Feature Visualization

The feature visualization tool provides interactive visual analysis and statistical exploration of extracted features from your dataset. It enables comprehensive examination of feature distributions, correlations, dimensionality reduction, and cell-type-specific patterns through an interactive web interface.

Overview

Feature visualization helps you understand the characteristics and quality of your extracted features before training MIL models.

Feature Visualizer Dashboard

The tool provides:

  • Statistical descriptors: Mean, median, standard deviation, skewness, and kurtosis

  • Distribution analysis: Histograms and kernel density estimates

  • Correlation analysis: Feature correlation matrices

  • Dimensionality reduction: PCA and t-SNE visualizations

  • Cell-type analysis: Cell type distributions and feature comparisons across cell types

  • Multi-slide analysis: Combined visualizations across entire datasets

CLI Usage

Basic Command

vis_features --dataset PATH

Required Arguments

--dataset PATH

Path to the dataset folder containing slides with extracted features. This should be the output directory from feature extraction or dataset creation steps.

Complete Example

vis_features --dataset ./results/my_dataset

This command will:

  1. Launch an interactive web application at http://127.0.0.1:8050

  2. Scan the dataset for available slides and feature extraction outputs

  3. Provide interactive controls to navigate through different feature sets

  4. Generate visualizations on-demand based on your selections

See Also