NeuronAlg: An Innovative Neuronal Computational Model for Immunofluorescence Image Segmentation.

Area of Science:

• Digital pathology
• Computational neuroscience
• Image analysis

Background:

• Accurate segmentation of regions of interest is crucial in digital pathology image analysis.
• Developing robust segmentation methods that do not rely on machine learning (ML) is of significant interest.
• Current ML approaches often require extensive tuning and can be sensitive to noise.

Purpose of the Study:

• To present a deterministic, non-ML approach for automated cell and nuclei segmentation in fluorescence images.
• To demonstrate the robustness and equivalent performance of this method compared to ML techniques.
• To provide a reliable tool for classifying and diagnosing indirect immunofluorescence (IIF) raw data.

Main Methods:

• A deterministic computational neuroscience approach was developed for cell and nuclei identification.
• The method is based on formally correct functions and does not require dataset-specific tuning.
• The algorithm, named NeuronalAlg, is fully automatic and optimized for diverse datasets.

Main Results:

• The NeuronalAlg method demonstrated robustness against variations in image size, mode, and signal-to-noise ratio.
• Validation on three independent datasets (Neuroblastoma, NucleusSegData, ISBI 2009) showed excellent performance.
• Quantitative indicators confirmed comparable or superior results to three published ML approaches.

Conclusions:

• Deterministic, formally correct methods ensure optimized and functionally accurate results in image analysis.
• The NeuronalAlg method offers a robust and reliable alternative to ML for cell and nuclei segmentation.
• This approach enhances digital pathology by providing accurate and consistent image analysis for diagnostics.