←Home

How to Use the Medical Text Analyzer

A comprehensive guide to analyzing medical documents with AI-powered entity recognition

Quick Start

Get started in 3 simple steps:

  1. Input your text - Upload a file or paste medical text directly
  2. Select a model - Choose which medical entities to detect
  3. Analyze - Click the analyze button and view results instantly

Step 1: Input Your Medical Text

Option A: Upload a File

Supported formats:

  • PDF - Medical reports, lab results, discharge summaries
  • DOCX - Word documents, clinical notes
  • TXT - Plain text files

Maximum file size: 10MB

Enhanced Extraction

When to enable Enhanced Extraction:

  • Scanned documents or images
  • PDFs with poor text quality
  • Documents with tables or complex layouts
  • Handwritten clinical notes

Note: Enhanced extraction uses AI-powered OCR and may incur additional processing time.

Option B: Paste Text Directly

Simply paste your medical text into the text area. Perfect for:

  • Quick analysis of clipboard content
  • Testing with sample text
  • Short clinical notes or snippets

Step 2: Choose Your Analysis Model

Select the type of medical entities you want to detect. Each model is specialized for different medical terminology:

🦠

Disease Recognition Model

Core Model

Detects diseases, conditions, disorders, and syndromes

Examples: Hypertension, Type 2 Diabetes, Acute Myocardial Infarction

💊

Pharmaceutical Identifier

Core Model

Identifies medications, drugs, and pharmaceutical compounds

Examples: Metformin, Lisinopril, Aspirin 81mg

🫀

Medical Anatomy Detector

Core Model

Recognizes anatomical structures, organs, and body parts

Examples: Left ventricle, Anterior descending artery, Liver

🔬

Pathology Detector

Core Model

Identifies pathological findings and abnormalities

Examples: Stenosis, Hypertrophy, Inflammation

🧬

Genomic Entity Extractor

Core Model

Extracts genomic and genetic information

Examples: BRCA1, TP53 mutation, HLA-B27

🔒

PII Detection System

Core Model

Identifies personally identifiable information for privacy

Examples: Patient names, dates of birth, MRN numbers

💡 Pro Tip: Core models are pre-loaded for instant analysis!

Additional specialized models (Protein, Chemical, Oncology, Species) are available and load on-demand when selected.

Step 3: Adjust Confidence Threshold

The confidence threshold controls which detected entities are shown in your results.

0%50%100%

Lower Threshold (30-50%)

More entities detected, but may include false positives

Good for: Exploratory analysis, catching all possibilities

Higher Threshold (70-90%)

Fewer entities, but higher accuracy and precision

Good for: Production use, reliable results

Recommended: 65% (Default)

Balanced approach that filters out most false positives while retaining true entities.

Step 4: Understanding Your Results

Results are displayed in three different views for comprehensive analysis:

📊 Table View

Sortable and filterable table of all detected entities

  • ✓Sort by confidence, type, or position
  • ✓Filter by entity type
  • ✓View confidence scores as progress bars
  • ✓See exact text positions

🔗 Graph View

Interactive visual representation of entities

  • ✓Color-coded nodes by entity type
  • ✓Pan and zoom for detailed inspection
  • ✓Quick overview of entity distribution
  • ✓Interactive exploration

✨ Highlighted Text

Original text with entities highlighted in context

  • ✓See entities in their original context
  • ✓Color-coded highlights by type
  • ✓Hover for confidence scores
  • ✓Legend showing entity counts

Statistics Summary

At the top of results, you'll see key metrics:

  • Total Entities: Number of entities detected
  • Average Confidence: Mean confidence score across all entities
  • Processing Time: How long analysis took (typically 50-500ms)
  • Model Used: Which model performed the analysis

Step 5: Export Your Results

Save your analysis results in multiple formats:

📄 JSON Export

Complete analysis data in machine-readable format

Includes: All entities, statistics, metadata, original text, timestamps

Best for: API integration, data processing, archival

📊 CSV Export

Entity data in spreadsheet-compatible format

Includes: Entity ID, type, text, confidence, position

Best for: Excel analysis, data manipulation, reporting

Tips & Best Practices

✅

For Best Results

  • •Use clear, well-formatted medical text
  • •Enable Enhanced Extraction for scanned documents
  • •Choose the model that matches your content (e.g., Disease model for diagnoses)
  • •Start with default 65% confidence and adjust based on results
  • •Review highlighted text view to verify entity accuracy
⚡

Performance Tips

  • •Shorter texts (< 5000 characters) process faster
  • •Core models load instantly - no wait time
  • •Optional models take 2-5 seconds to load first time
  • •File upload extracts text first, then analyzes (may take a few seconds)
🎯

Accuracy Tips

  • •Higher confidence threshold = more accurate but fewer results
  • •For exploratory analysis, try 50-60% threshold
  • •For production/clinical use, use 70-80% threshold
  • •Check multiple models if you need comprehensive entity detection
  • •Review entities in highlighted text to catch context-dependent meanings
💡

Use Case Examples

  • •Clinical notes: Use Disease + Anatomy + Pharma models
  • •Research papers: Use Genomic + Protein models
  • •De-identification: Use PII Detection model
  • •Lab reports: Use Pathology + Chemical models
  • •Drug safety: Use Pharma + Disease models

Troubleshooting

❓ File upload fails

Try these solutions:

  • →Check file size is under 10MB
  • →Ensure file format is PDF, DOCX, or TXT
  • →Try enabling Enhanced Extraction for scanned PDFs
  • →For very large files, try pasting text directly instead

❓ No entities detected

Try these solutions:

  • →Lower the confidence threshold to 40-50%
  • →Try a different model that matches your content
  • →Check if text contains medical terminology
  • →Verify text extracted correctly from uploaded file

❓ Too many false positives

Try these solutions:

  • →Increase confidence threshold to 70-80%
  • →Review entities in highlighted text view
  • →Use more specific model (e.g., Pharma instead of Disease)
  • →Filter by entity type in table view

❓ Analysis is slow

Try these solutions:

  • →Check if you're using Enhanced Extraction (takes 2-5s per page)
  • →Large files take longer to process
  • →First-time loading optional models adds 2-5 seconds
  • →Backend may be loading models on first startup (15-20 seconds)

Try It with Sample Text

Sample Medical Note

Patient presents with chest pain and shortness of breath. Past medical history includes hypertension and type 2 diabetes mellitus. Current medications: Metformin 1000mg twice daily, Lisinopril 20mg once daily, Aspirin 81mg daily. Vital signs: BP 145/92 mmHg, HR 88 bpm. Laboratory results: HbA1c 7.8%, LDL 145 mg/dL, Troponin elevated at 0.8 ng/mL. EKG shows ST elevation in leads II, III, aVF. Diagnosis: Acute inferior myocardial infarction. Plan: Cardiac catheterization, continue antiplatelet therapy.

Try analyzing this text with different models:

  • Disease model: Will find "hypertension", "type 2 diabetes mellitus", "myocardial infarction"
  • Pharma model: Will find "Metformin", "Lisinopril", "Aspirin"
  • Anatomy model: Will find "chest", mentions of heart-related structures
  • Pathology model: Will find "ST elevation", "Troponin elevated"

Need More Help?

Check out our comprehensive documentation for technical details, API reference, and advanced usage.