Cosmic Peak Correction

Interactive MATLAB GUI for detecting and removing cosmic ray spikes from spectral datasets using derivative-based detection and linear interpolation.

Uses the AppBase + uihtml architecture: HTML/CSS/JS frontend with MATLAB backend.

Features

• Derivative-Based Detection: Configurable derivative order (1st, 2nd, ...) with absolute value thresholding
• Automatic Threshold: Data-driven starting threshold from 10^floor(log10(max(|derivative|)))
• Manual Threshold Tuning: Logarithmic slider + click-to-set on the derivative chart
• All Spectra Overlay: View all individual spectra overlaid (with subsampling for large datasets)
• Signal-by-Signal Mode: Browse individual spectra with navigation badges
• Corrected Toggle: Show corrected only or both original + corrected
• Derivative Chart: Pointwise max |derivative| across all samples with sort mode (rank vs channel)
• Channel Expansion: Configurable number of channels to remove around each detected spike
• Batch Correction: Apply correction to all spectra simultaneously
• Progress Feedback: Step-by-step progress bar for loading, correction, and export
• Zoom/Pan: Mouse wheel zoom + drag pan on both charts
• Export: Corrected data exported as MATLAB workspace variable
• Dark Mode: Toggle via Ctrl+D or the + menu
• Resizable Panels: Drag handle between controls and charts
• Demo Data: Built-in Raman-like spectra with synthetic cosmic spikes

Installation

Add the CosmicPeakCorrection folder to your MATLAB path:

addpath('path/to/CosmicPeakCorrection');

Quick Start

% Empty (load data from GUI)
CosmicPeakCorrection();

% With a numeric matrix from workspace
CosmicPeakCorrection(spectra);

% With struct (data + optional wavelength)
s.data = spectra;
s.wavelength = wavelength;
CosmicPeakCorrection(s);

Input Data

Spectral Matrix

• 2D numeric matrix: samples (rows) x channels (columns)
• Must be real, finite (no NaN or Inf), and have at least 3 columns
• Supported types: double, single, integer types (converted to double internally)

X-Axis Vector (optional)

• 1D numeric vector matching the number of columns
• Selectable from the Load Data dialog (e.g., wavelength, wavenumber, Raman shift)
• If not provided, channel indices (1, 2, ..., N) are used

Algorithm

The correction follows five steps per sample:

1. Compute abs(diff(spectrum, d)) where d is the derivative order.
2. Compare each value against the threshold.
3. Flag channels where |derivative| > threshold as spike positions.
4. Expand each flagged position by k channels on each side.
5. Set flagged channels to NaN, then fillmissing('linear') to interpolate.

Parameter	Effect	Default	Typical Range
Derivative order	Higher orders amplify narrow spikes	1	1 -- 3
Channels to remove	Expansion radius around each spike	1	0 -- 5
Threshold	Lower = more aggressive detection	Auto	10^-2 -- 10^6

Architecture

CosmicPeakCorrection/
├── CosmicPeakCorrection.m                 # AppBase class (MATLAB backend)
├── CosmicPeak_test.m                      # Demo data generator
├── ui/
│   └── cosmic_peak_correction_ui.html     # HTML/CSS/JS frontend
├── business_logic/
│   ├── @CosmicPeakCorrector/
│   │   └── CosmicPeakCorrector.m          # Detection & interpolation logic
│   └── @DataValidator/
│       └── DataValidator.m                # Input validation
└── README.md

Three-Layer Separation

Layer	Files	Responsibility
UI	ui/cosmic_peak_correction_ui.html	Layout, sliders, charts (Canvas 2D), modals, dark mode
Backend	CosmicPeakCorrection.m	Action dispatcher, workspace I/O, orchestration
Business Logic	@CosmicPeakCorrector, @DataValidator	Algorithm, validation (no GUI code)

Programmatic Usage (No GUI)

% Create test data
CosmicPeak_test;  % generates 'spectra' and 'wavelength' in workspace

% Use the corrector directly
corrector = CosmicPeakCorrector();

% Correct all spectra
[correctedData, correctionMask] = corrector.correct( ...
    spectra, 1, 1, 100);

% Compute automatic threshold
threshold = corrector.autoThreshold(spectra, 1);

Requirements

• MATLAB R2020b or newer (for uihtml)
• No toolbox dependencies

Author

Adrian Gomez-Sanchez

License

MIT