University of Twente Student Theses


Optimizing navigation bronchoscopy by improving the biopsy tool and lesion tracking

Stienstra, L. and Stel, C. and Baan, J. and Roelofs, C. (2023) Optimizing navigation bronchoscopy by improving the biopsy tool and lesion tracking.

Full text not available from this repository.

Full Text Status:Access to this publication is restricted
Abstract:Introduction Lung cancer is the most deathly type of cancer in the Netherlands. Navigation bronchoscopy will probably become more important in the diagnostic process for small lesions. This upcoming procedure has several limitations which decrease its efficiency and the diagnostic yield. Objective This paper focuses on the improvement of taking the biopsy by looking at the biopsy tool and the imaging and tracking of the lungs and lesion. The objective of the biopsy tool is to have a more accurate biopsy and to maximize the amount of tissue obtained from the lesion. The objective of the lesion tracking method is to aim for a more accurate biopsy by tracking the lesion during the procedure and by visualizing the respiration-induced motion of the lungs. Method A new biopsy tool is designed and several design aspects were tested. Four different designs are made at two different scales using a resin 3D printer. During the experiment, the prototypes were inserted into a gelatin phantom that should represent lesion tissue. The weight of the samples is analysed and compared to the samples taken by a forceps biopsy tool. Tracking the lung lesion and analyzing the respiration-induced motion is accomplished by the use of a deep learning network and optical flow in Python. Results The impact of the tool design variants on the sample weight is significant. Most of the designs have a significant smaller sample weight than the forceps. The overlap rate for the masks is very high for all masks (98.2 ± 0.398 percent) and the differences in lung size ratio are small (0.0130 ± 0.0078) between the two manually drawn masks and the mask made by the deep learning model. The respiration-induced motion of the lungs can be analyzed with the color plot and color circle. The location of the augmented lesion can be updated frame by frame on the 2D fluoroscopy DICOM with the optical flow-based lesion tracking method. Conclusion The novel biopsy tool design does not take a significantly larger biopsy compared to the forceps biopsy tool in this experiment. As the experiment has several limitations, further research on tool design specifics are necessary. The optical flow-based lesion tracking method based on a deep learning model is a promising method for tracking the lung and its lesion in a 2D fluoroscopy for navigation bronchoscopy. In the future, the deep learning model should be trained with more input images from different angles. The optical flow-based lesion tracking method needs to be further improved and optimized before it can be used in real time in navigation bronchoscopy.
Item Type:Essay (Bachelor)
Radboudumc, Nijmegen, Nederland
Faculty:TNW: Science and Technology
Subject:44 medicine, 50 technical science in general
Programme:Technical Medicine BSc (50033)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page