Detection-Gated Glottal Segmentation with Zero-Shot Cross-Dataset Transfer and Clinical Feature Extraction

TL;DR

A detection-gated pipeline that integrates a YOLOv8-based detector with a U-Net segmenter facilitates the standardized, large-scale extraction of clinical biomarkers across diverse endoscopy platforms.

Abstract

Background: Accurate glottal segmentation in high-speed videoendoscopy (HSV) is essential for extracting kinematic biomarkers of laryngeal function. However, existing deep learning models often produce spurious artifacts in non-glottal frames and fail to generalize across different clinical settings. Methods: We propose a detection-gated pipeline that integrates a YOLOv8-based detector with a U-Net segmenter. A temporal consistency wrapper ensures robustness by suppressing false positives during glottal closure and instrument occlusion. The model was trained on a limited subset of the GIRAFE dataset (600 frames) and evaluated via zero-shot transfer on the large-scale BAGLS dataset. Results: The pipeline achieved state-of-the-art performance on the GIRAFE benchmark (DSC 0.81) and demonstrated superior generalizability on BAGLS (DSC 0.85, in-distribution) without institutional fine-tuning. Downstream validation on a 65-subject clinical cohort confirmed that automated kinematic features (Open Quotient, coefficient of variation) remained consistent with established clinical benchmarks. The coefficient of variation (CV) of the glottal area was found to be a significant marker for distinguishing healthy from pathological vocal function (p=0.006). Conclusions: The detection-gated architecture provides a lightweight, computationally efficient solution (~35 frames/s) for real-time clinical use. By enabling robust zero-shot transfer, this framework facilitates the standardized, large-scale extraction of clinical biomarkers across diverse endoscopy platforms. Code, trained weights, and evaluation scripts are released at https://github.com/hari-krishnan/openglottal.

Figures (5)

• Figure 1: Overview of the three main inference pipelines. Input (left) is the grayscale frame; each pipeline yields a segmentation mask (right). Solid arrows denote data flow; the gate symbol indicates that the output is set to zero when the detector does not fire (or after at most 4.0 consecutive missed frames), removing spurious detections.
• Figure 2: Effect of temporal hold duration (0--20 frames and $\infty$) on YOLO+UNet (GIRAFE test set): DSC (left axis) and Det.Recall (right axis). At 4000frames/s, 4.0 frames $=$ 1ms.
• Figure 3: Output of the YOLO+UNet pipeline on 12.0 evenly spaced frames from one patient (patient 1): glottal mask (green), YOLO bounding box (yellow), and per-frame area. The montage illustrates temporal consistency of the segmentation across the vibratory cycle.
• Figure 4: Effect of YOLO confidence threshold on YOLO-Crop+UNet performance (BAGLS test, 3500.0 frames, zero-shot). YOLO inference is run once; thresholds are applied in post-processing.
• Figure 5: Example glottal area waveforms: Healthy (Patient 14), Paresis (Patient 50), and Paralysis (Patient 46B1). Each panel shows the time-varying glottal area extracted by the pipeline from GIRAFE raw videos.