Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A labeled dataset of simulated phlebotomy procedures for medical AI: polygon annotations for object detection and human-object interaction

Raúl Jiménez Cruz, César Torres-Huitzil, Marco Franceschetti, Ronny Seiger, Luciano García-Bañuelos, Barbara Weber

TL;DR

This work presents a labeled dataset of 11,884 RGB frames from simulated phlebotomy on a training arm, annotated with polygon segmentations for five objects and organized into a 70/15/15 train/validation/test split. Frames are selected via SSIM filtering from high-definition 1920×1080 videos and anonymized to preserve privacy, enabling safe public sharing. An annotation pipeline combines a large manually labeled golden set (8,743 images) with model-assisted labeling and subsequent manual corrections, producing YOLOv8 segmentation-ready labels. The dataset supports fine-grained tool detection, human–object interaction analysis, and conformance checking in medical training contexts, and is publicly available on Zenodo under CC BY 4.0.

Abstract

This data article presents a dataset of 11,884 labeled images documenting a simulated blood extraction (phlebotomy) procedure performed on a training arm. Images were extracted from high-definition videos recorded under controlled conditions and curated to reduce redundancy using Structural Similarity Index Measure (SSIM) filtering. An automated face-anonymization step was applied to all videos prior to frame selection. Each image contains polygon annotations for five medically relevant classes: syringe, rubber band, disinfectant wipe, gloves, and training arm. The annotations were exported in a segmentation format compatible with modern object detection frameworks (e.g., YOLOv8), ensuring broad usability. This dataset is partitioned into training (70%), validation (15%), and test (15%) subsets and is designed to advance research in medical training automation and human-object interaction. It enables multiple applications, including phlebotomy tool detection, procedural step recognition, workflow analysis, conformance checking, and the development of educational systems that provide structured feedback to medical trainees. The data and accompanying label files are publicly available on Zenodo.

A labeled dataset of simulated phlebotomy procedures for medical AI: polygon annotations for object detection and human-object interaction

TL;DR

This work presents a labeled dataset of 11,884 RGB frames from simulated phlebotomy on a training arm, annotated with polygon segmentations for five objects and organized into a 70/15/15 train/validation/test split. Frames are selected via SSIM filtering from high-definition 1920×1080 videos and anonymized to preserve privacy, enabling safe public sharing. An annotation pipeline combines a large manually labeled golden set (8,743 images) with model-assisted labeling and subsequent manual corrections, producing YOLOv8 segmentation-ready labels. The dataset supports fine-grained tool detection, human–object interaction analysis, and conformance checking in medical training contexts, and is publicly available on Zenodo under CC BY 4.0.

Abstract

This data article presents a dataset of 11,884 labeled images documenting a simulated blood extraction (phlebotomy) procedure performed on a training arm. Images were extracted from high-definition videos recorded under controlled conditions and curated to reduce redundancy using Structural Similarity Index Measure (SSIM) filtering. An automated face-anonymization step was applied to all videos prior to frame selection. Each image contains polygon annotations for five medically relevant classes: syringe, rubber band, disinfectant wipe, gloves, and training arm. The annotations were exported in a segmentation format compatible with modern object detection frameworks (e.g., YOLOv8), ensuring broad usability. This dataset is partitioned into training (70%), validation (15%), and test (15%) subsets and is designed to advance research in medical training automation and human-object interaction. It enables multiple applications, including phlebotomy tool detection, procedural step recognition, workflow analysis, conformance checking, and the development of educational systems that provide structured feedback to medical trainees. The data and accompanying label files are publicly available on Zenodo.
Paper Structure (8 sections, 5 figures, 1 table)

This paper contains 8 sections, 5 figures, 1 table.

Table of Contents

  1. File formats
  2. Repository structure
  3. Acquisition setup
  4. Privacy and preprocessing
  5. Frame curation and standardization
  6. Annotation pipeline
  7. Quality assurance and validation
  8. Code and software

Figures (5)

  • Figure 1: Vision-based conformance checking of a phlebotomy process, combined with IoT data seiger2026online.
  • Figure 2: Phlebotomy simulation and monitoring setup using IoT sensors and cameras.
  • Figure 3: Side-by-side example of original and annotated frames showing the object classes.
  • Figure 4: Detailed annotated example with polygons and text labels (syringe, gloves, rubber band, training arm).
  • Figure 5: Training and validation curves over 150 epochs for YOLOv8 segmentation. Loss terms: box_loss (box regression), cls_loss (classification), seg_loss (segmentation masks), and dfl_loss (distribution focal loss for box refinement). Metrics: precision, recall, and mean Average Precision at IoU thresholds; mAP50 corresponds to IoU = 0.50 and mAP50--95 averages across 0.50:0.05:0.95.