Volume estimation in a sequence of freehand ultrasound images

Volume estimation is particularly important in clinical medicine. Accurate volume estimation can provide quantitative information from which the follow-up therapy can be derived. In this paper, an efficient approach to volume estimation in a sequence of freehand ultrasound images is proposed. By integral of vector areas along the path of centroids of serial cross-sections, 3D volume estimation can be represented as 2D area calculation, where a fast mapping algorithm generating 2D representation is presented so that the position of interpolation points can be calculated with high efficiency. Meanwhile, to improve the accuracy, the cubic spline with second-order continuity is proposed for interpolation of 2D representation. Volume estimation on simulating phantoms for parallel cutting, fan cutting and random cuttings is provided. The experimental results show that the 2D representation generated by the fast mapping algorithm is highly efficient with less than 0.001 ms for 100 cross-sections. Quantitative comparisons show that the proposed interpolation method can approximate the original volume more precisely as compared to the Catmull?Rom (CR) spline, especially in the case of small number of cross-sections. In all cases, our approach can obtain accurate results at an error of less than 2% for ten cross-sections. Additionally, volume estimation on a high intensity focused ultrasound (HIFU) lesion based on linear B-scan and rotational B-scan sequential images are also performed. The experiments show that the proposed approach is promising and may have potential in clinical applications.