Clinical significance of creative 3D-image fusion across multimodalities [PET+CT+MR] based on characteristic coregistration

Fusion imaging to evaluate the radiographic anatomical relationship between primary tumors and local recurrences in retroperitoneal soft tissue sarcoma

BACKGROUND

Local recurrence (LR) of retroperitoneal soft tissue sarcoma (RPS) is a common and life-threatening event. The evaluation of the exact anatomical patterns of local recurrence might help to improve local treatment in RPS.

METHODS

Of our local database we extracted ten patients with LR of RPS with axial MRI and/or CT datasets of the primary tumor (PT) and the LR. Using the Osirix DICOM viewer Version v.3.9.4 64-bit (Pixmeo, Geneva, Switzerland) we performed a three-step fusion algorithm consisting of: a) 3-point co-registration of the axial datasets depicting the PT and the LR using three abdominal landmarks b) re-orientation of the datasets and c) image fusion. We evaluated the feasibility of this technique with regard to categorizing the localization of LR as within or distant from the PT.

RESULTS

Fusion imaging was feasible in seven out of ten patients. In the other three patients anatomical shifting of organs after surgery led to a relevant mismatch of anatomical landmarks and impeded interpretation of the fused images. In five of seven patients with successful fusion imaging, local recurrences were located within the anatomical borders of the primary tumor, in two out of seven patients local recurrences were distant to the primary.

CONCLUSIONS

Fusion imaging of primary tumors and local recurrences is feasible in most patients with RPS. Most local recurrences occurred within the anatomical localization of the primary tumor. For further investigations validation of the technique in larger patient cohorts is required.

Registration of the Cone Beam CT and Blue-Ray Scanned Dental Model Based on the Improved ICP Algorithm

Multimodality image registration and fusion has complementary significance for guiding dental implant surgery. As the needs of the different resolution image registration, we develop an improved Iterative Closest Point (ICP) algorithm that focuses on the registration of Cone Beam Computed Tomography (CT) image and high-resolution Blue-light scanner image. The proposed algorithm includes two major phases, coarse and precise registration. Firstly, for reducing the matching interference of human subjective factors, we extract feature points based on curvature characteristics and use the improved three point's translational transformation method to realize coarse registration. Then, the feature point set and reference point set, obtained by the initial registered transformation, are processed in the precise registration step. Even with the unsatisfactory initial values, this two steps registration method can guarantee the global convergence and the convergence precision. Experimental results demonstrate that the method has successfully realized the registration of the Cone Beam CT dental model and the blue-ray scanner model with higher accuracy. So the method could provide researching foundation for the relevant software development in terms of the registration of multi-modality medical data.

The adaptive FEM elastic model for medical image registration

This paper proposes an adaptive mesh refinement strategy for the finite element method (FEM) based elastic registration model. The signature matrix for mesh refinement takes into account the regional intensity variance and the local deformation displacement. The regional intensity variance reflects detailed information for improving registration accuracy and the deformation displacement fine-tunes the mesh refinement for a more efficient algorithm. The gradient flows of two different similarity metrics, the sum of the squared difference and the spatially encoded mutual information for the mono-modal and multi-modal registrations, are used to derive external forces to drive the model to the equilibrium state. We compared our approach to three other models: (1) the conventional multi-resolution FEM registration algorithm; (2) the FEM elastic method that uses variation information for mesh refinement; and (3) the robust block matching based registration. Comparisons among different methods in a dataset with 20 CT image pairs upon artificial deformation demonstrate that our registration method achieved significant improvement in accuracies. Experimental results in another dataset of 40 real medical image pairs for both mono-modal and multi-modal registrations also show that our model outperforms the other three models in its accuracy.

Clinical and research applications of simultaneous positron emission tomography and MRI

Evaluation of the molecular processes responsible for disease pathogenesis and progression represents the new frontier of clinical radiology. Multimodality imaging lies at the cutting edge, combining the power of MRI for tissue characterization, microstructural appraisal and functional assessment together with new positron emission tomography (PET) tracers designed to target specific metabolic processes. The recent commercial availability of an integrated clinical whole-body PET-MRI provides a hybrid platform for exploring and exploiting the synergies of multimodal imaging. First experiences on the clinical and research application of hybrid PET-MRI are emerging. This article reviews the rapidly evolving field and speculates on the potential future direction.

Correcting for possible tissue distortion between provocation and assessment in skin testing: the divergent beam UVB photo-test

BACKGROUND

In tissue viability imaging (TiVi), an assessment method for skin erythema, correct orientation of skin position from provocation to assessment optimizes data interpretation. Image processing algorithms could compensate for the effects of skin translation, torsion and rotation realigning assessment images to the position of the skin at provocation.

METHODS

A reference image of a divergent, UVB phototest was acquired, as well as test images at varying levels of translation, rotation and torsion. Using 12 skin markers, an algorithm was applied to restore the distorted test images to the reference image.

RESULTS

The algorithm corrected torsion and rotation up to approximately 35 degrees. The radius of the erythemal reaction and average value of the input image closely matched that of the reference image's 'true value'.

CONCLUSION

The image 'de-warping' procedure improves the robustness of the response image evaluation in a clinical research setting and opens the possibility of the correction of possibly flawed images performed away from the laboratory setting by the subject/patient themselves. This opportunity may increase the use of photo-testing and, by extension, other late response skin testing where the necessity of a return assessment visit is a disincentive to performance of the test.