Source-linked AI summary
Constructing Predictive Surgical Path for AI-based Capsulorhexis Skill Transfer
Mohammad Javad Ahmadi, Hamid D. Taghirad
TL;DR
The paper addresses the challenge of transferring expert surgical skills to novice surgeons through AI-assisted training. It introduces an annotated capsulorhexis dataset and a deep-network platform that optimizes novice movement paths using expert features while preserving intent. The reported path improvement is at least 20 percent, with concrete indicators used to assess trainee improvement.
Problem
Effective transfer of surgical skills from expert surgeons to novices remains a central challenge in automating surgical training.
Method
The platform pre-trains a CNN-FFT network on JIGSAWS and ARAS-Farabi motion data, then optimizes novice paths using expert reference features and two loss functions.
Results
At least 20 percent improvement in the trainee surgeon’s movement path was reported while maintaining the surgeon’s intentional objective.
Takeaways & Limitations
The platform provides an AI agent for surgical skill transfer and uses tangible indicators to verify trainee improvement.
Takeaways & Limitations
The predictive-constructive process assumes a coefficient that may initially exceed one and can decrease over time to a value between zero and one.
Abstract
from arXiv · showhide
Automated training of surgeons is one of the most crucial factors that significantly minimize surgical training risks and expenses. With recent advances in artificial intelligence (AI) knowledge and available data from various surgeries, AI's involvement in surgical training is becoming very promising. It is recommended that at the early stages of AI development, it interferes in the surgery as a third agent alongside the trainer. As trust in AI increases, this process will lead to an AI agent acting as a trainer in the future. The first phase in which AI can intervene in the training process is to suggest an improved surgical path to the trainer. A platform must be constructed in the first step, to accomplish this task and to enhance the movement path of trainee surgeons. This paper introduces this platform along with an annotated capsulorhexis surgery dataset called the ARAS-Farabi dataset. In this research, a deep convolutional neural network is pre-trained with JIGSAWS and ARAS-Farabi surgical datasets that can extract surgical skill characteristics from surgery tool tip motion data. The proposed platform develops a reference model from the feature space of an expert surgeon's movement trajectory and proposes an improved path to enhance the skill of a novice surgeon. An optimization with two loss functions is utilized to create a path that raises the skill level of the novice surgeon's path while simultaneously predicting and preserving his/her intent. The results of this study reveal that, with the assistance of an AI agent, the trainee surgeon's movement path can be enhanced by at least 20 percent while maintaining his intentional objective. In addition to the recommended deep network, various tangible indicators have also been developed in this research to verify the level of trainee improvement.
1 INTRODUCTION
AI-assisted surgical training is presented as a way to address the difficulty of transferring expert skills to novices while reducing reliance on trainer surgeons. The proposed platform combines AI-based skill assessment with movement-path generation and can provide visual or haptic guidance.
- Transferring surgical skills effectively from expert surgeons to novices remains a major challenge in automating surgical training.
- Existing haptic and dual-user robotic systems support instruction and error correction, but most lack AI agents and still require substantial trainer effort.
- The paper proposes combining control structures with AI-based skill assessment to generate an expert path from a novice surgeon’s movements.
- The platform is intended to support an AI agent that initially assists a trained surgeon and could later provide surgical instruction independently.
- The study addresses AI trust by presenting a dataset and system that inject expert surgical information into novice motion while preserving established surgical procedures.
- The platform’s output can be delivered through visual analysis software or haptic feedback to the trainee surgeon.
2 METHODOLOGY
The platform combines two surgical datasets, video-derived motion tracking, and a convolutional network to extract skill features and construct improved trainee paths. Optimization incorporates expert skill characteristics while retaining the trainee’s surgical objective.
- Platform overview: The platform injects skill characteristics into a novice surgeon’s automatically generated path using a neural-network reference model.The proposed platform assesses trainee movements and uses a reference model developed from surgical skill features.
- Datasets: JIGSAWS provides general surgical-skill data from 103 surgeons performing suturing, needle passing, and knot tying.The dataset contains 76 kinematic variables, with skill levels primarily annotated using surgical experience.
- Datasets: ARAS-Farabi adds more than one hundred annotated capsulorhexis recordings classified by three expert surgeons as expert or novice.Motion data are retrieved from surgical videos using ARAS-Farabi software and OpenCV-based tracking algorithms.
- Motion-data extraction: The ARAS-Farabi software tracks surgical regions, permits observer correction of bounding boxes, and calculates position, speed, and acceleration.It addresses reflection and scale challenges through morphology operations, OpenCV libraries, and pupil-bounding-box normalization.
- Skill-related feature extraction: The network receives intervals of absolute or relative x, y, and z or velocity coordinates, emphasizing smoothness of the surgeon’s entry path.An FFT block supplies frequency-domain inputs, while convolutional layers, batch normalization, and ELU activations extract motion patterns.
- Predictive-constructive path: The predictive-constructive procedure uses bilateral optimization and a loss function designed to preserve the trainee’s intended path while making it appear more expert.The method combines decision and skill-related objectives during path construction.
3 RESULTS AND DISCUSSION
The platform constructs novice movement paths with more expert characteristics by optimizing skill improvement while preserving surgical intention. Across experiments, the proposed paths became smoother and more expert, with improvements varying by loss weighting, reference expertise, and transferred procedures.
- Network training: 99% F1 score was obtained after tuning the network parameters and hyperparameters.The trained model stores surgical skill knowledge for later path construction.
- Cross-procedure transfer: Transferring skills from four expert surgical procedures made the proposed novice path 55% more expert on average.The experiment used the same coefficients across the transferred procedures.
- JIGSAWS-based transfer: Applying the structure to x, y, and z motion variables produced paths with more expert traits and fewer surprising motions and tremors.The skill-loss coefficient controls how increasingly expert the suggested path becomes.
4 CONCLUSION
The paper presents a CNN-FFT platform that learns surgical skill features from JIGSAWS and ARAS-Farabi motion data to construct improved novice paths. Two loss functions preserve the trainee’s objective while increasing skill, supporting AI-assisted skill transfer in surgical training.
- Platform: A CNN-FFT network pre-trained on JIGSAWS and ARAS-Farabi motion data generates an expert reference feature space and constructs novice paths.The network uses expert data to form the reference model and novice movement data to generate a proposed path.
- Skill transfer: Two loss functions and optimization create paths that enhance trainee skill while preserving the trainee surgeon’s objective.The paper reports this as an advantage over previous studies for capsulorhexis surgery.