Algorithm for Calculating the Fractal Dimension of Medical Images Using Neutrophil Nuclei as an Example: Integration with Neural Network Segmentation to Simplify Training

Measurement Techniques · Springer Nature · 2026-06-01 · DOI: 10.1134/S000635092670048X

Аннотация

Abstract A hybrid algorithm for automated analysis of blood smears is proposed, combining the localization of neutrophil nuclei using the U-Net convolutional neural network and the subsequent calculation of their fractal dimension via the box-counting method along the contour highlighted by the color method (HSV segmentation). The approach allows for dividing the task into two stages: highlighting the area of interest and classification by scalar feature, which reduces the requirements for the volume and complexity of the the training dataset required for neural network training. Experiments on a dataset of 300 images demonstrated satisfactory segmentation quality (IoU ≈ 0.44) and a correlation between fractal dimension values and neutrophil maturation stages (ρ = 0.9350, p < 0.001). The method eliminates the need for manual contour delineation and ensures reproducibility of the analysis, which makes it promising for laboratory diagnostics.

Читать у издателя →

Похожие статьи