Development of a code-free machine learning model for the classification of cataract surgery phases

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Medical Imaging

Background:

• Cataract surgery phase classification from videos is crucial for training and analysis.
• Developing accurate automated systems traditionally requires expert-level coding and deep learning knowledge.

Purpose of the Study:

• To assess the performance of automated machine learning (AutoML) in classifying cataract surgery phases using surgical videos.
• To evaluate if a code-free AutoML model can achieve performance comparable to or better than expert-developed models.

Main Methods:

• Two ophthalmology trainees utilized Google Cloud AutoML Video Classification to train a deep learning model.
• The model was trained, validated, and tested on two open-access datasets (122 surgeries) and externally validated on 10 surgeries.

Main Results:

• The AutoML model achieved excellent performance, with an area under the precision-recall curve of 0.855.
• Overall metrics included 96.0% accuracy, 81.0% sensitivity, and 77.1% recall.
• Hydrodissection and phacoemulsification were the most accurately predicted phases (100% and 92.31% respectively).

Conclusions:

• A code-free AutoML model can accurately classify cataract surgery phases from videos.
• The performance of the AutoML model is comparable or superior to models developed by experts, indicating its utility in ophthalmology.