Deep learning prediction of error and skill in robotic prostatectomy suturing
2024

Using AI to Improve Robotic Surgery Skills

Sample size: 54 publication Evidence: moderate

Author Information

Author(s): Sirajudeen N., Boal M., Anastasiou D., Xu J., Stoyanov D., Kelly J., Collins J. W., Sridhar A., Mazomenos E.

Primary Institution: University College London

Hypothesis

This study aimed to validate surgical skill rating and error annotations in suturing gestures to inform the development and evaluation of AI models.

Conclusion

The study successfully demonstrated the feasibility of using detailed skill and error annotation in robotic surgical video, establishing a foundation for future advancements in automated skill assessment.

Supporting Evidence

  • Strong inter-rater reliability was demonstrated with a correlation coefficient range of r=0.70–0.89.
  • Skill estimation showed a prediction error of 17.92% for OSATS and 20.6% for M-GEARS.
  • The best performing models achieved a mean absolute precision of 37.14% and an area under the curve of 65.10%.

Takeaway

The researchers used videos of robotic surgery to teach computers how to spot mistakes and measure skills, which could help doctors perform better surgeries.

Methodology

The study used a dataset of robotic surgery videos, which were annotated for skill and errors, and employed deep learning models to analyze the data.

Potential Biases

Most videos were annotated by only one rater due to time constraints, which could introduce bias.

Limitations

The study focused only on simple suturing tasks and did not collect clinical outcomes, limiting predictive validity.

Participant Demographics

The videos analyzed were from 8 surgeons.

Statistical Information

P-Value

p<0.001

Confidence Interval

[0.806, 0.990]

Statistical Significance

p<0.001

Digital Object Identifier (DOI)

10.1007/s00464-024-11341-5

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication