Modeling and analysis of 3-D elongated shapes with applications to long bone morphometry

Nested star-shaped objects segmentation using diameter annotations

Most current deep learning based approaches for image segmentation require annotations of large datasets, which limits their application in clinical practice. We observe a mismatch between the voxelwise ground-truth that is required to optimize an objective at a voxel level and the commonly used, less time-consuming clinical annotations seeking to characterize the most important information about the patient (diameters, counts, etc.). In this study, we propose to bridge this gap for the case of multiple nested star-shaped objects (e.g., a blood vessel lumen and its outer wall) by optimizing a deep learning model based on diameter annotations. This is achieved by extracting in a differentiable manner the boundary points of the objects at training time, and by using this extraction during the backpropagation. We evaluate the proposed approach on segmentation of the carotid artery lumen and wall from multisequence MR images, thus reducing the annotation burden to only four annotated landmarks required to measure the diameters in the direction of the vessel's maximum narrowing. Our experiments show that training based on diameter annotations produces state-of-the-art weakly supervised segmentations and performs reasonably compared to full supervision. We made our code publicly available at https://gitlab.com/radiology/aim/carotid-artery-image-analysis/nested-star-shaped-objects.

Evaluation of computer-aided design software methods for assessment of the three-dimensional geometry of the canine radius

OBJECTIVE

To describe methods to measure the 3-D orientation of the proximal, diaphyseal, and distal segments of the canine radius by use of computer-aided design software (CADS) and to compare the repeatability and reliability of measurements derived by those methods.

SAMPLE

31 canine radii with biapical deformities and 24 clinically normal (control) canine radii.

PROCEDURES

Select CT scans of radii were imported into a CADS program. Cartesian coordinate systems for the humerus and proximal, diaphyseal, and distal radial segments were developed. The orientation of each radial segment in the frontal, sagittal, and transverse planes was measured in triplicate by 3 methods. The repeatability and reliability of those measurements were calculated and compared among the 3 measurement methods.

RESULTS

The mean ± SD within-subject repeatability of radial angular measurements for all 3 methods was 1.40 ± 0.67° in the frontal plane, 3.17 ± 2.21° in the sagittal plane, and 3.01 ± 1.11° in the transverse plane for control radii and 2.56 ± 1.95° in the frontal plane, 3.59 ± 2.39° in the sagittal plane, and 3.47 ± 1.19° in the transverse plane for abnormal radii. Mean ± SD bias between radial measurement methods was 1.88 ± 2.07° in the frontal plane, 6.44 ± 6.80° in the sagittal plane, and 2.27 ± 2.81° in the transverse plane.

CONCLUSIONS AND CLINICAL RELEVANCE

Results indicated that use of CADS to assess the 3-D orientation of the proximal, diaphyseal, and distal segments of normal and abnormal canine radii yielded highly repeatable and reliable measurements.

Modeling prostate anatomy from multiple view TRUS images for image-guided HIFU therapy

Current planning methods for transrectal high-intensity focused ultrasound treatment of prostate cancer rely on manually defining treatment regions in 15-20 sector transrectal ultrasound (TRUS) images of the prostate. Although effective, it is desirable to reduce user interaction time by identifying functionally related anatomic structures (segmenting), then automatically laying out treatment sites using these structures as a guide. Accordingly, a method has been developed to effectively generate solid three-dimensional (3-D) models of the prostate, urethra, and rectal wall from boundary trace data. Modeling the urethra and rectal wall are straightforward, but modeling the prostate is more difficult and has received much attention in the literature. New results presented here are aimed at overcoming many of the limitations of previous approaches to modeling the prostate while using boundary traces obtained via manual tracing in as few as 5 sector and 3 linear images. The results presented here are based on a new type of surface, the Fourier ellipsoid, and the use of sector and linear TRUS images. Tissue-specific 3-D models will ultimately permit finer control of energy deposition and more selective destruction of cancerous regions while sparing critical neighboring structures.

Object classification in 3-D images using alpha-trimmed mean radial basis function network

We propose a pattern classification based approach for simultaneous three-dimensional (3-D) object modeling and segmentation in image volumes. The 3-D objects are described as a set of overlapping ellipsoids. The segmentation relies on the geometrical model and graylevel statistics. The characteristic parameters of the ellipsoids and of the graylevel statistics are embedded in a radial basis function (RBF) network and they are found by means of unsupervised training. A new robust training algorithm for RBF networks based on alpha-trimmed mean statistics is employed in this study. The extension of the Hough transform algorithm in the 3-D space by employing a spherical coordinate system is used for ellipsoidal center estimation. We study the performance of the proposed algorithm and we present results when segmenting a stack of microscopy images.