Skip to main content

Longitudinal View

The Longitudinal View module offers a temporal perspective on validation results by presenting how diagnostic outcomes and model predictions evolve across time for individual patients. This page is especially useful for identifying clinical trends, monitoring disease progression, and evaluating model consistency.

1. Navigation​

  • Accessible via: Validation Analysis → Longitudinal View
  • Located alongside other analysis tabs such as:
    • Validation Metrics
    • Dual Comparison
    • Blind Spot Analysis
    • Validation Summary

2. Select a Validation Run​

  • Use the dropdown to choose a Validation or Batch Run (e.g., pain-earth, Covid-19 Stony Brook)
  • This filters the view to patient cases associated with the selected run

3. Sequential Patient Imaging Timeline​

  • Each patient case is displayed as a horizontal scroll of image thumbnails
  • Every thumbnail includes:
    • Date: When the image was captured
    • Ground Truth: Actual diagnostic outcome (e.g., Alive, Dead)
    • Prediction: Model's output for that specific time point

4. Detailed Image View​

  • Clicking on a thumbnail opens an enlarged version of the image
  • Alongside the visual, the following details are shown:
    • Acquisition Date
    • Ground Truth Label
    • Model Prediction (if available)

5. Temporal Diagnostic Review​

  • Allows visual inspection of disease progression over time
  • Enables analysts to:
    • Identify trends or abrupt changes in condition
    • Detect prediction inconsistencies
    • Assess early/late prediction accuracy

💡 Use Longitudinal View to explore how patient conditions evolve and how well the model captures that evolution across multiple imaging sessions.