Access Type
Open Access Embargo
Date of Award
January 2020
Degree Type
Dissertation
Degree Name
Ph.D.
Department
Electrical and Computer Engineering
First Advisor
Abhilash Pandya
Abstract
Incorporating visual intelligence within surgical robotics can enhance the performance of robotic surgeries. During minimally invasive surgery (either robotic or traditional laparoscopic), vascular injuries may occur because of inadvertent surgical tool movements or actions. These vascular injuries can lead to arterial or venous bleeding with varying degrees of severity that may be life‐threatening. Given that a bloody spot is characterized by homogenous and uniform texture, our algorithm automatically scans the entire surgical video frame‐ by‐ frame using a local entropy filter to segment each image into different regions sequence. By comparing changes in entropy in the frame’s sequences, the algorithm detects the moment of bleeding occurrence and its pixel location. In addition, we have observed patterns of abrupt movements that are good indications that bleeding may occur. Our results using surgical videos as input show that the algorithm can detect bleeding within 0.635 s, on average, after their occurrences and locate the bleeding sources within, on average, 2.5% of discrepancy in pixels from their origins. In addition, results show that the algorithm is 88% accurate and 90% precise in predicting bleeding. The average error of prediction time is 0.662 seconds. The validated preliminary results based on various recorded robotic and laparoscopic videos show that our system can predict unexpected bleeding and detect and localize the bleeding in the early stages.
Recommended Citation
Daneshgar Rahbar, Mostafa, "Visual Intelligence For Robotic And Laparoscopic Surgery: A Real-Time System For Bleeding Detection And Prediction" (2020). Wayne State University Dissertations. 2482.
https://digitalcommons.wayne.edu/oa_dissertations/2482