Article

Machine Learning Helps Identify Eyes With Pathologic Myopia

Author(s):

Using a support vector machine model, researchers were able to distinguish between eyes with pathologic myopia and healthy myopia.

According to a study published in Scientific Reports, researchers developed a model that provides an accurate modality for identification of patients with pathologic myopia and may help prioritize these patients for further treatment.

Myopia (nearsightedness) complications are a major cause of visual impairment around the world, while eyes with pathologic myopia can develop ocular pathologies in the macula, peripheral retina and optic nerve, authors wrote.

The introduction of fundus photography has led to many suggested methods of differentiating pathologic myopic eyes. However, several limitations are present in the methods, and there are currently no quantitative specifications of pathologic myopia that fully describe the condition of the posterior globe.

To address this knowledge gap, investigators proposed an easy-to-use and clinically available model to identify these patients based on quantitative measurements of the posterior globe.

“This method uses the fovea, optic disc, and deepest point of the eye (DPE) as the 3 major markers (i.e., key indicators) of the posterior globe to quantify the relative tomographic elevation of the posterior sclera (TEPS),” they explained.

A total of 860 patients recruited from 2 tertiary ophthalmology institutes in Korea were included in the analysis (n = 728 with healthy myopia eyes, n=132 with pathologic myopia eyes). Different machine learning classifiers (n = 11) were tested to see if they could discriminate the presence of pathologic myopia. Patients with additional retinal or choroidal disorders were excluded from the study.

“The International Myopia Institute defines pathologic myopia as an excessive axial elongation associated with myopia that leads to structural changes in the posterior segment of the eye (including posterior staphyloma, myopic maculopathy, and high myopia-associated optic neuropathy) as well as a loss in best-corrected visual acuity,” authors said.

After patients underwent comprehensive clinical evolutions, 2 authors agreed on patients’ diagnosis and classification of posterior staphyloma using stereoscopic fundus photography. Two ophthalmologists also determined the designation of healthy myopia or pathologic myopia.

Mean (SD) patient age was 52.43 (14.14) years while the majority (59.3%) of patients were male. Patients with pathologic myopia tended to be significantly older, had significantly longer axial length (AxL), and exhibited significantly worse best-corrected visual acuity.

Analyses revealed:

  • The support vector machine (SVM)-based machine learning classifier predicted pathologic myopia with an area under the receiver operating characteristic curve (AUROC) of 0.828, with 77.5% sensitivity and 88.07% specificity.
  • AxL and choroidal thickness, the existing quantitative indicators of pathologic myopia, only reached an AUROC of 0.758, with 75.0% sensitivity and 76.61% specificity
  • When all 6 indices were applied (4 TEPS, AxL, and subfoveal choroidal thickness [SCT]), the discriminative ability of the SVM model demonstrated an AUROC of 0.868, with 80.0% sensitivity and 93.58% specificity

Notably, the SVM model is based on measurable parameters that can be collected in ophthalmology clinics where optical coherence tomography is available. Although AxL and refractive errors (RE) are the “gold standards for representation of myopia,” researchers found “the use of AxL and RE for pathologic myopia discrimination presents a risk of missing pathologic eyes that may not have a long AxL or RE.” Results also showed SCT alone did not have any discriminative ability.

The retrospective design of the study and the fact the TEPS index only described a portion of the sclera curvature mark limitations to the study. Results also may not be generalizable to the wider population as patients included had a high prevalence of pathologic myopia and were mostly of Korean ethnicity.

“Future studies and machine learning algorithm development will focus on validation of our model with respect to community-based populations and multi-ethnic groups,” researchers concluded.

Reference

Kim YC, Chang DJ, Park SJ, et al. Machine learning prediction of pathologic myopia using tomographic elevation of the posterior sclera. Sci Rep. Published online March 26, 2021. doi:10.1038/s41598-021-85699-0

Related Videos
Quint Petris
Mina Massaro-Giordano, MD
Quint and Petris
Quint Petris
Image of the UPMC eyeVan
Image of the UPMC eyeVan
Screenshot of Byron Lam, MD
Screenshot of Byron Lam, MD
Screenshot of Byron Lam, MD
AJMC Managed Markets Network Logo
CH LogoCenter for Biosimilars Logo