v1.0.0
Which signs of diseases does RetinaLyze OCT look for?
RetinaLyze OCT is able to detect more than 70 pathological signs. The algorithm analyzes the macula area and is able to detect signs of the following retinal conditions (mainly):
Asteroid Hyalosis
Atrophic AMD
Branch Central Retinal Artery Occlusion
Branch Retinal Vein Occlusion
Central Retinal Artery Occlusion
Central Retinal Vein Occlusion
Central Serous Retinopathy
Chorioretinitis
Choroidal rupture
Classic Choroidal Neovascularization
Cystoid Macular Edema
Degenerative Myopia
Diabetic Retinopathy
Diabetic Macula Edema
Diffuse Macular Edema
Drusen
Dry Age-Related Macular Degeneration
Epiretinal Fibrosis
Epiretinal Hemorrhages
Geographic Atrophy
Intraretinal Hemorrhages
Lamellar Macular Hole
Macular Hole
Macular Telangiectasia
Maculopathy
Multifocal Choroiditis
Myopia
Neurosensory Epithelium Detachment
Pigment Choroidal Nevus
Pigment Epithelium Detachment
Pseudohole
Retinal Detachment
Retinitis Pigmentosa
Solar Maculopathy
Subretinal Fibrosis
Subretinal Hemorrhages
Tapetoretinal Dystrophy
Vitelliform Dystrophy
Vitreo-Macular Traction Syndrome
Vitreous Detachment
Wet Age-Related Macular Degeneration
I get the "Scan could not be analyzed" screening result - What do I do?
This result indicates a less than ideal scan quality. Please capture a new OCT scan photo (making sure that the macula is present in the image) and then analyze the new scan.
How does RetinaLyze OCT work?
RetinaLyze OCT is trained using AI-techniques and uses analysis of the retinal structure to detect anomalies. The software has been trained using anonymized OCT retinal scans collected over more than 12 years.
The training set consists of 500.000 b-scans of Caucasians/Europeans (100.000 b-scans before augmentation). The scans were genereated using multiple OCTs from leading manufacturers such as Zeiss, Optopol, Topcon, Huvitz, Nidek and Canon.
The training dataset was annotated with 8 separate ophthalmologists. All b-scans were annotated by at least two ophthalmologists, and a third ophthalmologist would decide the result, if they disagreed.
How has the performance of the algorithm been validated?
The RetinaLyze OCT algorithm has been validated through internal studies. The model shows a high level of performance with multi-disease classification accuracy more than 96%. Find the latest studies on our Research page.
Which factors can affect the use of the algorithm and the validity of the result?
Important guidelines about scan quality
It is very important that a high scan quality is maintained to ensure a reliable screening result. Please see common causes for poor scan quality below:
Small pupil size, Dry eye and Lens opacity/Cataract
Small pupil size, dry eye and cataract are the main reasons for poor scan quality since they may reduce the signal-to-noise ratio (introduce a lot of noise in the scan). Patients with slightly blurred scans from initial cataracts can be analyzed.
Floaters and other vitreous opacities
Floaters and other vitreous opacities may decrease scan quality by interfering with the light beam path. If these floaters are positioned within the scan area (specifically in the macula area), a new scan is recommended.
Patients blinking
Patients blinking is another common cause of poor scan quality. Patients should be encouraged to blink a few times immediately before scan capture to ensure uniform tear film distribution and to preserve adequate scan quality. This may also improve patient comfort, thus decreasing the likelihood of blinks or motion artifacts during scan acquisition. However, in some cases, artificial tears or gel may be necessary for adequate ocular surface lubrication.
Patients moving
Similarly motion artifacts (from the patient moving their eye or head) can contribute heavily to noise and/or misalignment. Please capture a new scan.
The system will automatically reject the scan if the scan quality is too poor.
Any screening or diagnostic system has a limited degree of accuracy, sensitivity and specificity. RetinaLyze OCT is quite resistant to moderate defects in scan quality, but its results will be optimal by selecting the best possible scans, just as with manual grading.
Can RetinaLyze OCT run offline/in an on-premise environment?
RetinaLyze OCT is only available online at the moment. If offline/on-premise operation is a necessity, please contact us.
What type of scans/OCT devices can be used with the algorithm?
RetinaLyze OCT is powered by http://Altris.ai