[Navigation-assisted sonography for soft tissues in the head and neck region]

[Development and evaluation of ultrasound navigation for free-hand biopsies of small masses in the head and neck area]

BACKGROUND

Ultrasound is an important imaging method in the head and neck area. It is readily available, dynamic, inexpensive, and does not involve radiation exposure. Interventions in the complex head and neck anatomy require good orientation, which is supported by navigation systems.

OBJECTIVE

This work aimed to develop a new ultrasound-controlled navigation system for taking biopsies of small target structures in the head and neck region.

METHODS

A neck phantom with sonographically detectable masses (size: 8-10 mm) was constructed. These were automatically segmented using a ResNet-50-based deep neural network. The ultrasound scanner was equipped with an individually manufactured tracking tool.

RESULTS

The positions of the ultrasound device, the masses, and a puncture needle were recorded in the world coordinate system. In 8 out of 10 cases, an 8‑mm mass was hit. In a special evaluation phantom, the average deviation was calculated to be 2.5 mm. The tracked biopsy needle is aligned and navigated to the masses by auditory feedback.

CONCLUSION

Outstanding advantages compared to conventional navigation systems include renunciation of preoperative tomographic imaging, automatic three-dimensional real-time registration that considers intraoperative tissue displacements, maintenance of the surgeon's optical axis at the surgical site without having to look at a navigation monitor, and working freely with both hands without holding the ultrasound scanner during biopsy taking. The described functional model can also be used in open head and neck surgery.

[Automatic registration of patients with A-mode ultrasound for computer-assisted surgery. Laboratory proof of concept]

BACKGROUND

The main source of error in 3D navigation is the patient-to-image registration process. Anatomical landmarks or adhesive markers perform sub-optimally. Bone-anchored invasive markers significantly change the clinical workflow of navigated ENT surgery, are invasive and cause patient discomfort. In order to minimize registration errors and to further streamline the clinical use of intraoperative 3D navigation we demonstrate that A-mode ultrasound allows an accurate 3D surface profile of the os occipitale to be created which can be reliably registered on preoperative patient CT data.

METHODS

The transducer is mechanically positioned with sub-millimeter accuracy on the patient's occiput. From the sound echos a 3D surface is generated and registered to the preoperative CT images with the iterative closest point (ICP) algorithm. The evaluation of our setup was performed on three anatomic specimens and one bony skull.

RESULTS

The ultrasound echoes from the occiput allowed the creation of an adequate 3D surface which could be registered to a segmentation of the CT image with an accuracy greater than 1.5 mm. The experiments were evaluated by an intuitive representation of the spatial deviation between CT and ultrasound data as a color-coded map.

CONCLUSION

The approach to scan the posterior skull with A-mode ultrasound enables automatic intraoperative registration and can be integrated into the intraoperative setup.