ArthroPhase: a novel dataset and method for phase recognition in arthroscopic video
These videos have been matched automatically. Contact us if they are not relevant.
These videos have been matched automatically. Contact us if they are not relevant.
View abstract on PubMed
Summary
This summary is machine-generated.This study introduces a new dataset and transformer model for recognizing surgical phases in arthroscopic Anterior Cruciate Ligament (ACL) reconstruction. The model achieves high accuracy, aiding surgical training and efficiency.
Area Of Science
- Computer Vision
- Medical Imaging
- Orthopedic Surgery
Background
- Surgical phase recognition is crucial for arthroscopic procedures like Anterior Cruciate Ligament (ACL) reconstruction.
- Existing methods face challenges due to limited field of view, occlusions, and visual distortions in arthroscopy.
Purpose Of The Study
- To develop a novel transformer-based model for accurate surgical phase recognition in arthroscopic procedures.
- To establish a benchmark for arthroscopic surgical phase recognition using a new dataset.
Main Methods
- Developed the ACL27 dataset with 27 labeled arthroscopic ACL reconstruction videos.
- Employed a transformer-based architecture with ResNet-50 for spatio-temporal feature extraction.
- Introduced a Surgical Progress Index (SPI) to quantify surgery progression.
Main Results
- Achieved 72.9% accuracy on the ACL27 dataset and 92.4% on the Cholec80 dataset.
- SPI demonstrated low error rates (10.6% on ACL27, 9.8% on Cholec80) for reliable surgery progression estimation.
- Performance comparable to state-of-the-art methods on the Cholec80 dataset.
Conclusions
- The proposed transformer model and dataset significantly advance arthroscopic surgical phase recognition.
- The model shows effectiveness and generalizability, with potential to enhance surgical training and efficiency.
- Publicly available dataset and code will foster future research in orthopedic surgery.
These videos have been matched automatically. Contact us if they are not relevant.
These videos have been matched automatically. Contact us if they are not relevant.
