Standardisation of Convex Ultrasound Data Through Geometric Analysis and Augmentation

TL;DR

The paper tackles the lack of standardised ultrasound datasets by introducing a geometry-based approach that automatically extracts the convex US plane and encodes it as an annulus sector with parameters $O$, $ heta$, $r_{inner}$, and $r_{outer}$. It presents a four-stage pipeline—Plane Masking, Centre calculation, Radial Boundary Detection, and Annulus Parameters Calculation—followed by scan-line extraction and linearisation to a consistent representation for augmentation. Experimental validation on private intraoperative data and public noisy datasets demonstrates accurate key-point estimation, low re-projection error (about $0.0064$ px), and robust linearisation with MS-SSIM around $0.69$ when RF data is involved. This approach offers a practical path to standardise US datasets for data-driven methods, improving reproducibility and enabling geometry-preserving augmentation.

Abstract

The application of ultrasound in healthcare has seen increased diversity and importance. Unlike other medical imaging modalities, ultrasound research and development has historically lagged, particularly in the case of applications with data-driven algorithms. A significant issue with ultrasound is the extreme variability of the images, due to the number of different machines available and the possible combination of parameter settings. One outcome of this is the lack of standardised and benchmarking ultrasound datasets. The method proposed in this article is an approach to alleviating this issue of disorganisation. For this purpose, the issue of ultrasound data sparsity is examined and a novel perspective, approach, and solution is proposed; involving the extraction of the underlying ultrasound plane within the image and representing it using annulus sector geometry. An application of this methodology is proposed, which is the extraction of scan lines and the linearisation of convex planes. Validation of the robustness of the proposed method is performed on both private and public data. The impact of deformation and the invertibility of augmentation using the estimated annulus sector parameters is also studied. Keywords: Ultrasound, Annulus Sector, Augmentation, Linearisation.

Figures (9)

• Figure 1: Medical imaging modality deep learning research output trends.
• Figure 2: Annulus sector geometric properties, created using https://www.geogebra.org.
• Figure 3: Extracting the US plane. Left shows the original image. Middle shows a plot of the z-score, where in this circumstance there is only one spike as the background pixel is constant at 0. Right shows the output of the connected components process, where the US plane is the largest object and represented in white (the other colours represent other objects, belonging to the GUI).
• Figure 4: (Right) An example of the calculation of the vertical axis of symmetry using the sliding window method. The blue line represents the intensities of vector $S$, the red line is the vertical line passing through point $m$ and the green lines are the vertical lines through points $min_l, min_r$. (Left) The binarised US plane with the same lines superimposed.
• Figure 5: Examples from the controlled experiment with the privaye data. The left images show the processing of the example images using the proposed method. The white plane is the output of the connected components process. The purple dots are the edge points of the mask, and the blue lines and orange crosses are the output of the RANSAC and the estimated radial boundaries. The middle and right images show the estimated and ground truth key points.