analyze-evidence

name: analyze_evidence description: Analyze dental X-ray images to detect teeth, pathologies, and estimate age/gender.

Analyze Evidence Skill

Use this skill to perform forensic analysis on dental X-ray images. This involves detecting teeth (FDI notation), identifying pathologies, and estimating biological profile (age/gender).

Dependencies

• DentalID.Core.Interfaces.IAiPipelineService (The AI Engine)
• DentalID.Application.Services.ForensicAnalysisService (High-level orchestration)
• DentalID.Core.DTOs.AnalysisResult (The output format)

Workflow

1. Validate Input:

   • Ensure the file is a supported image format (.png, .jpg, .bmp, .dcm).
   • SECURITY: Do NOT pass file paths directly to the AI engine if possible. Open a FileStream (Read-only) and pass the Stream. This ensures file locks and prevents race conditions.

2. Run Analysis:

   • Call IAiPipelineService.AnalyzeImageAsync(stream).
   • Note: This method is computationally expensive.

3. Interpret Results:

   • Teeth: Check AnalysisResult.Teeth.
   • Pathologies: Check AnalysisResult.Pathologies.
   • Demographics: Check EstimatedAge and EstimatedGender.

4. Advanced: Batch Processing:

   • When processing multiple images (e.g., a full case folder), use Task.WhenAll to maximize CPU/GPU usage.
   • Memory Warning: Process in chunks of 5-10 images to avoid OutOfMemory exceptions on standard laptops.

5. Error Recovery:

   • Low Confidence: If Teeth.Count < 5, the image might be blurry.
     • Action: Apply Histogram Equalization (using OpenCV/Skia) and retry.
   • Upside Down: If 11 is found at the bottom of the image.
     • Action: Rotate 180 degrees and retry.

Example Usage (C#)

using DentalID.Core.Interfaces;
using DentalID.Core.DTOs;

public async Task<AnalysisResult> PerformForensicAnalysis(string imagePath, IAiPipelineService aiService)
{
    // 1. Secure Open
    using var stream = File.OpenRead(imagePath);
    
    // 2. Analyze with Retry Logic
    try 
    {
        return await aiService.AnalyzeImageAsync(stream);
    }
    catch (Exception ex)
    {
        Console.WriteLine($"[Error] First pass failed: {ex.Message}. Retrying with enhancement...");
        // TODO: call ImagePreprocessor.Enhance(stream)
        return await aiService.AnalyzeImageAsync(stream); // Retry
    }
}

Troubleshooting

• Zero Teeth Detected: The image might be upside down or have massive artifacts. Try DentalID.Core.Utilities.ImagePreprocessor.EnhanceContrast.
• "Model Not Found": Ensure the models/ directory contains teeth_detect.onnx and pathology_detect.onnx.