MRI is a valuable tool in the assessment of the extent of myocardial fibrosis. Late Gadolinium Enhancement (LGE)[1] is the noninvasive gold standard for focal fibrosis, parametric mapping with and without contrast-media enable detection of diffuse fibrosis[2]. We developed a non-rigid registration method to superimpose LGE images and T₁-Maps allowing for pixel-wise comparison of LGE extent and abnormal T₁ times. This may give new insights into disease development. Furthermore direct visual comparison is more easily assessible in a fused image.

A non-rigid, edge-detection based algorithm was developed in-house in MATLAB(Natwick, USA). It consists of a two-fold registration, by first matching anatomical, manually-set control points with an affine transformation and then aligning the endo- and epicardium with a non-rigid point set based approach. Contouring of the endo-and epicardium and selecting of the anterior insertion of the right ventricle was done by a CMR-experienced physician. Further segmentation was based on the 16-segment AHA-model[3]. We applied the tool in 12 patients with non-ischemic heart disease: Myotonic Dystrophy type 2 (MD2) n=6, age 59$$$\pm$$$8 years, all female and Hypertrophic Cardiomyopathy (HCM) patients, n=6, age: 51$$$\pm$$$17 years, 4 male and 2 female. In all patients T₁-mapping pre- and post-contrast media application as well as LGE-imaging were performed. Hematocrit was measured in 5 patients with MD2 immediately prior to CMR scan. The MR-protocol was as follows: MD2 at 1.5T (MAGNETOM Avanto, Siemens Healthcare, Erlangen, Germany) Modified Look-Locker inversion-recovery [4] (MOLLI) scheme (TE=1.08ms, FA=35°, FOV=(270 x 360)mm2, matrix= 169x224, slice thickness= 6 mm, GRAPPA acceleration factor 2)for native T₁-Mapping, for post-contrast T₁-Mapping 15 minutes after contrast-agent application (0.2 mmol/kg body weight, Gadoteridol) and LGE imaging technique (2D-PSIR; TR=744ms, TE=5.17ms, FOV=(350×262)mm², matrix=256×162, slice thickness=7mm, GRAPPA acceleration factor 2). HCM at 3.0T (MAGNETOM Verio, Siemens Healthcare, Erlangen, Germany) using a MOLLI (TR=2.6-2.7ms, TE=1.0-1.1ms, FA=35°, FOV= (270×360)mm², matrix=156×208 to 168×224,slice thickness=6mm, GRAPPA acceleration factor 2) scheme for native T₁-Mapping, post-contrast T₁-Mapping 10 minutes after contrast-agent application (0.2 mmol/kg body weight, Gadobutrolum) and LGE imaging technique (2D-PSIR; TR=10.5ms, TE=5.4ms, FA=30°, FOV=(350×262)mm²,matrix=256×162,slice thickness=6mm, GRAPPA acceleration factor 2.) ECV was calculated based on [5]. LGE positive areas were defined by signal intensity in remote myocardium+2 standard deviations(SD). T₁-mapping reference values for 1.5T were derived from our own normal values [6], normal values for 3.0T were published recently [7]. The threshold was chosen to be $$$\pm$$$2 SD from the mean value for post- and pre-contrast T₁-Mapping respectively and mean+2 SD for ECV.

Predominantly, slices with LGE were analyzed. In MD2 patients the basal slice was always affected, therefore all basal segments (36) of the 6 MD2 patients were evaluated for native T₁, T₁ post-contrast and LGE. In 6 HCM patients, midventricular (5 patients) and basal segments (1 patient) were used, depending on best LGE differentiation. ECV was calculated in 30 segments only, as hematocrit was missing in one MD2 patient and all HCM patients. Statistical analysis was conducted using the Wilcoxon signed rank test. First we compared the mean ECV of all segments from analysis done in CVI42 (Calgary, Canada) and after the registration in our implementation and found no statistically significant difference (p>0.05) in either left- or right-sided signed rank test. We found ECV and both native and post-contrast T1 abnormality areas to be significantly larger than LGE positive areas in all MD2 patients(p<0.05), but only post-contrast T1 abnormality region sizes were significantly increased in HCM patients. We calculated the percentage of pixels in each segment beyond the respective threshold as illustrated in the figures.In 6 MD2 patients, we showed that it is possible to register LGE images to ECV and T₁-Mapping images with no statistically significant change in data. Therefore we were able to generalize our tool from patients with MD2 to patients with HCM. We observed larger areas of abnormalities in ECV and both native and post-contrast T₁-Mapping compared to LGE posive areas. However, LGE positive area could not be fully detected by any of the proposed methods. This could be an error created by malregistration.Our proof of concept shows that feature extraction in registrated images is feasible. Furthermore a significant increase in the size of areas with abnormal native and post-contrast T₁ values compared to LGE positive areas was observed segment-wise. However, on a pixel-basis, we observed, that ECV, T₁ native or post-contrast abnormalities do not always include the LGE positive area completely.