Efficient Precision Control in Object Detection Models for Enhanced and Reliable Ovarian Follicle Counting

2025-01-23

UNSURE@MICCAI

AI Summary

In assisted reproductive ultrasound, automatic follicle counting faces challenges including sub-millimeter size, severe occlusion, and speckle noise, leading to inaccurate localization and missed detections. To address these, this paper proposes a YOLOv8-based single-stage detection framework with key innovations: (1) a dynamic precision control mechanism featuring a lightweight confidence-localization co-calibration module embedded in the detection head—enabling fine-grained, demand-driven accuracy allocation for the first time; (2) integration of multi-scale feature enhancement, adaptive non-maximum suppression (NMS), and uncertainty-aware regression loss; and (3) clinical prior-constrained post-processing. Evaluated on a multi-center ultrasound dataset, the model achieves 96.2% recall, 94.7% precision, and 92.4% of cases with follicle count error ≤ ±1, while operating at 47 FPS—meeting real-time clinical deployment requirements.

Technology Category

• Computer Vision: Other Foundations of Computer Vision
• Machine Learning: Calibration & Uncertainty Quantification
• Intelligent Robots: State Estimation

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Problem

• Ovarian Follicle Counting
• Model Accuracy
• Reproducibility

Innovation

• High-definition Virtual Slide Scanning
• Machine Learning
• Biological Knowledge Integration