AI-based detection of the inner limit of the minimal waist of the nerve fiber bundles in the ONH in subjects with peripapillary atrophy

• Purpose: To verify that a deep learning model developed by our research group does not wrongly identify the outer edge of the atrophic zone in eyes with peripapillary atrophy (PPA) rather than a point adjacent to the external limit of the minimal waist of the nerve fiber bundles in the ONH.Methods: Subjects with at least one eye with PPA were included. The ONH of the eye with PPA was imaged with colour photography, and the three-dimensional structure of the ONH was captured three times at one occasion with SS-OCT (Topcon Triton, Japan). Each capture was exported to a custom-made software for analysis. The custom-made software, AutoPimd, allowed fully automatic localization of the external limit of the minimal waist of the nerve fiber layer in space using deep learning-model. An en face visualization of the frontal plane of the ONH and user measurement in the en face view is possible. The en face view in the OCT volume was verified to render PPA as imaged in the photograph. The distance between the ONH center and the extreme edge of the PPA was measured in the en face view of the OCT. Then, the frontal plane distance between the ONH center and the fully automatically detected external limit of the minimal waist of the nerve fiber layer in the ONH was measured along the same frontal plane angle.Results: A 95 % confidence interval for the mean difference between the distance from center of the ONH to the extreme edge of the PPA, and the distance from the center of the ONH to the corresponding fully automatically detected outer limit of the minimal waist of the nerve fiber layer was estimated to 692 ± 192 μm (d.f = 5).Conclusion: Our AI model does not wrongly localize the outer limit of PPA as the external limit of the minimal waist of the nerve fiber layer. The structural representation of the external limit of the minimal waist of the nerve fiber bundles localized by our fully automatic AI model in eyes with PPA remains to be identified.

Ämnesord

MEDICIN OCH HÄLSOVETENSKAP  -- Klinisk medicin -- Oftalmologi (hsv//swe)

MEDICAL AND HEALTH SCIENCES  -- Clinical Medicine -- Ophthalmology (hsv//eng)

Nyckelord

ONH

nerve fibers

minimal waist

OCT

artificial intelligence

deep learning

peripapillary atrophy

Ophtalmology

Oftalmiatrik

Publikations- och innehållstyp

vet (ämneskategori)

ovr (ämneskategori)