SRAS-net: Low-resolution chromosome image classification based on deep learning.

Area of Science:

• Genetics
• Medical Imaging
• Artificial Intelligence

Background:

• Prenatal karyotype diagnosis is crucial for identifying fetal genetic and congenital diseases.
• Chromosome classification is a vital yet labor-intensive component of karyotype analysis.
• Current deep learning methods struggle with low-quality images and sex chromosome classification.

Purpose of the Study:

• To develop an efficient and accurate chromosome classification method for prenatal diagnosis.
• To address limitations in existing deep learning models regarding image quality and sex chromosome classification.

Main Methods:

• Proposed SRAS-net, combining a super-resolution network (Self-Attention Negative Feedback Network) with traditional neural networks.
• Input low-resolution chromosome images into the super-resolution network to generate high-resolution images.
• Utilized the SMOTE algorithm to balance sex chromosome samples for improved model learning.

Main Results:

• Achieved a classification accuracy of 97.55%.
• Demonstrated superior performance compared to existing state-of-the-art methods.
• Successfully improved the accuracy of sex chromosome classification.

Conclusions:

• SRAS-net offers an effective solution for accurate chromosome classification in prenatal diagnosis.
• The integration of super-resolution and data augmentation significantly enhances classification performance.
• This method holds promise for improving the efficiency and reliability of genetic disease screening.