Improved detection of liver metastasis by registration and subtraction of dynamic contrast-enhanced magnetic resonance imaging.

Dipartimento di Ingegneria Biomedica, Politecnico di Milano, Italy.
Conference proceedings: ... Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Conference 02/2007; 2007:59-62. DOI: 10.1109/IEMBS.2007.4352222
Source: PubMed

ABSTRACT The purpose of this work was to validate a subtraction method for dynamic contrast-enhanced magnetic resonance volumes acquired at portal and delayed (3 hours) phases, to obtain a better depiction and delineation of hepatic secondary lesions. As the clinical readability of subtraction images depends on the degree of co-registration between volumes, we applied a 3D non-rigid registration procedure to realign portal and delayed-phase volumes before subtraction. Our method combines a rigid registration, based on Normalized Mutual Information maximization and a non-rigid registration, based on a multiresolution analysis performed through Wavelet Transform. The feasibility of the method was tested on thirty-five secondary lesions. Results showed that the proposed subtraction produces images of easy interpretation and improves diagnostic confidence. In particular, subtraction image of subcentimetric lesions was scored as diagnostic in the 89% of the cases.

  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Subtraction of Dynamic Contrast-Enhanced 3D Magnetic Resonance (DCE-MR) volumes can result in images that depict and accurately characterize a variety of liver lesions. However, the diagnostic utility of subtraction images depends on the extent of co-registration between non-enhanced and enhanced volumes. Movement of liver structures during acquisition must be corrected prior to subtraction. Currently available methods are computer intensive. We report a new method for the dynamic subtraction of MR liver images that does not require excessive computer time. Nineteen consecutive patients (median age 45 years; range 37-67) were evaluated by VIBE T1-weighted sequences (TR 5.2 ms, TE 2.6 ms, flip angle 20 degrees , slice thickness 1.5 mm) acquired before and 45s after contrast injection. Acquisition parameters were optimized for best portal system enhancement. Pre and post-contrast liver volumes were realigned using our 3D registration method which combines: (a) rigid 3D translation using maximization of normalized mutual information (NMI), and (b) fast 2D non-rigid registration which employs a complex discrete wavelet transform algorithm to maximize pixel phase correlation and perform multiresolution analysis. Registration performance was assessed quantitatively by NMI. The new registration procedure was able to realign liver structures in all 19 patients. NMI increased by about 8% after rigid registration (native vs. rigid registration 0.073 +/- 0.031 vs. 0.078 +/- 0.031, n.s., paired t-test) and by a further 23% (0.096 +/- 0.035 vs. 0.078 +/- 0.031, p < 0.001, paired t-test) after non-rigid realignment. The overall average NMI increase was 31%. This new method for realigning dynamic contrast-enhanced 3D MR volumes of liver leads to subtraction images that enhance diagnostic possibilities for liver lesions.
    BMC Medical Imaging 01/2006; 6:5. · 0.98 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: To date various image filtering approaches are reported, most of them enhance the signal-to-noise ratio (SNR) in different levels with loss of some useful information. In this paper, an adaptive template filtering method is described which can be used to increase SNR and keep the important edge information of medical images. In the proposed method, for each pixel, an optimal template of different shape and coefficients is selected automatically depending on its neighbor pixels. Not being simple extension of the 2D method, sequential filtering of slices in three orthogonal directions implemented the 3D algorithm, which improved existing adaptive template filtering technique, avoiding a burden of enormous search. Simulation and MRI image tests, both 2D and 3D, show that the new adaptive template filter provides higher SNR and sharper edges.
    Engineering in Medicine and Biology Society, 2001. Proceedings of the 23rd Annual International Conference of the IEEE; 02/2001
  • [Show abstract] [Hide abstract]
    ABSTRACT: The purpose of this work was to determine the technical feasibility and value of dynamic subtraction (postcontrast-precontrast) magnetic resonance (MR) imaging for the assessment of hyperintense lesions on precontrast T1-weighted images in the cirrhotic liver. One hundred four hyperintense lesions on T1-weighted precontrast and arterial phase postcontrast images were subjected to analysis of their subtraction qualities depending on the lesion size, location and/or the degree of misregistration between the source images in 27 different MR imaging sets. The quality of subtraction images was always diagnostic for lesions larger than 2 cm in diameter (n=8) but not diagnostic for 73% (40 of 55 lesions) of small subcentimetric lesions. Thirty-one subcapsular lesions always showed a variable degree of coregistration artifact. Only 3 of 35 lesions with a slice misregistration of 3 mm or more gave rise to subtraction images of diagnostic quality. For determining the contrast enhancement, the area under the receiver operating characteristic curve of 30 verified lesions was significantly larger (P <0.001) for subtraction images than for conventional arterial phase images. Depending on the lesion size and/or location or the degree of misregistration between the source images, dynamic subtraction MR imaging can be useful for the characterization of hyperintense lesions on precontrast T1-weighted imaging.
    Journal of Computer Assisted Tomography 01/2005; 29(1):51-8. · 1.60 Impact Factor