Peltrini, Roberto (2023) Artificial Intelligence-assisted Near-INfrared FluorEscence Angiography with Indocyanine green after colorectal resections: The NINFEA study. [Tesi di dottorato]

Anteprima

Testo Peltrini_Roberto_36.pdf Download (1MB) | Anteprima

Abstract

As the uniformity of the brightness in indocyanine green based fluorescence (ICG) at the time of the anastomosis construction consists only in a qualitative, empirical evaluation, a machine learning algorithm to automatically assess blood perfusion of the bowel stump using NIR ICG angiography was proposed. Afterwards, the implementation of the algorithm provided the intraoperative identification of the optimal transection line.The algorithm adopts a Feed Forward Neural Network receiving as input a feature vector based on the histogram of the green band of the input image. In particular, the algorithm provides an output that classifies the perfusion as adequate or inadequate. It was used to acquire information related to perfusion during laparoscopic colorectal surgery and suggest ed the section point based on the ICG intensity support ing surgeon s in assessing objectively the procedure. The algorithm was validated on videos captured during surgical procedures carried out at the University Hospital Federico II in Naples. The results show a classification accuracy of 99.9%, with repeatability of 1.9%. Finally, the real time operation of the proposed algorithm was tested by analysing the video streaming captured directly from an endoscope available in the operating room.

Downloads

Actions (login required)

Modifica documento