Comparison of image quality of 3D ultrasound: motorized acquisition versus freehand navigated acquisition, a phantom study

Clinical validation of the accuracy of an intra-operative assessment tool using 3D ultrasound compared to histopathology in patients with squamous cell carcinoma of the tongue

BACKGROUND

Histopathological analysis often shows close resection margins after surgical removal of tongue squamous cell carcinoma (TSCC). This study aimed to investigate the agreement between intraoperative 3D ultrasound (US) margin assessment and postoperative histopathology of resected TSCC.

METHODS

In this study, ten patients were prospectively included. Three fiducial cannulas were inserted into the specimen. To acquire a motorized 3D US volume, the resected specimen was submerged in saline, after which images were acquired while the probe moved over the specimen. The US volumes were annotated twice: (1) automatically and (2) manually, with the automatic segmentation as initialization. After standardized histopathological processing, all hematoxylin-eosin whole slide images (WSI) were included for analysis. Corresponding US images were found based on the known WSI spacing and fiducials. Blinded observers measured the tumor thickness and the margin in the caudal, deep, and cranial directions on every slide. The anterior and posterior margin was measured per specimen.

RESULTS

The mean difference in all measurements between manually segmented US and histopathology was 2.34 (SD: ±3.34) mm, and Spearman's rank correlation coefficient was 0.733 (p < 0.001). The smallest mean difference was in the tumor thickness with 0.80 (SD: ±2.44) mm and a correlation of 0.836 (p < 0.001). Limitations were observed in the caudal region, where no correlation was found.

CONCLUSION

This study shows that 3D US and histopathology have a moderate to strong statistically significant correlation (r = 0.733; p < 0.001) and a mean difference between the modalities of 2.3 mm (95%CI: -4.2; 8.9). Future research should focus on patient outcomes regarding resection margins.

Implementing a deep learning model for automatic tongue tumour segmentation in ex-vivo 3-dimensional ultrasound volumes

Three-dimensional (3D) ultrasound can assess the margins of resected tongue carcinoma during surgery. Manual segmentation (MS) is time-consuming, labour-intensive, and subject to operator variability. This study aims to investigate use of a 3D deep learning model for fast intraoperative segmentation of tongue carcinoma in 3D ultrasound volumes. Additionally, it investigates the clinical effect of automatic segmentation. A 3D No New U-Net (nnUNet) was trained on 113 manually annotated ultrasound volumes of resected tongue carcinoma. The model was implemented on a mobile workstation and clinically validated on 16 prospectively included tongue carcinoma patients. Different prediction settings were investigated. Automatic segmentations with multiple islands were adjusted by selecting the best-representing island. The final margin status (FMS) based on automatic, semi-automatic, and manual segmentation was computed and compared with the histopathological margin. The standard 3D nnUNet resulted in the best-performing automatic segmentation with a mean (SD) Dice volumetric score of 0.65 (0.30), Dice surface score of 0.73 (0.26), average surface distance of 0.44 (0.61) mm, Hausdorff distance of 6.65 (8.84) mm, and prediction time of 8 seconds. FMS based on automatic segmentation had a low correlation with histopathology (r = 0.12, p = 0.67); MS resulted in a moderate but insignificant correlation with histopathology (r = 0.4, p = 0.12, n = 16). Implementing the 3D nnUNet yielded fast, automatic segmentation of tongue carcinoma in 3D ultrasound volumes. Correlation between FMS and histopathology obtained from these segmentations was lower than the moderate correlation between MS and histopathology.