Abstract
Purpose:
Electromagnetic based navigated bronchoscopy using preoperative CT images has reached the clinic during the last decade. One of the challenges is the “CT to patient anatomy alignment” of the CT images acquired days or even weeks ahead of bronchoscopy. An automatic registration method, without manual registration of anatomical landmarks, was developed, implemented, and evaluated in the current study.
Methods:
The registration method aligns automatically the preoperative CT images to the patient's anatomy during the initial part of the bronchoscopy. The algorithm is a modified version of an iterative closest point algorithm, which in addition to the positions also utilizes the orientation of the bronchoscope and the running direction of the CT centerline. The method was evaluated both by model-based simulated bronchoscopies and by clinical data from six real bronchoscopies. In the clinical evaluation, an electromagnetic position sensor was placed temporarily in the working channel close to the tip of a conventional bronchoscope. Position data, which were acquired while the bronchoscope was moving inside the airways, were registered to the centerline extracted from the airways in the CT image.
Results:
A mean registration accuracy of 3.0 ± 1.4 mm was found when simulating bronchoscopies. In the clinical part of the study, the registration method was successfully applied to the data from all six patients. The positions of the bronchoscope tip aligned to the CT centerline with a mean distance range 4.7–6.5 mm.
Conclusions:
The authors have developed and evaluated an automatic registration algorithm for electromagnetic navigated bronchoscopy. It functioned to its purpose and did not affect the workflow for the bronchoscopic investigation of the six patients included in the study.
Electromagnetic based navigated bronchoscopy using preoperative CT images has reached the clinic during the last decade. One of the challenges is the “CT to patient anatomy alignment” of the CT images acquired days or even weeks ahead of bronchoscopy. An automatic registration method, without manual registration of anatomical landmarks, was developed, implemented, and evaluated in the current study.
Methods:
The registration method aligns automatically the preoperative CT images to the patient's anatomy during the initial part of the bronchoscopy. The algorithm is a modified version of an iterative closest point algorithm, which in addition to the positions also utilizes the orientation of the bronchoscope and the running direction of the CT centerline. The method was evaluated both by model-based simulated bronchoscopies and by clinical data from six real bronchoscopies. In the clinical evaluation, an electromagnetic position sensor was placed temporarily in the working channel close to the tip of a conventional bronchoscope. Position data, which were acquired while the bronchoscope was moving inside the airways, were registered to the centerline extracted from the airways in the CT image.
Results:
A mean registration accuracy of 3.0 ± 1.4 mm was found when simulating bronchoscopies. In the clinical part of the study, the registration method was successfully applied to the data from all six patients. The positions of the bronchoscope tip aligned to the CT centerline with a mean distance range 4.7–6.5 mm.
Conclusions:
The authors have developed and evaluated an automatic registration algorithm for electromagnetic navigated bronchoscopy. It functioned to its purpose and did not affect the workflow for the bronchoscopic investigation of the six patients included in the study.
