Multidimensional data analysis and classification using SMIAL.

Area of Science:

• Life Sciences
• Biotechnology
• Medical Imaging

Background:

• Label-free imaging generates multidimensional data but requires programming expertise for analysis.
• Current software solutions are fragmented and necessitate custom coding for machine learning workflows.
• Reproducible analysis of complex imaging data remains a challenge for many researchers.

Purpose of the Study:

• Introduce SMIAL, an open-source graphical user interface (GUI) software for label-free, multidimensional imaging data analysis.
• Enable end-to-end machine learning (ML) workflows without requiring programming skills.
• Facilitate reproducible image analysis through parameter saving and reloading.

Main Methods:

• Developed SMIAL as a stand-alone, 64-bit Windows executable GUI.
• Integrated pre-processing, segmentation, feature generation, feature selection, and classification functionalities.
• Supported direct input of multichannel image stacks or pre-processed data (feature tables, ML models).

Main Results:

• Demonstrated SMIAL's utility in three distinct applications: melanoma detection, mitochondrial response tracking, and food quality assessment.
• Enabled users to perform complex ML analyses on label-free imaging data without coding.
• Ensured reproducibility of analysis pipelines via parameter saving and reloading features.

Conclusions:

• SMIAL provides an accessible, integrated platform for machine learning-based analysis of multidimensional label-free imaging data.
• The software empowers researchers lacking programming expertise to conduct sophisticated image analysis.
• SMIAL enhances the reproducibility and efficiency of label-free imaging data analysis in life sciences.