CIR-Net

CIR-Net classifies human chromosomes from G-band karyotype images using an optimized Inception-ResNet deep learning architecture combined with Chromosome Data Augmentation (CDA) to improve automated karyotyping accuracy.

Key Features:

• Deep Learning-Based Approach: Uses an optimized Inception-ResNet architecture for chromosome image classification.
• Augmentation Method - CDA: Implements Chromosome Data Augmentation (CDA) to augment limited training datasets and improve model generalization.
• High Classification Accuracy: Reports 95.98% classification accuracy on clinical G-band chromosome datasets, improving from 87.46% (an increase of >8.5%) when CDA is applied.

Scientific Applications:

• Karyotyping: Automates chromosome classification to support construction of karyotypes from G-band images.
• Medical Diagnostics: Supports cytogenetic analysis in clinical diagnostics by classifying human chromosomes in clinical G-band datasets.
• Drug Development and Biomedical Research: Enables scalable chromosome analysis for drug development studies and biomedical research.

Methodology:

Training an optimized Inception-ResNet deep neural network on chromosome image data with Chromosome Data Augmentation (CDA) to address limited datasets.