Label-free imaging enables multidimensional data acquisition across spectral and temporal domains and is increasingly used in life sciences. However, users working with multidimensional label-free data often lack the programming expertise needed to build reproducible image analysis pipelines. Existing software require custom programming or switching between fragmented platforms for feature generation and classification tasks. In this work, we introduce SMIAL, an open-source graphical user interface (GUI) software available as a stand-alone package runnable as a 64-bit Windows executable that enables end-to-end machine learning (ML) workflows for multichannel 2D imaging data without requiring programming skills. Users can input multichannel image stacks directly in SMIAL or work with pre-processed multidimensional data via feature tables or pre-trained ML models. Additionally, SMIAL supports parameter saving and reloading to ensure reproducibility across its pre-processing, segmentation, feature generation, feature selection and classification panels. We demonstrate SMIAL’s utility across three representative applications of label-free, multispectral imaging: melanoma detection, tracking mitochondrial response to rotenone treatment and non-invasive assessment of food quality. SMIAL is an open-source graphical user interface (GUI) software designed for reproducible machine learning workflows on multichannel 2D label-free imaging data, enabling classification across biomedical and non-biomedical applications.
Knab et al. (Mon,) studied this question.