⚠️ For licensed healthcare professionals only. Physician verification required. Beta: LLM-dependent, verify clinically.
🏥 Doctor Opus
⚠️ Beta clinical support software. For licensed healthcare professionals only. AI outputs may be inaccurate and depend on third-party LLM capabilities. Mandatory physician verification is required before any clinical decision.

📘 Physician Manual

📘 Doctor Opus — Physician User Guide

Important: Doctor Opus is beta Clinical Decision Support Software (CDSS) for licensed healthcare professionals only. It is not a medical device and does not provide final diagnosis or treatment orders. AI outputs depend on third-party LLM capabilities and may be incomplete or inaccurate. All AI-generated outputs require independent physician verification. Intended-use restriction: not for regulated clinical deployment in EU/US/UK jurisdictions. You bear full responsibility for all clinical decisions.

Doctor Opus accelerates your clinical workflow by providing AI-assisted interpretation of medical imaging, lab data, genetic reports, and clinical notes. Each section contains contextual tips — review them on first use.

The application runs on desktop and mobile. Both can work independently or in tandem via the cross-device sync module.

📱 Install as a Mobile App (PWA)

The platform is a Progressive Web App — install it on your home screen for native-like access.

iPhone (Safari)

  1. Open doctor-opus.online in Safari
  2. Tap the Share button (square with arrow) at the bottom
  3. Scroll down and select "Add to Home Screen"
  4. Tap "Add" in the top-right corner
  5. Done — the Doctor Opus icon will appear on your home screen

Android (Chrome)

  1. Open doctor-opus.online in Chrome
  2. Tap the three-dot menu (⋮) in the top-right corner
  3. Select "Add to Home Screen" or "Install App"
  4. Confirm installation
  5. Done — the icon will appear on your home screen

Once installed, the app opens full-screen, is accessible via icon, and works even with a poor connection (except AI features that require the network).

🏠 Home

Overview dashboard with quick navigation to all sections.

⚖️ Mandatory Legal Workflow

After first login (or after legal version update), you must confirm legal acceptance before using clinical modules.

When saving an AI result to a patient record:

  • A physician verification modal appears
  • You must confirm personal review and clinical responsibility
  • Audit log stores only case identifier + confirmation/hash metadata (no patient personal details are sent)

🤖 AI Assistant

The core intelligence of the platform. Supports open-ended clinical dialogue, case discussion, differential diagnosis, literature review, and multi-file analysis.

Available models (dropdown):

  • GPT-5.6 Sol — Best for 80% of imaging, MRI, and general clinical questions. Concise and efficient.
  • Claude Opus 5.5 — Deepest reasoning. Best for complex cases, genetics, and rare pathologies. Slower, higher cost.
  • Claude Sonnet 5 — Balanced. Excellent for quick consultations and fracture assessment.
  • Gemini 3 Flash — Fastest. Ideal for quick reference and data extraction.

The assistant can:

  • Answer clinical questions and sustain multi-turn dialogue
  • Accept results exported from any other section
  • Specialize as a consultant (Cardiologist, Neurologist, Orthopedist, etc.)
  • Perform literature review and evidence-based searches
  • Process uploaded files (images, PDFs, Word documents)
  • Use your Personal Library (RAG) — when enabled, the assistant draws answers from your own uploaded PDF guidelines and references

PHI reminder: Do not enter patient names, dates of birth, or other identifying information in the chat. Use anonymized descriptions (e.g., "65 y.o. male, smoker, 3-week cough").

📚 Personal Library (RAG)

Upload your own PDF clinical guidelines, textbooks, and atlases. Once processed, the AI Assistant can search them and cite relevant excerpts directly in the analysis output.

  • Capacity: up to ~1 GB (larger collections may slow the browser on low-end hardware)
  • Files are processed on your local server — not sent to external services
  • Use text-searchable PDFs (not scans) for best results

📝 Clinical Protocol (Voice-to-Note)

Converts unstructured dictation or typed notes into a structured, specialty-specific clinical note — ready to download as a Word (.docx) file, edit, and sign.

How to use

  1. Select your specialty from the dropdown (Cardiology, Neurology, Orthopedics, etc.)
  2. Dictate or type the encounter notes in any order — the AI structures them automatically
  3. Click Generate Protocol — the formatted note appears in the right panel
  4. Download as .docx, review, and sign

Note structure follows SOAP / H&P format:

  • S — Subjective (CC, HPI, PMH, medications, allergies)
  • O — Objective (vitals, PE findings)
  • A — Assessment (working diagnosis, differential)
  • P — Plan (diagnostics, treatment, follow-up)

You can customize any template to match your workflow. The customized version can be pinned as your personal standard.

Recommended models: GPT-5.6 Sol or Claude Sonnet 5.

🗣️ Medical Translator

Open it from the home page. You speak; the patient hears the translation.

  1. Choose your language and the patient’s language.
  2. Press Start and allow the microphone. Use headphones.
  3. The right column is what the listener hears. Now the patient passes the microphone to the patient. Now the doctor passes it back. The languages stay as you set them. End closes the session.
  4. If a dose, number, or clinical term may not match, check it yourself. The text is not corrected automatically.

Cost: 5.1 credits per minute while the microphone is on. Recording a conversation for the clinical protocol is separate: 62 credits per hour.

🧮 Medical Calculators

Launches an integrated third-party calculator suite. Runs client-side — no credits consumed, no data transmitted.

📋 Clinical Guidelines

Search current international clinical guidelines by condition, syndrome, or drug class.

Search depth options:

  • Standard — concise protocol summary with key recommendations
  • Clinical Review — deep analysis: differential, scoring scales (CHADS₂, Wells, CURB-65, etc.), step-by-step management, treatment algorithms
  • Real-time Search — latest 2024–2025 publications with source links

After receiving results, you can continue the conversation with follow-up questions in context.

🔬 Specialized Analysis Modules

📈 ECG Analysis

Workflow:

  1. Upload an ECG image (JPG, PNG, or PDF scan)
  2. Add clinical context (CC, HPI, relevant medications) — significantly improves accuracy
  3. Use the 🛡️ anonymization buttons before submitting:
    • Quick: Auto-redacts edges and corners
    • Precision: Brush editor for exact redaction
  4. Select analysis mode (Fast / Optimized / Expert Validated)

Additional tools:

  • Digital Caliper: Drag blue markers to measure PR, QRS, QT intervals. Calibrate using the ECG grid (1 sec = 5 large cells at 25 mm/s)
  • Search Library: After analysis, click to find matching cases or descriptions in your personal PDF library

Recommended models: GPT-5.6 Sol (general) · Claude Sonnet 5 (arrhythmia detail)

🩻 X-Ray Analysis

Upload single or multiple images (folder or DICOM series). Add clinical context for significantly better output.

Anonymization:

  • Quick: auto-redacts standard PHI zones
  • Precision: manual brush editor
  • DICOM: metadata stripped automatically

Comparison mode: Enable Before/After to compare two time points or views side-by-side.

Best models: GPT-5.6 Sol (80% of cases) · Claude Sonnet 5 (fractures, 83% accuracy)

🧠 CT Analysis

Upload CT images or an entire DICOM folder.

3D Viewer (DICOM series):

  • MPR 2×2: Axial / Coronal / Sagittal slices + volumetric model
  • Cinematic 3D ✨: Full-screen photorealistic rendering with soft shadows
  • Clinical presets: Bone, Soft tissue (X-ray effect), Glow (highlights pathological foci)
  • Scroll slices with mouse wheel · Zoom · Free 3D rotation
  • M1 chip: hardware-accelerated rendering

PHI anonymization: manual and automatic (DICOM metadata auto-stripped).

🧠 MRI Analysis

Identical workflow to CT. Supports multi-sequence DICOM series with full MPR and Cinematic 3D rendering.

🔊 Ultrasound Analysis (Cine-loop)

Upload a static image or a video loop (cine-loop).

Frame extraction:

  • Auto-Extract: System automatically extracts 5–12 key frames
  • Manual Capture: Step through with ±0.1s buttons and capture the exact frame

All frames are anonymized before submission (black bars on edges).

🔬 Dermatoscopy Analysis

Upload dermatoscopy images. Add clinical context (lesion location, duration, observed changes). Supports ABCDE criteria analysis and malignancy risk assessment.

🧪 Laboratory Data Analysis

Upload a lab report (PDF, Excel, CSV, or photo of a paper form).

Smart extraction: System automatically recognizes parameters, values, and reference ranges — even from multi-page PDFs or handwritten forms.

Analysis output includes:

  • Critical value flagging
  • Clinical interpretation in the context of provided HPI
  • Trend graphs (if patient is in your database)

🧬 Genetic Analysis

Upload a genetic report in .VCF format (raw lab output) or PDF.

Workflow:

  1. Upload file
  2. Stage 1 (Extract): Gemini 3 Flash extracts rsIDs and genotypes from the report
  3. Stage 2 (Interpret): Claude Opus 5.5 provides clinical risk interpretation
  4. Continue dialogue with the Genetics specialist for follow-up questions

Always anonymize before submitting (name and address auto-redacted on the preview screen).

🎬 Video Clinical Analysis

Upload any video file (patient gait, endoscopy, echocardiography, ultrasound loop, etc.).

Two modes:

| Mode | Description | Use when | |---|---|---| | Safe (frame extraction) | System extracts 5–12 frames, anonymizes each, shows preview | Default — any video with or without PHI | | Full video | Entire file sent unprocessed | Only for already-anonymized files |

⚠️ In Full Video mode, frames are NOT anonymized automatically. Confirm absence of PHI before using.

🔍 Comparative Analysis

Side-by-side comparison of medical images across time or location.

Comparison modes:

  • Over Time — progression assessment (before/after treatment)
  • By Location — comparing scans of different anatomical regions
  • General — free-form multi-image comparison

Supports both single images and video/DICOM folder batches.

🔬 Advanced Analysis (Image + Context)

Upload a primary image plus optional additional files (PDFs, Word documents, photos). Add detailed clinical context. Receive a unified clinical directive combining all inputs.

🧊 Advanced 3D Visualization (Cinematic)

Dedicated high-fidelity volumetric rendering for MRI and CT DICOM series.

  • Cinematic Mode: Volume scattering for photorealistic organ rendering
  • Vessel Highlight: Vessels and contrast-enhanced areas shown in red; surrounding tissue becomes semi-transparent
  • Adaptive quality: Lower resolution while rotating for smooth performance; restores to HQ at rest
  • Apple M1 optimization: Automatic downsampling for heavy studies to maintain frame rate

📄 Document Scanning

Turns your smartphone camera into a document scanner.

Local Copier (browser mode)

Works entirely in your browser — no AI, no internet required. 100% private.

Features:

  • Brightness, contrast, and grayscale adjustments
  • Export to Word (.docx) or PDF (via system print dialog)
  • Free — no credits consumed

Smart OCR (AI mode)

Extracts text and tables from scanned documents for further processing.

PHI protection:

  • Mandatory anonymization toggle
  • Auto-redaction of names and addresses when enabled
  • 🎨 Manual Redact editor: paint over any sensitive area before submitting

👥 Patient Database

Local patient records stored in your browser's IndexedDB — data never leaves your device.

Features:

  • Add patients with name (anonymized alias recommended), age, gender, diagnosis, notes
  • Save analysis results to patient records from any analysis section
  • View analysis history, timeline, and lab value trend charts
  • AI case summary: one-click narrative summary of all saved analyses for a patient

🔌 Direct Device Connection (USB)

Read data from ECG monitors, pulse oximeters, glucometers, and other serial-interface devices directly through the browser — no drivers required.

  1. Go to the Devices section
  2. Select baud rate (usually 115200)
  3. Click Connect and select your device from the browser prompt (Chrome / Edge only)
  4. View live ECG curve or sensor data
  5. Click Analyze fragment for immediate AI interpretation of the current segment

🛡️ Privacy & Data Handling

Doctor Opus is built on a Local-First principle — patient data stays on your device.

| Data type | Storage location | Leaves the device? | |---|---|---| | Patient cards & analysis history | Browser IndexedDB | Never | | Medical images during analysis | Browser RAM | Only anonymized fragments | | AI results (saved) | Browser IndexedDB | No | | User account & credit balance | Cloud PostgreSQL | Yes (no medical data) | | Analysis statistics (anonymized) | Cloud PostgreSQL | Yes (no PHI) |

Three-level anonymization before any AI call:

  1. Browser-side text regex — names, dates, IDs stripped
  2. Image canvas redaction — PHI zones painted black
  3. Server-side recursive scrub — all request fields cleaned before OpenRouter

No Personal Health Information (PHI) or Personally Identifiable Information (PII) linked to medical scans is stored in the cloud database.

💰 Credit System

Credits are consumed when using advanced AI models. Simple reference lookups and local tools are free.

| Operation | Credit cost (approx.) | |---|---| | Fast analysis (Gemini 3 Flash) | ~0.3 – 0.8 cr. | | Optimized analysis (Sonnet 5) | ~0.8 – 1.5 cr. | | Expert Validated (Opus 5.5 / GPT-5.6 Sol) | ~1.5 – 3.5 cr. | | PDF page (Vision processing) | ~0.3 cr. per page | | Medical Translator | 5.1 cr. per minute while the microphone is on | | Conversation recording for a clinical protocol | 62 cr. per hour of audio | | Local copier / calculators | Free |

Packages:

  • Starter: 50 credits — $6.99
  • Standard: 180 credits — $19.99
  • Pro: 600 credits — $59.99

Payment flow (current):

  • Select package on the Subscription page
  • Send exact USDT amount via TRON (TRC20) to the wallet shown in the invoice
  • Paste transaction hash (txHash) to confirm payment automatically
  • Credits are added after transaction validation

There are no trial credits for AI features.

Exact cost of each request is shown in the result block immediately after analysis completes. Full transaction history is available in Balance & History.

💡 Tips for Best Results

  • Always add clinical context (CC, HPI, key PMH) — it significantly improves relevance and accuracy
  • Use text-searchable PDFs (not image scans) for the Personal Library
  • For ECG: use Claude Sonnet 5 in Optimized mode for arrhythmia detail
  • For fractures: Claude Sonnet 5 outperforms other models (83% accuracy)
  • For complex genetics or rare pathology: use Claude Opus 5.5 (Expert Validated mode)
  • The system improves over time through your feedback — please rate AI responses after tests