Classification of Metaphase Chromosomes Using Deep Convolutional Neural Network.

Area of Science:

• Medical Imaging
• Computational Biology
• Genetics

Background:

• Karyotype analysis is crucial for diagnosing diseases like birth defects and hematological tumors.
• Accurate chromosome identification from G-banded images is challenging but vital for clinical applications.
• The increasing volume of patient samples necessitates automated solutions for medical testing.

Purpose of the Study:

• To develop an automated method for chromosome classification using artificial intelligence.
• To evaluate the performance of a convolutional neural network (CNN) for identifying chromosomes in medical images.
• To assess the potential of CNNs in constructing an automated karyotyping platform.

Main Methods:

• A convolutional neural network (CNN) architecture was designed with 6 convolutional, 3 pooling, 4 dropout, and 2 fully connected layers.
• The CNN model was trained using a labeled dataset of G-banded metaphase images.
• A softmax activation function was employed for mapping chromosome classifications into 24 distinct classes.

Main Results:

• The developed CNN classifier achieved a high accuracy of 93.79% for chromosome identification.
• The model demonstrated robust performance in distinguishing between different chromosome classes.
• The results indicate significant potential for AI in automating complex cytogenetic analyses.

Conclusions:

• The CNN model shows considerable promise for accurate and automated chromosome classification.
• This approach can significantly contribute to the development of efficient automated karyotyping systems.
• The findings support the integration of AI in clinical diagnostics for genetic disorders.