Waves of precision Intra-operative 3D ultrasound for margin assessment of resected tongue cancer

Open Access
Authors
  • N.M. Bekedam
Supervisors
Cosupervisors
Award date 05-03-2025
ISBN
  • 9789465068824
Number of pages 144
Organisations
  • Faculty of Dentistry (ACTA)
Abstract
The first choice for resectable squamous cell carcinoma of the tongue is surgery. The surgeon aims to remove the tumor with an adequate margin. Close margins are found in up to 85% of patients and are associated with a higher recurrence rate and a lower survival rate. Intra-operative margin assessment could prevent close margins by directly performing an additional resection. This thesis focuses on developing an intra-operative margin assessment tool by 3D ultrasound (US).
Applying electromagnetic tracking to the US transducer enables 3D acquisition.
The image quality of this freehand acquisition approach was compared to that of a motorized acquisition, and it showed that the motorized acquisition is more stable and more accurate in axial measurements.
Manual segmentation of 3D US volumes was too time-consuming and labor-intensive for assessment during surgery. Implementing a nnUNet model reduced the segmentation time to 8 seconds. The model’s performance was very high for the specimen regions but left room for improvement for the tumor regions.
The 3D US assessment was clinically validated by comparing US with histopathology measurements in ten patients. The agreement was 2.3 mm (95%CI: -4.2; 8.9) with a correlation of r=0.733 (p<0.001), which means that the ex-vivo 3D US should consider 7.3 mm to achieve a mean histopathological margin of >5mm.
Tumor border delineation can be difficult in the US and consequently results in incorrect ground truth labels. Multimodal image registration from 3D US to histopathology is possible within a 0.4 mm registration error. The registered histopathological tumor annotation (HTA) in US resulted in an improved agreement (2.16 mm; 95%CI: -1.31; 5.63) and correlation (r=0.924; p<0.001), suggesting ground truth labels should be based on registered HTA.
In conclusion, 3D US can be utilized as an intra-operative margin assessment tool. Multimodal image registration delivers accurate ground truth annotation. Future research should focus on the impact of the intra-operative 3D US margin assessment on patient outcomes regarding the resection margin in tongue specimens.
Document type PhD thesis
Language English
Downloads
Permalink to this page
cover
Back