Implicitly assessing and dynamically prioritizing high-quality synthetic chromosome anomalies during training

Determine implicit assessment criteria and dynamic prioritization mechanisms that enable the selection of high-quality synthetic abnormal chromosome images during training so as to maximize their utility for downstream structural chromosomal anomaly detection.

Background

Even when synthetic anomalies can be generated, the absence of ground-truth labels makes it difficult to evaluate their authenticity and usefulness. Without reliable selection criteria, synthetic samples may not improve and can even hinder detector training.

The paper explicitly identifies the unresolved challenge of how to implicitly assess and dynamically prioritize synthetic samples during training to maximize their contribution to downstream anomaly detection.

References

Therefore, two key challenges remain unresolved: how to generate structurally realistic and diverse synthetic anomalies in the absence of sufficient real abnormal data, and how to implicitly assess and dynamically prioritize high-quality synthetic samples during training to maximize their utility for downstream anomaly detection.

— Perturb-and-Restore: Simulation-driven Structural Augmentation Framework for Imbalance Chromosomal Anomaly Detection  (2604.00854 - Zhang et al., 1 Apr 2026) in Section 1: Introduction