University of Twente Student Theses

Towards RObot assisted iMage guided surgery using deeP lEarning Techniques (TROMPET): the calibration of flexible instruments for stereotactic navigation in locally advanced and recurrent rectal carcinoma

Brink, R.S.A. ten (2023) Towards RObot assisted iMage guided surgery using deeP lEarning Techniques (TROMPET): the calibration of flexible instruments for stereotactic navigation in locally advanced and recurrent rectal carcinoma.

Full text not available from this repository.

Full Text Status:	Access to this publication is restricted
Abstract:	Introduction The current rate of irradical resections in primary locally advanced and recurrent rectal cancer is unacceptably high. This is associated with a high chance of locoregional recurrence, development of metastases and reduced life expectancy, quality of life, health care efficiency and medical capacity. Locally advanced and recurrent rectal cancer extends beyond the mesorectal fascia which makes it difficult for surgeons to determine where to cross the anatomical excision planes. To give surgeons real-time guidance during surgery, optical stereotactic navigation is currently being investigated as a method of increasing radicality. To allow for future use in robot-assisted surgery, this thesis aims to develop a proof-of-concept embedded system that allows for stereotactic navigation of flexible robotic instruments. Methods The proposed embedded system consists of a rigid endoscope that is tracked using the Brainlab® optical stereotactic system. A custom application is written to track the flexible instruments from the endoscopic video feed in 3D relative to the endoscope. Both coordinates are combined to determine the position of the flexible instrument relative to the patient. This position can then be visualised in a preoperative CT scan. A phantom study of this system was conducted. Results Sending endoscope coordinates is achieved through OpenIGTLink. The flexible instrument is recognized and located in XY by DeepLabCut and located in Z by MiDaS, both of which are both deep learning networks. This process is visualised by 3D slicer. However, accuracy in the Z-axis was low and the position took long to calculate on a CPU. Conclusion This thesis has proven that the developed embedded system, which allows for optical stereotactic navigation of flexible robot instruments, works as a proof-of-concept. However, for clinical use, the accuracy needs to be improved and the calculation speed needs to be increased.
Item Type:	Essay (Master)
Clients:	UMCG, Groningen, Netherlands
Faculty:	TNW: Science and Technology
Subject:	44 medicine, 50 technical science in general
Programme:	Technical Medicine MSc (60033)
Link to this item:	https://purl.utwente.nl/essays/94512
Export this item as:	BibTeX EndNote HTML Citation Reference Manager

Show download statistics for this publication

Repository Staff Only: item control page