Ocular Diseases
Conditions
Keywords
ocular diseases, Ophthalmic Multimodal AI-Assisted Medical Decision-Making, Artificial Intelligence
Brief summary
This is a multi-center, retrospective clinical study designed to evaluate the application and effectiveness of an AI-assisted medical decision support system, leveraging multimodal data fusion, in ophthalmic clinical practice.
Detailed description
Visual impairments significantly affect an individual's quality of life. Early screening, diagnosis, and treatment of ocular diseases are crucial for preventing the onset and progression of vision disorders. In clinical practice, ophthalmologists often need to integrate a wide range of patient data, including demographic information, medical history, biochemical markers such as blood glucose and lipid levels, risk factors, as well as various ophthalmic data, such as fundus images, OCT scans, and visual field tests, to make an accurate diagnosis and develop an appropriate treatment plan. In an era where precision and personalized medicine are at the forefront of healthcare, the early detection and diagnosis of eye diseases, as well as the selection of suitable diagnostic and therapeutic strategies at different stages of the disease, have become significant challenges in clinical settings. Recent advancements in medical imaging and analysis techniques have greatly enhanced the accuracy and effectiveness of ocular disease diagnosis. This study aims to develop an ophthalmic artificial intelligence-assisted decision-making system by integrating multimodal data from imaging and electronic medical records, in combination with deep learning techniques. The objective is to improve diagnostic accuracy, streamline clinical workflows, and provide more personalized treatment options for patients. Ultimately, this system seeks to enhance treatment outcomes and improve the overall quality of life for patients suffering from ocular diseases.
Interventions
DIAGNOSTIC_TESTDiagnostic Test: AI-Based Diagnostic and Prognostic Model for Ocular Diseases
This intervention involves an AI system that leverages multimodal data fusion to support the clinical decision-making and evaluation of ophthalmic diseases. It integrates multi-modal data, including fundus photography, optical coherence tomography (OCT), and patient clinical records, to provide real-time, precise, and personalized diagnostic support. Unlike other models, this system utilizes a longitudinal patient dataset to predict disease progression and treatment outcomes.Key distinguishing features include: 1. Multi-Modal Data Integration: Combines imaging, clinical, and genetic data for comprehensive analysis. 2. Predictive Capability: Offers advanced prognostic predictions, enabling personalized treatment plans. 3. Deep Learning Framework: Employs state-of-the-art deep learning algorithms for improved diagnostic accuracy and efficiency. 4. Real-World Validation: Validated using a large cohort of diverse patient data, ensuring generalizability and robustness.
Sponsors
The Eye Hospital of Wenzhou Medical University
Study design
Observational model
CASE_ONLY
Time perspective
RETROSPECTIVE
Eligibility
Sex/Gender
ALL
Healthy volunteers
No
Inclusion criteria
1.All patients who have received treatment at multiple centers, including The Eye Hospital of Wenzhou Medical University, First Affiliated Hospital of Wenzhou Medical University, Second Affiliated Hospital of Wenzhou Medical University, ZhuHai Hospital, and Macau University of Science and Technology Hospital. 2.Availability of comprehensive electronic health records (EHR), including: Ophthalmic images (e.g., fundus photography, OCT, or slit-lamp images). Electronic medical records (e.g., diagnosis, treatment, and follow-up notes). Examination results (e.g., visual acuity, intraocular pressure, or laboratory tests). 3.Patients with a clear and confirmed diagnosis of one or more ocular diseases. 4.Patients with sufficient follow-up records to allow assessment of disease progression or prognosis, if applicable. 1. All ophthalmology patients who have previously received treatment at the Department of Ophthalmology, the Eye Hospital of Wenzhou Medical University, First Affiliated Hospital of Wenzhou Medical University, Second Affiliated Hospital of Wenzhou Medical University, Zhuhai People's Hospital, and the University Hospital. 2. Availability of comprehensive electronic health records (EHR), including: Ophthalmic images (e.g., fundus photography, OCT, or slit-lamp images). Electronic medical records (e.g., diagnosis, treatment, and follow-up notes). Examination results (e.g., visual acuity, intraocular pressure, or laboratory tests). 3. Patients with a clear and confirmed diagnosis of one or more ocular diseases. 4. Patients with sufficient follow-up records to allow assessment of disease progression or prognosis, if applicable.
Exclusion criteria
1. Incomplete or missing critical EHR components. 2. Cases with ambiguous or unverified diagnoses that cannot be clearly categorized. 3. Duplicated or redundant data from the same patient.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Decision Time Efficiency | 1 years | Average time (seconds) required for physicians to make diagnostic and treatment decisions, before and after AI assistance. |
| Specificity | 1 years | Specificity (also called True Negative Rate) measures the proportion of actual negative cases correctly identified by the model. No unit (a ratio or percentage, typically expressed as a percentage). |
| False Positive Rate | 1 years | False Positive Rate (FPR) measures the proportion of actual negative cases that are incorrectly identified as positive by the model. No unit (a ratio or percentage, typically expressed as a percentage). |
| False Negative Rate | 1 years | False Negative Rate (FNR) measures the proportion of actual positive cases that are incorrectly identified as negative by the model. No unit (a ratio or percentage, typically expressed as a percentage). |
| Postoperative Complication Rate | 1 years | Percentage (%) of patients experiencing postoperative complications. |
| Recurrence Risk Rate | 1 years | Percentage (%) of patients experiencing recurrence during the follow-up period. |
| Survival Rate | 1 years | Percentage (%) of patients alive, calculated using Kaplan-Meier survival curves. |
| Effectiveness of Decision Support | 1 years | Percentage (%) improvement in the accuracy of treatment decisions with AI assistance compared to traditional decisions. |
| Area Under the Curve (AUC) | 1 years | AUC of the ROC curve, used to quantify diagnostic accuracy. No unit (a ratio or percentage, typically expressed as a number between 0 and 1). |
| Sensitivity | 1 years | Sensitivity (also called True Positive Rate) is a measure of how well a model identifies positive instances. It is defined as the proportion of actual positive cases correctly identified by the model. No unit (a ratio or percentage, typically expressed as a percentage). |
| Accuracy Accuracy Accuracy | 1 years | Accuracy measures the proportion of all correct predictions (true positives and true negatives) out of the total number of cases evaluated by the model. No unit (a ratio or percentage, typically expressed as a percentage). |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| AI System Response Time | 1 years | Average time (seconds) taken for the AI to provide recommendations after data input. |
| System Failure Rate | 1 years | Frequency of AI system failures, measured as failures per thousand hours of use (failures/thousand hours). |
| User Interface Design Satisfaction | 1 years | Evaluated using the User Experience Questionnaire (UEQ), with scores ranging from 1-7. |
| Patient Satisfaction Score | 1 years | Measured using the Patient Satisfaction Questionnaire (CSQ-8), with scores ranging from 8-32. |
| Treatment Adherence | 1 years | Percentage (%) of patients adhering to personalized treatment plans and regular follow-up visits. |
| Physician Acceptance of AI System | 1 years | Evaluated using the Technology Acceptance Model (TAM) scale, with scores ranging from 1-7. |
| System Usability Score | 1 years | Evaluated using the System Usability Scale (SUS), with scores ranging from 0-100. |
Countries
China, Macau
Contacts
Primary ContactLan Wang, MD
Outcome results
None listed