Electrochemiluminescence Interface

Desktop application for end-to-end ECL biosensor analysis — from image capture to ML-based concentration prediction.

Built at the MEMS, Microfluidics and Nanoelectronics Lab, BITS-Pilani Hyderabad Campus.

Code for https://doi.org/10.1016/j.compbiomed.2024.109546

---

What it does

1. Image Analysis — Batch-processes ECL sensor images (JPG/PNG/GIF). Extracts light intensity by masking pixels in the reagent's HSV hue range and computing the mean of the region of interest. Exports intensity–concentration pairs to Excel.
2. Data Analysis — Trains multiple regression models (Linear, RANSAC, Huber, Theil-Sen, Decision Tree, Random Forest, AdaBoost, Gradient Boost, KNN, SVM) on the exported data. Saves trained models as .pkl files and generates scatter plots + error metrics (R², MAE, RMSE).
3. Prediction — Loads trained models and predicts concentration from either a new image or a manually entered intensity value.
4. Calibration — Interactive HSV/LAB color-space masking tool with live sliders. Lets you visually tune a reagent's hue range and apply it directly to the analysis pipeline.
5. Real-Time (Raspberry Pi only) — Live ECL analysis from PiCamera v2.

Supported reagents: Luminol (hue 110–130) and Ruthenium (hue 0–20). Auto-detection is available for both.

---

Requirements

Component	Minimum
Python	3.9+
OS	Windows / macOS / Linux
RAM	4 GB
For real-time tab	Raspberry Pi 3+ with PiCamera v2+

---

Setup

git clone https://github.com/Shash976/ECLInterface.git
cd ECLInterface
pip install -r requirements.txt
python src/main.py

---

Running the tests

pip install pytest
pytest tests/ -v

Tests cover the full scientific pipeline (image processing, ML training, prediction) and run headlessly — no display or camera required.

---

Project structure

ECLInterface/
├── src/
│   ├── main.py               # Entry point
│   ├── ml_gui_pyqt5.py       # Main window and all tabs
│   ├── image_processor.py    # Batch-processing state and timer (ImageProcessor)
│   ├── image_analysis.py     # HSV masking, intensity extraction
│   ├── processing.py         # ML training pipeline, Excel export
│   ├── prediction.py         # Model loading and inference
│   ├── calibration.py        # Interactive color-space calibration tool
│   ├── cameraApp.py          # Real-time PiCamera tab
│   ├── model_def.py          # ML model registry, Reagent definitions, DataAxis
│   ├── util.py               # Shared utilities (crop, GIF, platform open)
│   └── media/                # Images used by the UI
├── tests/
│   ├── conftest.py           # Qt mock, matplotlib backend, sys.path setup
│   ├── test_util.py
│   ├── test_model_def.py
│   ├── test_image_analysis.py
│   ├── test_processing.py
│   └── test_prediction.py
└── requirements.txt

---

Image folder convention

The batch image analysis tab expects a folder laid out as:

experiment_folder/
├── 0.1 uM/
│   ├── image1.jpg
│   └── image2.jpg
├── 0.5 uM/
│   └── image1.jpg
└── 1.0 uM/
    └── image1.gif

Each subfolder name must contain a numeric value — that value becomes the concentration label. Subfolders are sorted numerically before processing. GIF files are handled by taking the max-intensity frame across all frames.

---

Calibrating a reagent

1. Open the Calibrate tab and browse to a representative image.
2. Switch to HSV mode.
3. Adjust the six sliders (lower H/S/V and upper H/S/V) until only the emission region is visible in the mask panel.
4. Select the target reagent from the dropdown and click Apply to Reagent. The hue range is applied to the analysis pipeline immediately for the current session.