement
- Goals and Objectives
- Solution
- System Architecture
- Functional Modules
- Key Performance Indicators
- Competitive Analysis
- Target Audience
- Market Analysis
- My Role and Contribution
- Security Implementation
- Technology Stack
- Team
- Project Context
- Note on Source Code
IMed Helix is an integrated software–hardware solution that simplifies the process of booking appointments and visiting medical specialists. The system combines a self-service kiosk, a neural network–based pre-diagnosis engine, and a real-time clinic optimization layer.
The product is designed to reduce administrative load on medical staff, eliminate long queues, and provide patients with a fast, intuitive, and personalized experience — all while operating fully offline within the clinic's local infrastructure.
The project was developed as part of the university research initiative at Tyumen Industrial University (Тюменский индустриальный университет, ТИУ) during the 2025 academic cycle.
Even in regions with high digitization, patients continue to experience friction when scheduling and visiting doctors.
Quantitative findings:
| Indicator | Value |
|---|---|
| Patients in major clinics and medical centers (sample) | 20,000+ |
| Patients reporting dissatisfaction with current booking systems | 55% |
| Average time doctors spend per patient (incl. admin tasks) | 15 min |
Regional digitization rates (Russia):
| Region | Digitization |
|---|---|
| Saint Petersburg | 75% |
| Moscow / Republic of Buryatia | ~49% |
| Krasnodar Krai | 48% |
| Sverdlovsk Oblast / Tyumen | 45% |
Core pain points:
- Long appointment booking times (2–3 minutes on existing platforms).
- Patients are routed to the wrong specialist due to lack of pre-screening.
- Doctors lose ~30% of their consultation time on administrative work.
- Existing systems require stable internet — unreliable in remote regions.
- Limited integration between booking platforms and clinic-internal Medical Information Systems (MIS).
Revolutionize the medical appointment process by making it fast, intuitive, and efficient through the application of artificial intelligence — saving time for both patients and medical staff.
- Optimize patient flow to reduce queues and bottlenecks at registration.
- Implement AI-driven pre-diagnosis to route patients to the correct specialist.
- Radically simplify the booking process down to a 30-second self-service interaction.
- Increase the operational efficiency of medical institutions through schedule optimization and predictive analytics.
IMed Helix is a self-service kiosk deployed at the entrance of a medical institution. The patient interacts with a touchscreen interface, authenticates via a personal QR code, describes their symptoms, and receives an automatic appointment with the most relevant specialist — all in approximately 30 seconds.
In parallel, the system continuously learns from anonymized usage patterns to:
- Predict daily and weekly patient load.
- Redistribute doctors' schedules dynamically.
- Personalize the booking experience based on each patient's history.
- Provide administrators with real-time analytics dashboards.
┌────────────────────────────┐
│ AI / ML Inference Layer │
│ (PyTorch, scikit-learn) │
└─────────────┬──────────────┘
│
┌────────────┐ ┌───────────────▼────────────────┐ ┌───────────────┐
│ Patient | ── | Intelligent Booking │ ── | Doctor / |
│ (Kiosk) │ │ Module │ │ Specialist │
└────────────┘ └───────────────┬────────────────┘ └───────────────┘
│
┌────────────────────────────┼────────────────────────────┐
▼ ▼ ▼
┌────────────────┐ ┌──────────────────┐ ┌────────────────────┐
│ Schedule │ │ Patient History │ │ Notification & │
│ Optimizer │ │ & Preferences │ │ Reminder Service │
└────────────────┘ └──────────────────┘ └────────────────────┘
│ │ │
└────────────────────────────┼────────────────────────────┘
▼
┌────────────────────────────┐
│ REST / WebSocket │
│ API │
└─────────────┬──────────────┘
│
┌─────────────▼──────────────┐
│ Clinic MIS Integration │
│ (Electronic Records, EHR, │
│ Insurance Systems) │
└────────────────────────────┘
| # | Module | Description |
|---|---|---|
| 1 | Intelligent Booking Module | Analyzes patient-described symptoms via NLP + classifier and routes to the correct specialist. |
| 2 | AI Pre-Diagnosis Engine | Neural network providing probabilistic diagnosis suggestions to assist the doctor. |
| 3 | Preference & History Analyzer | Personalizes the booking experience based on the patient's previous visits and preferences. |
| 4 | Patient Flow Predictor | Forecasts patient load to help administrators allocate resources proactively. |
| 5 | Notification & Reminder Service | Sends appointment confirmations and reminders via SMS / Telegram / email. |
| 6 | Schedule Optimizer | Dynamically redistributes doctor availability based on real-time demand. |
| 7 | Analytics Dashboard | Provides administrators and physicians with KPIs and operational insights. |
| 8 | Feedback Processing Module | Collects and analyzes patient feedback for continuous service improvement. |
| 9 | MIS Integration API | Bridges the kiosk system with the clinic's existing Medical Information System. |
All modules communicate through an internal asynchronous event bus, allowing the system to function as a single, cohesive organism.
| KPI | Target | Description |
|---|---|---|
| Pre-diagnosis Accuracy | 85% | Accuracy of AI-based symptom-to-specialty classification. |
| Booking Time | 30 sec | Average time required for a patient to complete an appointment via the kiosk. |
| Throughput Increase | +40% | Improvement in the number of patients a clinic can process per day. |
| Queue Wait Time | −70% | Reduction in average patient wait time in the registration queue. |
| Feature | IMed Helix | EMIAS / Gosuslugi | Microsoft AI Health | Legacy Kiosks |
|---|---|---|---|---|
| AI-based pre-diagnosis | ✅ | ❌ | ✅ | ❌ |
| Offline / local deployment | ✅ | ❌ | ❌ | ⚠ Partial |
| QR-code authentication | ✅ | ⚠ App only | ❌ | ❌ |
| Full MIS integration | ✅ | ⚠ State only | ❌ | ⚠ Basic |
| Personal data protection | ✅ (on-premise) | ✅ | ⚠ Cloud | ✅ |
| Booking time | 30 sec | 2–3 min | 1–2 min | 1–2 min |
| Intuitive interface | ✅ | ⚠ Complex nav | ✅ | ⚠ Outdated UI |
| Self-learning system | ✅ | ❌ | ✅ | ❌ |
Key differentiators: offline-first deployment, on-premise data isolation, sub-minute booking flow, and a self-improving AI core.
The product targets three interconnected user groups:
| Group | Share of Market / Users | Notes |
|---|---|---|
| Medical Institutions | 65% | Primary buyers — public clinics (40%) and private clinics (25%). |
| Medical Personnel | 20% | Physicians (12%) and administrative staff (8%). |
| Patients | 15% | Direct end-users of the kiosk interface. |
Key insight: medical institutions are the purchasers, but value is distributed across all three groups. The system exhibits a synergistic effect — the more participants use it, the higher its overall efficiency becomes.
The market size was calculated using the top-down methodology.
| Institution Type | Count |
|---|---|
| State hospitals | 5,065 |
| Private clinics | 3,692 |
| Non-commercial medical organizations | 312 |
| Total | 9,069 |
| Metric | Value |
|---|---|
| TAM (Total Addressable Market) | ₽171,395,000 |
| Competitor share | 33.9% |
| SAM (Serviceable Addressable Market) | ₽8,989,600 |
| Realistic capture share | 11% |
| SOM (Serviceable Obtainable Market) | ₽961,887 |
| Projected revenue at 10% institution coverage | ₽17,139,500 |
| Payback period | 201 days |
The 201-day payback period makes IMed Helix highly attractive from an investment perspective, especially given the recurring nature of clinic licensing.
Saveliy Golubev — Neural Network Algorithm Developer & Security Engineer.
- Designed the symptom-to-specialty classification model achieving 85% accuracy.
- Built the personalization layer that adapts the booking flow to individual patient history.
- Developed the patient flow prediction model used by the schedule optimizer.
- Engineered the full ML pipeline: dataset preparation, feature engineering, training, validation, and inference deployment.
- Integrated the inference layer with the backend API via asynchronous endpoints.
Took on the additional responsibility of designing the system's security architecture. See the next section for detail.
Given that IMed Helix processes sensitive medical data, security was treated as a first-class architectural concern. As the security engineer, I implemented:
- Patient data encryption — at rest (AES-256) and in transit (TLS 1.3).
- QR-code authentication — short-lived signed tokens with session-bound nonces to prevent replay attacks.
- Input validation & sanitization — strict schema validation on every API endpoint to mitigate injection vectors.
- Role-Based Access Control (RBAC) — separate permission tiers for patients, medical staff, and administrators.
- Rate limiting — adaptive throttling on authentication and booking endpoints to mitigate brute-force and abuse.
- Secure API design — JWT-based session management, CSRF protections, and full security header coverage (CSP, HSTS, X-Frame-Options).
- HIPAA-aligned data handling — minimal data retention, audit logging of all access to patient records, and a clear data lifecycle policy.
- Local-first architecture — sensitive data never leaves the clinic's local network unless explicitly authorized.
| Layer | Technologies |
|---|---|
| Language | Python 3.11+ |
| Backend | FastAPI, asyncio, WebSocket |
| AI / ML | PyTorch, scikit-learn, pandas, NumPy |
| Frontend (Kiosk UI) | React, TypeScript |
| Database | PostgreSQL (clinic data), SQLite (kiosk-local cache) |
| Authentication | JWT, signed QR tokens |
| Deployment | Docker, on-premise local servers |
| Integration | REST API, HL7/FHIR-compatible adapters for MIS |
| Member | Role | Responsibilities |
|---|---|---|
| Polina Goglacheva | Project Manager | Team coordination and project execution |
| Saveliy Golubev | Neural Network Developer & Security Engineer | AI models, ML pipeline, system security |
| Lavr Albychev | Backend Developer | Server-side architecture, API, data persistence |
| Eva Sklyarenko | UI/UX Designer | Kiosk interface, user-experience design |
| Sofya Sokolova | Data Analyst | Data collection, processing, and market research |
- Institution: Tyumen Industrial University (Тюменский индустриальный университет, ТИУ)
- Course: ПР7 — Programming in Python
- Project type: Research & Development (R&D)
- Year: 2025
- Format: Cross-functional team of five
- Outcome: Working prototype, full project documentation, market analysis, and presentation to the academic review board.
The full source code of IMed Helix is the intellectual property of Tyumen Industrial University and is therefore not published in this repository. This repository serves as a public technical showcase and portfolio document describing the project's goals, architecture, methodology, and the author's individual contribution.
For inquiries regarding access to source materials, please contact the university directly.
Maintained by @NovaCode37 — Saveliy Golubev.