Abstract
Nowadays, laparoscopic surgery is one of the most used surgeries due to the minimal
invasion to the patient (mainly because of the small incisions performed), which is
why it is included in the group of Minimally Invasive Surgery (MIS). A thin tool
with a video camera and a light at the end is used to obtain images from the tissues
and organs of the patient. Said images are displayed in a monitor allowing the
surgeon to see what is happening inside the body and perform the surgery accurately.
This technique requires for surgeons to have a high-level training, for which new
pedagogical approaches are being developed. One useful technology is the creation
and application of surgical interactive videos for the learning processes of residents.
There are several training platforms which uses interactive videos as part of their
surgical training curriculum (WebSurg, surgical theater. . . ).
However, there are not many authoring tools which allow for the creation of original
contents by the teachers or content creators in charge of the residents' training.
AMELIE is an authoring tool which provides teachers and content creators with the
means for creating interactive videos. Its functionalities include, among others, the
ability to show anatomical structures of importance for the understanding of a procedure.
This is obtained by means of applying computer vision in those structures.
The aim of this project is precisely to implement an improved segmentation and
tracking algorithm based on artificial vision capable of recognizing any anatomical
structure selected by the teachers or content creators interacting with the authoring
tool in a laparoscopic video frame, and following its position during the rest of
the selected frames. This information will lead to an improvement in the training
processes of surgical residents.
The developed algorithm is based on the extraction and matching of image features.
It essentially finds important regions in each frame and compares them with
the initial Region of Interest (ROI) selected by the user to find the new ROI. Those
regions that are similar are recognized as the ROI of the image.
After the implementation of this new tracking method, it has been compared to a
previously used method in the AMELIE authoring tool, obtaining better results at
tracking structures in laparoscopic videos, more concretely a 18,35% better taking
into account the average F1 score of both algorithms. This new method improving
the overall learning experience of surgeons.
