SVP-ML: SVP Detection Using Machine Learning
Study Details
Study Description
Brief Summary
This diagnostic study will use 410 retrospectively captured fundal videos to develop ML systems that detect SVPs and quantify ICP. The ground truth will be generated from the annotations of two independent, masked clinicians, with arbitration by an ophthalmology consultant in cases of disagreement.
Condition or Disease | Intervention/Treatment | Phase |
---|---|---|
|
Study Design
Arms and Interventions
Arm | Intervention/Treatment |
---|---|
Patients aged ≥18 years with presumed normal intracranial pressure
|
Diagnostic Test: Machine Learning Model
Automated machine learning system for the detection of spontaneous venous pulsations and quantification of intracranial pressure
|
Patients aged ≥18 years with suspected raised intracranial pressure
|
Diagnostic Test: Machine Learning Model
Automated machine learning system for the detection of spontaneous venous pulsations and quantification of intracranial pressure
|
Outcome Measures
Primary Outcome Measures
- Area-under-the receiver operating characteristic (AUROC) for spontaneous venous pulsations detection [1 year]
Binary classification performance of the machine learning model
Secondary Outcome Measures
- Localisation of spontaneous venous pulsations [1 year]
Bounding box overlap for the machine learning model
- Quantification of intracranial pressure [1 year]
Mean absolute error for the prediction of the intracranial pressure
Eligibility Criteria
Criteria
Inclusion Criteria:
-
Patients aged ≥18 years with presumed normal ICP undergoing routine dilated OCT scans.
-
Patients undergoing a LP or continuous ICP monitoring with implanted transcranial pressure transducer devices at in- or out-patient neurology, neurosurgery or neuro-ophthalmology services.
Exclusion Criteria:
-
Glaucoma diagnosis or glaucoma suspects in either eye.
-
Bilateral restricted fundal view, e.g. advanced bilateral cataracts.
-
Bilateral retinal vein or artery occlusion.
Contacts and Locations
Locations
No locations specified.Sponsors and Collaborators
- King's College London
Investigators
None specified.Study Documents (Full-Text)
None provided.More Information
Publications
None provided.- 1.0