Multimodal Document AI & Diagnostic Imaging

Medaius's Document AI turns static clinical files into structured, actionable medical data. By combining Google Gemini's multimodal vision and acoustic models with clinical NLP pipelines, Document AI can ingest chest radiographs, lab panels, PDF referral letters, and consultation audio files, extracting diagnostic findings, anatomical impressions, and standardized billing codes in seconds.

---

1. Document Architecture & Patient File Repository

All patient records, medical imaging, lab results, and audio consultations are organized within the patient's centralized document gallery. Navigate to Patient Chart > Documents to access the repository:

Category-Driven File Organization

Documents are automatically classified into structured clinical folders:

• Imaging: Radiographs (X-rays), ultrasound captures, CT/MRI slices, and photographic wound documentation.
• Consultations: Ambient audio recordings, dictation audio clips, and specialist referral letters.
• Lab Reports: Hematology panels, metabolic profiles, urinalysis records, and pathology results.
• Administrative / Consent: Patient intake agreements, insurance cards, and signed consent waivers.

Each document card displays vital clinical metadata:

• File Title & Thumbnail: Visual thumbnail preview of images or acoustic waveform icons for audio.
• Category Badge: Distinctive color-coded pills (e.g., Imaging, Consultation).
• File Size & Format: Format indicators such as image/jpeg, audio/mp4, or application/pdf.
• Upload Attribution: Timestamp and clinician or staff member who uploaded the asset.

---

2. Dedicated AI Analysis Section in the Document Viewer

When you open any file from the document gallery, the fullscreen Document Viewer provides three specialized inspection tabs:

1. Document Preview: Displays the high-resolution image with pan and zoom capabilities, or renders an interactive audio/video player.
2. Details & Metadata: Shows technical file properties, MIME types, file size, upload history, and access logs.
3. AI Analysis Tab: Houses the comprehensive multimodal diagnostic report generated by Gemini, complete with clinical findings, medical impressions, and recommended billing codes.

---

3. Multimodal Chest X-Ray (CXR) Analysis

Medaius allows clinicians to upload high-resolution diagnostic imaging (such as standard Posteroanterior Chest Radiographs) and run immediate multimodal computer vision evaluations:

Structured Radiographic Report Breakdown

When Document AI evaluates a chest radiograph, it generates a comprehensive, five-tier clinical report:

1. Examination & Technique

Documents projection and radiographic quality (e.g., Technique: PA projection. Adequate inspiration, no significant patient rotation or motion artifact. Diagnostic quality: High).

2. Anatomical Findings by Subsystem

• Lungs & Airway: Evaluates lung volumes, parenchymal opacities, focal consolidations, interstitial markings, and tracheal alignment.
• Cardiomediastinal Silhouette: Assesses cardiothoracic ratio (normal $\le 0.50$), aortic arch contour, and hilar structures.
• Osseous Structures: Examines clavicles, ribs, thoracic spine, and humeral heads for acute fractures or lytic/blastic lesions.
• Diaphragm & Pleural Spaces: Evaluates hemidiaphragm contour, costophrenic angles for blunting or effusions, and subdiaphragmatic free air.
• Soft Tissues: Evaluates extrathoracic soft tissues and surgical clips or subcutaneous emphysema.

3. Diagnostic Impression

A concise, synthesized diagnostic conclusion (e.g., Normal chest radiograph. No acute cardiopulmonary disease. Clear lung fields bilaterally without focal consolidation, pneumothorax, or pleural effusion).

4. Automated Clinical Billing & Coding Suggestions

Document AI automatically correlates clinical findings with standardized diagnostic and procedure codes:

• ICD-10-CM: Suggests appropriate diagnostic codes (e.g., Z00.00 - Encounter for general adult medical examination without abnormal findings).
• CPT / Procedure Codes: Suggests compliant procedural billing codes (e.g., 71045 - Radiologic examination, chest; single view).

[!TIP] Clinician Oversight: AI radiographic evaluations serve as preliminary clinical decision support. Clinicians can review findings side-by-side with the high-resolution radiograph before accepting suggestions into the patient's problem list or billing ledger.

---

4. Consultation Audio & Synchronized Transcript Viewer

When audio recordings (such as consultation sessions or ambient scribes) are archived in the patient document repository, Document AI provides an interactive acoustic review suite:

Features of the Audio Document AI Tab

• Continuous Audio Playback: Dedicated playback scrub bar with pause, resume, and timeline indicators.
• Timestamped Diarized Dialogue: Conversations are parsed into speaker turns with millisecond timestamps ([00:00], [00:02], [00:06]).
• Interactive Audio Synchronization: Clicking any line in the transcript immediately jumps the audio player to that exact moment in the consultation.
• Permanent Clinical Audit Archive: Audio files and their corresponding transcripts remain safely stored under the patient's permanent record, accessible for future chart audits, medico-legal documentation, or clinical handoffs.

---

5. Workflow Summary: Uploading and Running Document AI

1. Open Patient Chart > Documents.
2. Click Upload Document and select an imaging file or audio recording.
3. Select the appropriate folder category (Imaging, Consultations, or Lab Reports).
4. Once uploaded, click the document card to launch the Document Viewer.
5. Select the AI Analysis tab. If analysis hasn't completed automatically, click Run AI Analysis.
6. Review the generated anatomical findings, impression, and diagnostic billing codes.