Clinical Relevance of Choroidal Thickness in Obese and Healthy Children: A Machine Learning Study

Erkan Bulut1, Sümeyra Köprübaşı2, Özlem Dayi3

  • 1Gelişim University, Vocational School of Health Services, Department of Opticianry, İstanbul, Türkiye.

Insights

Obesity impacts children's choroidal thickness, particularly in the subfoveal and outer peripapillary regions. Machine learning algorithms like Random Forest and Support Vector Machine accurately classify obese children based on these measurements.

Area of Science:

  • Ophthalmology
  • Pediatrics
  • Biomedical Engineering

Background:

  • Childhood obesity is a growing public health concern.
  • Obesity is associated with various systemic health issues, potentially affecting ocular structures.
  • Understanding ocular changes in obese children is crucial for early detection and management.

Purpose of the Study:

  • To investigate the effect of macular choroidal thickness (MCT) and peripapillary choroidal thickness (PPCT) on classifying obese versus healthy children.
  • To compare the performance of Random Forest (RF), Support Vector Machine (SVM), and Multilayer Perceptron (MLP) algorithms in this classification task.

Main Methods:

  • Prospective comparative study involving 59 obese and 35 healthy children (aged 6-15 years).
  • Optical coherence tomography (OCT) used to measure MCT and PPCT at various distances (500, 1000, 1500 μm) from the fovea and optic disc.
  • Feature selection algorithms identified key differentiating features, followed by classification using RF, SVM, and MLP.

Main Results:

  • The correlation feature selection algorithm yielded the best results.
  • Key features for distinguishing groups included PPCT at temporal 500 μm and 1500 μm, nasal 1500 μm, inferior 1500 μm, and subfoveal MCT.
  • Classification accuracies were: RF 98.6%, SVM 96.8%, and MLP 89%.

Conclusions:

  • Obesity significantly affects choroidal thickness in children, especially in the subfoveal and outer peripapillary areas.
  • RF and SVM algorithms demonstrate high accuracy and effectiveness in classifying obese children based on choroidal thickness measurements.
Abstract