Computer Aided Auto Measurement Of Pupil Diameter Based On The Graph-Cut And Object Detection

The pupil diameter size and changes are important clinical parameters, regular quantitative monitoring of pupils can reflect physiological arousal, heart rate changes, evaluate the autonomic nervous activity and achieve non-contact non-destructive measurement of heart rate variability. Therefore quantitative analysis and evaluation of pupils has important clinical significance, especially in the emergency and intensive care departments. There are many measured manners in clinic, including doctors’ artificial pupil diameter measurement, infrared pupil measuring instrument, corneal topography, wave-front analyzer and ultrasound bio-microscope. All the methods above require the patient a certain pose and need to be poked eyelid, causing inconvenient and secondary damage in patient’s eyes. Other way, because of the real-time, inexpensive and noninvasive advantages, ultrasonic imaging is widely used in medical imaging. Using ultrasonic imaging to measure pupil diameter can solve the above deficiencies. But because of the speckle noise and low resolution, the method would need an accuracy and stability segmentation method about ultrasonic tissue.This paper proposed an automatic pupil diameter measuring algorithm based on object recognition and graph cuts(Graph-Cut). Firstly, the algorithm used Histogram of Oriented Gradient(HOG) and support vector machine (SVM) to train classifier. Then it used classifier to identify the pupil position, confirming the region of interest (ROI). Secondly, the algorithm used Graph-Cut to segment pupil. Lastly, we can measure pupil diameter in the segmented image. This set of algorithms can be used for continuous pupil image segmentation, producing dynamic curve of pupil diameter.Through statistical analysis we know that the correlation coefficient between the proposed results and the doctor artificial measurement is up to 0.99, and the autocorrelation coefficient of the proposed method is up to 0.999. The results show that the proposed algorithm for quantitative clinical examination and monitoring of pupil diameter has high feasibility and good stability, and it can be used to monitor the dynamic pupil diameter.