Deep Learning Chromosome Recognition via ResNet and MLP

The patent replaces the manual mechanical recognition process with an automated deep learning system. The deep learning model automatically extracts features from chromosome images and classifies chromosome types, substituting the doctor's manual visual inspection and decision-making process. This automation dramatically reduces examination time from two weeks or longer to a much shorter automated processing time while maintaining high recognition accuracy through the model's learned patterns from training data.

The deep learning model achieves self-service by automatically learning chromosome recognition patterns from training data without requiring subjective human judgment. The model consistently applies the same learned features and classification rules to all chromosome images, eliminating variability caused by different doctors' expertise levels, fatigue, or external environmental influences. This self-service automation ensures reliable and consistent recognition results across different cases and time periods.

The patent applies preliminary action by pre-training the deep learning model on a large dataset of chromosome images before deployment. The model learns chromosome features, patterns, and classification rules in advance during the training phase. When deployed for actual chromosome analysis, the pre-trained model can quickly and accurately recognize chromosome types without requiring complex real-time processing or manual intervention, thus improving productivity while the one-time training complexity is amortized over many subsequent efficient inferences.