Intelligent Evaluation and Supervision of Cataract Surgery

Research purpose: intelligent identification and evaluation of cataract surgery steps Research methods: A total of 9 items (such as gender, age, visual acuity, etc.) were extracted from the surgical videos of senile cataract patients and the clinical data recorded by the electronic medical record system. The machine learning algorithm 3D-CNN was applied to identify the 11 steps in cataract surgery and the pictures (blank pictures) without instrument manipulation on the eyeball during the operation. Six key cataract surgery steps were scored using deep learning algorithms (probability smoothing window and softmax). We employ precision, precision, recall, and F1-score to evaluate the model's performance for recognizing surgical steps. To evaluate the reliability of the model's scoring of surgical steps, we used a human-machine comparison method to calculate the agreement (kappa value) between machine and expert scores.

Study Overview

• Status: Completed
• Conditions: Cataract
• Intervention / Treatment: Other: Evaluation test: cataract surgery steps

• Study Type: Observational
• Enrollment (Actual): 344

Contacts and Locations

Study Locations

• China
  • Guangdong
    • Guangzhou, Guangdong, China, 510060
      • Zhognshan Ophthalmic Center, Sun Yat-sen University

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: 46 years to 96 years (Adult, Older Adult)
• Accepts Healthy Volunteers: N/A
• Genders Eligible for Study: All

• Sampling Method: Probability Sample

Study Population

Participants who had senile cataracts and phacoemulsification and IOL implantation in the Zhongshan Ophthalmic Centre (ZOC, Guangzhou, Guangdong, China) and Shenzhen Eye Hospital (Shenzhen, Guangdong, China)

Description

Inclusion Criteria:

-Videos of phacoemulsification and IOL implantation for senile cataracts will be included

Exclusion Criteria:

-The peak signal-to-noise ratio (PSNR) is utilized to assess whether a video was blurred. If the PSNR of a video was less than 20 decibels (dBs), the whole video was discarded.

Study Plan

How is the study designed?

Design Details

• Observational Models: Other
• Time Perspectives: Retrospective

Number of groups / cohorts

3

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
Development Dataset; 12 cataract surgery steps including(1) main incision formation, (2) side incision formation, (3) ophthalmic viscoelastic device (OVD) injection, (4) capsulorrhexis formation, (5) hydrodissection, (6) phaco, (7) cortical material removal, (8) intraocular lens (IOL) implantation, (9) OVD removal, (10) IOL centration and (11) wound closure through corneal hydration, and (12) idle phases.	Other: Evaluation test: cataract surgery steps; The development datasets were used to train the deep learning model. The validation and test group were used to optimize hyperparameters
Validation Dataset; 12 cataract surgery steps including(1) main incision formation, (2) side incision formation, (3) ophthalmic viscoelastic device (OVD) injection, (4) capsulorrhexis formation, (5) hydrodissection, (6) phaco, (7) cortical material removal, (8) intraocular lens (IOL) implantation, (9) OVD removal, (10) IOL centration and (11) wound closure through corneal hydration, and (12) idle phases.	Other: Evaluation test: cataract surgery steps; The development datasets were used to train the deep learning model. The validation and test group were used to optimize hyperparameters
Test Dataset; 12 cataract surgery steps including(1) main incision formation, (2) side incision formation, (3) ophthalmic viscoelastic device (OVD) injection, (4) capsulorrhexis formation, (5) hydrodissection, (6) phaco, (7) cortical material removal, (8) intraocular lens (IOL) implantation, (9) OVD removal, (10) IOL centration and (11) wound closure through corneal hydration, and (12) idle phases.	Other: Evaluation test: cataract surgery steps; The development datasets were used to train the deep learning model. The validation and test group were used to optimize hyperparameters

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Accuracy	The investigators will calculate accuracy of deep learning system and compare this index between deep learning system and human doctors	baseline

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
kappa	Cohen's kappa coefficient was calculated to assess the agreement between the grades given by human doctors and DeepSurgery	baseline

Collaborators and Investigators

• Sponsor: Sun Yat-sen University
• Investigators: Study Chair: Yizhi Liu, M.D., Ph.D., Zhongshan Ophthalmic Center, Sun Yat-sen University

Study record dates

Study Major Dates

• Study Start (Actual): January 1, 2019
• Primary Completion (Actual): September 30, 2021
• Study Completion (Actual): December 30, 2021

Study Registration Dates

• First Submitted: February 27, 2022
• First Submitted That Met QC Criteria: February 27, 2022
• First Posted (Actual): March 2, 2022

Study Record Updates

• Last Update Posted (Actual): March 2, 2022
• Last Update Submitted That Met QC Criteria: February 27, 2022
• Last Verified: February 1, 2022

More Information

Terms related to this study

• Keywords: cataract surgery
• Additional Relevant MeSH Terms: Eye Diseases; Lens Diseases; Cataract
• Other Study ID Numbers: CS-2022

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No