Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

RLPeri: Accelerating Visual Perimetry Test with Reinforcement Learning and Convolutional Feature Extraction

Tanvi Verma, Linh Le Dinh, Nicholas Tan, Xinxing Xu, Chingyu Cheng, Yong Liu

TL;DR

RLPeri, a reinforcement learning-based approach to optimize visual perimetry testing that results in a 10-20% reduction in examination time while maintaining the accuracy as compared to state-of-the-art methods.

Abstract

Visual perimetry is an important eye examination that helps detect vision problems caused by ocular or neurological conditions. During the test, a patient's gaze is fixed at a specific location while light stimuli of varying intensities are presented in central and peripheral vision. Based on the patient's responses to the stimuli, the visual field mapping and sensitivity are determined. However, maintaining high levels of concentration throughout the test can be challenging for patients, leading to increased examination times and decreased accuracy. In this work, we present RLPeri, a reinforcement learning-based approach to optimize visual perimetry testing. By determining the optimal sequence of locations and initial stimulus values, we aim to reduce the examination time without compromising accuracy. Additionally, we incorporate reward shaping techniques to further improve the testing performance. To monitor the patient's responses over time during testing, we represent the test's state as a pair of 3D matrices. We apply two different convolutional kernels to extract spatial features across locations as well as features across different stimulus values for each location. Through experiments, we demonstrate that our approach results in a 10-20% reduction in examination time while maintaining the accuracy as compared to state-of-the-art methods. With the presented approach, we aim to make visual perimetry testing more efficient and patient-friendly, while still providing accurate results.

RLPeri: Accelerating Visual Perimetry Test with Reinforcement Learning and Convolutional Feature Extraction

TL;DR

RLPeri, a reinforcement learning-based approach to optimize visual perimetry testing that results in a 10-20% reduction in examination time while maintaining the accuracy as compared to state-of-the-art methods.

Abstract

Visual perimetry is an important eye examination that helps detect vision problems caused by ocular or neurological conditions. During the test, a patient's gaze is fixed at a specific location while light stimuli of varying intensities are presented in central and peripheral vision. Based on the patient's responses to the stimuli, the visual field mapping and sensitivity are determined. However, maintaining high levels of concentration throughout the test can be challenging for patients, leading to increased examination times and decreased accuracy. In this work, we present RLPeri, a reinforcement learning-based approach to optimize visual perimetry testing. By determining the optimal sequence of locations and initial stimulus values, we aim to reduce the examination time without compromising accuracy. Additionally, we incorporate reward shaping techniques to further improve the testing performance. To monitor the patient's responses over time during testing, we represent the test's state as a pair of 3D matrices. We apply two different convolutional kernels to extract spatial features across locations as well as features across different stimulus values for each location. Through experiments, we demonstrate that our approach results in a 10-20% reduction in examination time while maintaining the accuracy as compared to state-of-the-art methods. With the presented approach, we aim to make visual perimetry testing more efficient and patient-friendly, while still providing accurate results.
Paper Structure (24 sections, 8 equations, 5 figures, 3 tables, 2 algorithms)

This paper contains 24 sections, 8 equations, 5 figures, 3 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Background
  3. Perimetry
  4. ZEST Procedure
  5. Reinforcement Learning
  6. Knowledge Based Learning
  7. RLPeri: Reinforcement Learning Based Perimetry
  8. State Space and State Features
  9. Action Space
  10. Reward Function
  11. Experiments
  12. Dataset
  13. Simulator
  14. Evaluation Metric
  15. Hyperparameter and Computing Infrastructure
  16. ...and 9 more sections

Figures (5)

  • Figure 1: Three decision-making steps of visual field testing.
  • Figure 2: \ref{['fig:sensitivity']} The inverse relationship between light intensity and sensitivity to light. Individuals with a high sensitivity to light are able to detect even faint lights, while those with low sensitivity to light can only see very bright lights. \ref{['fig:fos']} Frequency-of-seeing (FOS) curve. Sensitivity threshold is defined as the intensity at which probability of seeing is 50%. Note the inverse relationship between sensitivity and stimulus intensity. \ref{['fig:vf']} 2D mapping of sensitivity threshold values for 54 locations of 24-2 test pattern. Darker color indicates low sensitivity.
  • Figure 3: The test's status is depicted using a pair of 3D matrices, encompassing the patient's responses. Each matrix is designated for seen responses and not seen responses, respectively. Group-wise and point-wise convolutional kernels are employed to capture features related to the threshold values of the surrounding locations as well as the patient's responses to different stimulus values presented at a given location.
  • Figure 4: Framework for RLPeri. Q-values are learned independently for each action dimension.
  • Figure 5: Ground truth values, reconstructed visual fields, initial stimulus presented, sequence of locations and number of stimuli presented at each location for testing of three different visual fields A, B and C.