Flemboda artificial intelligence: hybrid fuzzy-convolutional neural network for efficient chromosome abnormality...

Area of Science:

• Clinical Genetics
• Computational Biology
• Bioinformatics

Background:

• Accurate chromosomal abnormality detection is crucial for clinical genetics diagnosis and treatment planning.
• Existing learning-based methods struggle with capturing diverse features, limiting classification performance.
• Automated detection systems need enhanced efficiency, accuracy, and robustness.

Purpose of the Study:

• To propose FLEMBODA AI, an advanced computational framework for automated chromosome defect detection.
• To improve the efficiency, accuracy, and robustness of existing methods.
• To provide a reliable solution for clinical diagnostics and large-scale genetic analysis.

Main Methods:

• Karyotype image acquisition, augmentation, and pre-processing (normalization, G-bending enhancement).
• Chromosome segmentation using U-Net.
• Feature extraction and classification using a Hybrid Fuzzy-Convolutional Neural Network (Hybrid Fuzzy-CNN) with VGG-16 for weight assignment.
• Chromosome defect localization using Mask Region-centric CNN (Mask R-CNN).

Main Results:

• FLEMBODA AI achieved a recall of 95.3%, precision of 94.8%, and F1-score of 95.0%, outperforming baseline models.
• The U-Net segmentation model attained 93.8% accuracy.
• The framework demonstrated significant improvements in abnormality localization and classification.

Conclusions:

• FLEMBODA AI offers a reliable and effective solution for automated chromosomal abnormality detection.
• The proposed framework shows strong potential for clinical diagnostics and future genetic analysis.
• Enhanced feature capture and classification methods contribute to superior performance.