Deep Learning for Classification of Scheimpflug Corneal Tomography Images

Classification of Color-Coded Scheimpflug Camera Corneal Tomography Images Using Deep Learning

Keratoconus is a common disorder. An early diagnosis influences the disease prognosis in the affected patients and prevents postoperative complications in patients with keratoconus considering refractive surgery. Machine learning approaches have been widely used for image classification. Here, we will assess the ability of deep learning to enable high-performance image classification of the color-coded corneal maps obtained by Scheimpflug camera in patients with keratoconus, subclinical keratoconus, and normal individuals.

Inclusion Criteria:

Keratoconus group:

• Presence of a central protrusion of the cornea with Fleischer ring, Vogt striae, or both by slitlamp examination.
• irregular cornea determined by distorted keratometry mires and distortion of retinoscopic red reflex or both in addition to the following topographic findings as summarized by Pińero and colleagues :
• focal steepening located in a zone of protrusion surrounded by concentrically decreasing power zones
• focal areas with diopteric (D) values >47.0D
• inferior- superior(I-S) asymmetry measured to be > 1.4 D
• Angling of the hemimeridians in an asymmetric or broken bowtie pattern with skewing of the steepest radial axis(SRAX) .

Suspicious group:

• Defined as subtle corneal tomographic changes as the aforementioned keratoconus abnormalities in the absence of slit- lamp or visual acuity changes typical of keratoconus (forme fruste keratoconus).

Normal group:

• Refractive surgery candidates
• Refractive error of less than 8.0 D sphere
• Less than 3.0 D of astigmatism
• without clinical, topographic or tomographic signs of keratoconus or suspect keratoconus.

Exclusion Criteria:

• Systemic disease
• Other corneal disease such as pellucid marginal degeneration
• History of trauma
• Corneal surgery such as corneal cross- linking for progressive keratoconus.