Deformable registration of trans-rectal ultrasound (TRUS) and magnetic resonance imaging (MRI) for focal prostate brachytherapy.

PURPOSE: Focal therapy in low-risk prostate cancer may provide the best balance between cancer control and quality of life preservation. As a minimally invasive approach performed under TRUS guidance, brachytherapy is an appealing framework for focal therapy. However, the contrast in TRUS images is generally insufficient to distinguish the target lesion from normal prostate tissue. MRI usually offers a much better contrast between the lesion and surrounding tissues. Registration between TRUS and MRI may therefore significantly improve lesion targeting capability in focal prostate brachytherapy. In this paper, we present a deformable registration framework for the accurate fusion of TRUS and MRI prostate volumes under large deformations arising from dissimilarities in diameter, shape and orientation between endorectal coils and TRUS probes. METHODS: Following pose correction by a RANSAC implementation of the ICP algorithm, TRUS and MRI Prostate contour points are represented by a 3D extension of the shape-context descriptor and matched by the Hungarian algorithm. Eventually, a smooth free-form warping is computed by fitting a 3D B-spline mesh to the set of matched points. RESULTS: Quantitative validation of the registration accuracy is provided on a retrospective set of ten real cases, using as landmarks either brachytherapy seeds (six cases) or external beam radiotherapy fiducials (four cases) implanted and visible in both modalities. The average registration error between the landmarks was 2.49 and 3.20 mm, for the brachytherapy and external beam sets, respectively, that is less than the MRI voxels' long axis length ([Formula: see text]). The overall average registration error (for brachytherapy and external beam datasets together) was 2.56 mm. CONCLUSIONS: The proposed method provides a promising framework for TRUS-MRI registration in focal prostate brachytherapy.

A functional test for the detection of infusion lines extravasation.

Extravasation during intravenous (IV) infusion is a common secondary effect with potentially serious clinical consequences. The correct positioning of the needle in the vein may be difficult to confirm when no blood return is observed. In this paper, a novel method is proposed for the detection of extravasation during infusion therapy. A small volume of a sodium bicarbonate solution is administrated IV and, following its consecutive dissociation, an excess of carbon dioxide (CO2) is rapidly exhaled by the lungs. The analysis of the exhaled CO2 signal by a pattern recognition algorithm enables the robust detection of the CO2 excess release, thereby confirming the absence of extravasation. Initial results are presented for the application of the method on a group of 89 oncology patients.