Artificial Intelligence, Facial Plastic and Reconstructive Surgery, Orbital Diseases, Periocular Diseases
Conditions
Keywords
Facial Plastic and Reconstructive Surgery, Periocular Diseases, Orbital Diseases, Artificial Intelligence
Brief summary
Computer vision using deep learning architecture is broadly used in auto-recognition. In the research, the deep learning model which is trained by categorized single-eye images is applied to achieve the good performance of the model in blepharoptosis auto-diagnosis.
Detailed description
This auto-diagnosis system of blepharoptosis using machine learning architecture will assist in telemedicine, such as early screening of childhood ptosis for prompt referral and treatment. People could use this software via mobile devices to get a primitive diagnosis before they reach the physicians. Furthermore, in primary health care, where there is no oculoplastic surgeon, the software could assist primary care physicians or general ophthalmologists, in identifying the need for a referral.
Interventions
None listed
Sponsors
Stanford University
National Taiwan University Hospital
Study design
Observational model
CASE_ONLY
Time perspective
RETROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
20 Years to 65 Years
Healthy volunteers
No
Inclusion criteria
\[Inclusion Criteria\] 1. The participants who were 20-year-old or above, 2. Surgical informed consent was endorsed by the participants themselves, 3. Participants who have surgical indications of the oculofacial surgeries, and 4. The participants who agreed on photograph taking after explanation by the surgeon at outpatient clinics. \[
Exclusion criteria
\] 1. The participants who were 19-year-old or under, 2. The participants who don't have surgical indications of the oculofacial surgeries, 3. The participants who were designed for minimal invasive treatments, such as Botox or any kind of fillers injection, 4. The participants who refused photograph taking for any reason, and 5. The participants who are not available for standard quality of photograph taking, such as bedridden patients.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| The model performance is evaluated by accuracy | Through study completion, an average of 1 year | An Artificial Intelligence Approach |
| AUC (Area Under the Curve) | Through study completion, an average of 1 year | An Artificial Intelligence Approach |
| ROC (Receiver Operating Characteristics) curve. | Through study completion, an average of 1 year | An Artificial Intelligence Approach |
| An Artificial Intelligence Approach to Identifying Facial, Periocular, and Orbital Diseases | Through study completion, an average of 1 year | The model interpretability is accessed by Grad-CAM (Class Activation Maps). |
Countries
Taiwan
Outcome results
None listed