Development and Validation of Deep Neural Networks for Blinking Identification and Classification

Overview

Primary objective of this study is the development and validation of a system of deep neural networks which automatically detects and classifies blinks as "complete" or "incomplete" in image sequences.

This method is based on iris and sclera segmentation in both eyes from the acquired images, using state of the art deep learning encoder-decoder neural architectures (DLED). The sequence of the segmented frames is post-processed to calculate the distance between the eyelids of each eye (palpebral fissure) and the corresponding iris diameter. Theses quantities are temporally filtered and their fraction is subject to adaptive thresholding to identify blinks and determine their type, independently for each eye. The two DLEDs were trained with manually segmented images and the post-process was parameterized using a 4-minute video. After DLED training, the proposed system was tested on 8 different subjects, each one with a 4-10-minute video. Several metrics of blink detection and classification accuracy were calculated against the ground truth, which was generated by 3 independent experts, whose conflicts were resolved by a senior expert. Two independent blink identifications are assumed to be in agreement, if and only if there is sufficient temporal overlapping and the type of blink is the same between the DLED system and the ground truth.

Conditions

Interventions

Eligibility

Locations (2)

Outcomes