Overview

Code example

Below is a simple example of how RamanSPy can be used to load, preprocess and analyse Raman spectroscopic data. Here, we load a data file from a commercial Raman instrument; apply a preprocessing pipeline consisting of spectral cropping, cosmic ray removal, denoising, baseline correction and normalisation; perform spectral unmixing; and visualise the results.

For more examples, check out Tutorials and Examples.

import ramanspy as rp

# load data
image_data = rp.load.witec("<PATH>")

# apply a preprocessing pipeline
pipeline = rp.preprocessing.Pipeline([
    rp.preprocessing.misc.Cropper(region=(700, 1800)),
    rp.preprocessing.despike.WhitakerHayes(),
    rp.preprocessing.denoise.SavGol(window_length=9, polyorder=3),
    rp.preprocessing.baseline.ASPLS(),
    rp.preprocessing.normalise.MinMax()
])
data = pipeline.apply(image_data)

# perform spectral unmixing
nfindr = rp.analysis.unmix.NFINDR(n_endmembers=5)
amaps, endmembers = nfindr.apply(data)

# plot results
rp.plot.spectra(endmembers)
rp.plot.image(amaps)
rp.plot.show()

Features

RamanSPy offers a range of features to support Raman research and analysis. This includes the following:

Complete workflow support

RamanSPy streamlines the entire Raman spectroscopic data analysis lifecycle by providing accessible, easy-to-use tools for loading, preprocessing, analysing and visualising diverse Raman spectroscopic data. All functionalities of RamanSPy are completely application-agnostic and work equally well on any data loaded into the framework, regardless of their spectroscopic modality (single-point spectra, imaging, volumetric) and instrumental origin. This allows users to construct entire analysis workflows with minimal software requirements which are out-of-the-box transferable across different datasets and projects.

Preprocessing pipelines

The package also improves the availability, consistency and reproducibility of preprocessing processes by providing a pipelining infrastructure that streamlines the compilation, customisation and execution of complex preprocessing pipelines, as well as provides a library of already established and validated preprocessing protocols. This not only improves preprocessing automation, but also ensures that the same preprocessing pipeline is applied consistently across different, as well as that users can easily access and reproduce the preprocessing steps used in published research.

Integrative analysis

Furthermore, RamanSPy has been designed such that it offers direct integration with the entire Python ecosystem, thereby allowing smooth incorporation with other Python packages for spectroscopic research (e.g. pysptools), statistical modelling and machine learning (e.g. scikit-learn), deep learning (e.g. pytorch, tensorflow) and many others. With that, we aim to facilitate the integration of new (AI-based) methods and applications into the Raman spectroscopic workflow and catalyse the emerging effort to bridge the gap between Raman spectroscopy and AI & ML.

Model development

Finally, RamanSPy has been equipped with a range of tools for algorithmic development and evaluation, including a library of diverse Raman spectroscopic datasets, as well as a set of metrics for the evaluation of model performance. This allows users to more efficiently and consistently develop and evaluate new (AI-based) methods and algorithms for Raman spectroscopy applications.