Flemboda artificial intelligence: hybrid fuzzy-convolutional neural network for efficient chromosome abnormality classification.
Journal:
Molecular genetics and genomics : MGG
Published Date:
Jan 21, 2026
Abstract
Chromosomal abnormality detection is a fundamental task in clinical genetics, as accurate identification of structural and numerical defects is essential for reliable diagnosis and treatment planning. However, many existing learning-based approaches failed to effectively capture diverse discriminative features, limiting their classification performance. To address this challenge, this study proposes FLEMBODA AI, an advanced computational framework designed to enhance the efficiency, accuracy, and robustness of automated chromosome defect detection. At first, the karyotype image is collected. Then, the input is augmented and pre-processed using normalization and a G-bending enhancement approach to ensure that the model's effectiveness generalizes across diverse datasets. After that, the chromosomes are segmented from the pre-processed image using a U-Net approach to isolate chromosomes from the karyotype image. Then, the segmented chromosomes are given as input to the Hybrid Fuzzy-Convolutional Neural Network (Hybrid Fuzzy-CNN) approach. In this, the CNN model performs the Feature Extraction (FE) process, and the Fuzzy logic is used for the classification of chromosomal abnormalities. Here, the VGG-16 is used for weight assignment for the classification. Then, using the Mask Region-centric CNN (Mask R-CNN), chromosome defect localization is performed. Experimental results demonstrate that FLEMBODA AI achieves a recall of 95.3%, precision of 94.8%, and an F1-score of 95.0%, outperforming baseline models. Additionally, the U-Net segmentation model attains an accuracy of 93.8%, contributing significantly to improved abnormality localization and classification performance. Overall, the proposed FLEMBODA AI framework provides a reliable as well as effective solution for automated chromosomal abnormality detection, with strong potential for application in clinical diagnostics and future large-scale genetic analysis systems.
