MALDI-MSI_2PLSM_Histology

Multimodal image analysis pipeline combining MALDI mass spectrometry imaging, two-photon laser scanning microscopy (2PLSM), and H&E histology to characterize spatial heterogeneity in colorectal cancer tissue. Accompanies the publication in npj Imaging (2024).

Figure 1: Imaging protocol for correlation of 2PLSM, MALDI-MSI, and histology.

---

Features

2PLSM (collagen texture)

• Structure tensor analysis — local fiber coherence (0 = chaotic, 1 = organised) and orientation (0–180°)
• Local density maps and fibrotic area quantification
• Orientation vector field visualisation

Histology (H&E)

• StarDist nuclei segmentation (deep learning, 2D_versatile_he model)
• Kernel density estimation and nuclear morphometry (eccentricity, area)
• H&E stain normalisation (Ruifrok & Johnston colour deconvolution)
• Proximity network graphs and Voronoi tessellation of nuclei

MALDI-MSI

• Open-source pipeline replacing SCiLS Lab: imzML import, TIC normalisation, peak alignment
• ROC analysis per m/z, discriminative ion selection
• PCA and bisecting k-means spatial segmentation
• QuPath annotation import (GeoJSON)

---

Installation

git clone https://github.com/arora-bharti/MALDI-MSI_2PLSM_Histology.git
cd MALDI-MSI_2PLSM_Histology
python -m venv venv && source venv/bin/activate
pip install -r requirements.txt

Or with Docker (no install required — reproduces the exact analysis environment):

docker build -t maldi-pipeline .
docker run -p 8501:8501 maldi-pipeline

---

Usage

Streamlit web app

streamlit run app/streamlit_app.py

Command-line scripts

# Collagen texture analysis
python scripts/analyze_texture.py -i image.tif -o results/

# Nuclei segmentation
python scripts/segment_nuclei.py -i histology.tif -o results/

Jupyter notebooks

Notebook	Purpose
Collagen_textureanalysis.ipynb	Step-by-step 2PLSM texture analysis
Nuclei_segmentation.ipynb	Nuclei detection and density maps
Low_high_coherence_percentage.ipynb	Coherence statistics across images
Orientation_vectorfield.ipynb	Fibre orientation quiver overlays
MALDI_analysis.ipynb	Full MALDI-MSI pipeline

Tests

pytest tests/ -v

Tests for TensorFlow/StarDist-dependent functions are skipped automatically if those libraries are not installed.

---

Parameters

Texture analysis

Parameter	Description	Typical range
filter_sigma	Gaussian smoothing sigma	1–5
local_sigma	Structure tensor window	5–20
local_density_kernel	Kernel size for density	10–30

Nuclei segmentation

Parameter	Description	Default
prob_thresh	Detection probability threshold	0.4
nms_thresh	Non-maximum suppression threshold	0.3
normalization_percentiles	Input normalisation range	[25, 85]

---

Output files

Texture analysis: *_filtered.tif, *_density.tif, *_coherence.tif, *_orientation.tif, Results_*.png, *.csv

Nuclei segmentation: labeled/*.tif, density/*.tif, circularity/*.tif, area/*.tif, *_mosaic.png, results/*.csv

---

Citation

@article{arora2024maldi,
  title={MALDI imaging combined with two-photon microscopy reveals local differences in the heterogeneity of colorectal cancer},
  author={Arora, Bharti and Kulkarni, Ajinkya and Markus, M. Andrea and Ramos-Gomes, Fernanda and Bohnenberger, Hanibal and Str{\"o}bel, Philipp and Alves, Frauke and Klein, Oliver},
  journal={npj Imaging},
  volume={2},
  number={35},
  year={2024},
  doi={10.1038/s44303-024-00041-3}
}

---

License

GNU Affero General Public License v3.0 — see LICENSE.

Authors

Bharti Arora and Ajinkya Kulkarni Max Planck Institute for Multidisciplinary Sciences, Göttingen, Germany

Acknowledgements

Funded by: EU Horizon 2020 Marie Sklodowska-Curie (No 857894 – CAST), Ministry for Science and Culture of Lower Saxony (ABA), Federal Ministry of Education and Research (MSTAR #031L0220A), BCRT and Berlin Institute of Health.

---

Repository structure

MALDI-MSI_2PLSM_Histology/
├── src/
│   ├── modules_2photon.py     # 2PLSM texture analysis
│   ├── modules_histo.py       # Histology / nuclei analysis
│   ├── modules_maldi.py       # MALDI-MSI analysis
│   └── __init__.py
├── notebooks/                 # Jupyter notebooks (one per modality)
├── scripts/
│   ├── analyze_texture.py     # CLI for texture analysis
│   ├── segment_nuclei.py      # CLI for nuclei segmentation
│   └── Merging.ijm            # ImageJ macro for image overlay
├── app/
│   └── streamlit_app.py       # Interactive web app
├── tests/                     # pytest unit tests (52 tests, synthetic data)
├── sample_data/               # Place sample .tif images here
├── Dockerfile                 # Reproducible environment
├── requirements.txt
└── CITATION.cff