Deep Learning Method for Classifying Spatial Patterning of Early Differentiated Human Induced Pluripotent Stem Cells...

Area of Science:

• Stem Cell Biology
• Computational Biology
• Genetics

Background:

• Micropattern cultures facilitate analysis of human induced pluripotent stem cells (hiPSCs) differentiation into three germ layers.
• Genetic abnormalities in hiPSCs can lead to highly variable spatial patterning during differentiation.
• Characterizing these abnormal patterns is crucial for understanding hiPSC behavior.

Purpose of the Study:

• To develop and evaluate a Convolutional Neural Network (CNN) structure with global average pooling for classifying micropattern images of early differentiated hiPSCs.
• To analyze the differentiation status of hiPSCs with genetic abnormalities using image classification.

Main Methods:

• A CNN structure was designed with successive downsampling and an increased number of filters at each stage.
• Global average pooling was implemented to mitigate overfitting and optimize classification.
• Seven classes of ectodermal cell images representing various genetic abnormalities (e.g., trisomy, uniparental disomy, loss of heterozygosity) were used for training and validation.

Main Results:

• The proposed CNN model achieved an accuracy of 81.4% in classifying micropattern images of early differentiated hiPSCs.
• The method demonstrated effectiveness in distinguishing between different types of differentiation patterns associated with genetic abnormalities.

Conclusions:

• The developed CNN approach is a useful tool for analyzing the early differentiation of hiPSCs, particularly those with genetic abnormalities.
• This method aids in characterizing abnormal spatial patterning, offering insights into the impact of genetic variations on stem cell differentiation.