Artificial Intelligence (AI), Cataract
Conditions
Brief summary
Cataract is a major cause of blindness due to eye diseases. Methods for evaluating the degree of lens opacification in cataracts are divided into subjective and objective methods. The commonly used subjective method is the Lens Opacification Classification System (LOCS Ⅲ), while the objective methods mainly include the Dysfunctional Lens Index (DLI) of the Ray Tracing aberration analysis system, the PNS score of the Pentacam anterior segment analysis system, etc. Subjective diagnosis may lead to certain misjudgments, which have affected clinical diagnosis and treatment. There is an urgent need to add objective diagnostic measures to assist clinical work. The Scanning Source Optical Coherence Tomography (SS - OCT) biometer - IOL Master 700 forms an OCT imaging of the eye based on the swept - source optical coherence tomography (OCT) biometric technology. It can visually show the longitudinal section of the entire lens, and the clear display of the patient's lens tomographic OCT image is obtained through image visualization measurement. The main purpose of this study is to analyze the lens images obtained by the IOLmaster 700. Based on the current mainstream algorithm models such as ResNet - 34 and XGBoost, develop a heterogeneous accelerated artificial intelligence algorithm according to our research needs to accurately calculate the degree of lens opacification. And write image analysis software by ourselves to automatically calculate the required indicators and output them. Establish a heterogeneous accelerated artificial intelligence - assisted lens opacification grading and prediction system, supporting software for biometer equipment, and a cataract lens image database. The software provides online service functions, and all researchers can use the image analysis function of the software after logging in, truly realizing the sharing of large instrument supporting software operations. Thereby improving the accuracy and efficiency of clinical diagnosis and treatment, the prognostic prediction level of patients after cataract surgery, guiding clinical diagnosis and treatment more accurately, and at the same time, it can be used as a tool for community screening.
Interventions
DEVICEIOL-MASTER 700
patients who were diagnosed cataract would go through tests with IOL-MASTER 700 to achieve ocular biometry parameters.
Sponsors
Shanghai 10th People's Hospital
Study design
Observational model
COHORT
Time perspective
RETROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
40 Years to No maximum
Healthy volunteers
No
Inclusion criteria
* A. Age between 18 and 90 years B. Diagnosed with age-related and/or complicated cataract (diagnosed according to LOCS III classification) C. The patient has signed an informed consent form
Exclusion criteria
* A. Exclude patients with corneal diseases, uveitis, vitreoretinal diseases, or refractive media opacities caused by conditions such as retinal detachment with silicone oil tamponade B. History of previous ophthalmic disease treatment or surgery C. Poor-quality or missing imaging data D. Pupil diameter \< 2.5 mm or loss of fixation during examination, resulting in inability to obtain sufficient lens data
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Cataract Grade | 3 months | Achieved cataract images from IOL-MASTER 700 will be graded by several experienced doctors. And each image will be graded in three parts: cortical, nuclear and posterior subcapsular opacification. |
| AI predicted cataract grade | 3 months | Achieved cataract images from IOL-MASTER 700 will be used to train, validate and test the AI algorithm. Predicted grade will be carried out and each image will be graded by AI in three parts: cortical, nuclear and posterior subcapsular opacification. |
| AI model performance | 3 months | To evaluate the model performance, several indices are introduced: Sensitivity, Specificity, Accuracy, Precision, F1 score, receiver operating characteristic curve and area under the curve. |
Contacts
Primary ContactYiwen Hu
Outcome results
None listed