Search

Background: In oncology, the correct determination of nodal metastatic disease is essential for patient management, as patient treatment and prognosis are closely linked to the stage of the disease. The aim of the study was to develop a tool for automatic 3D detection and segmentation of lymph nodes (LNs) in computed tomography (CT) scans of the thorax using a fully convolutional neural network based on 3D foveal patches., Methods: The training dataset was collected from the Computed Tomography Lymph Nodes Collection of the Cancer Imaging Archive, containing 89 contrast-enhanced CT scans of the thorax. A total number of 4275 LNs was segmented semi-automatically by a radiologist, assessing the entire 3D volume of the LNs. Using this data, a fully convolutional neuronal network based on 3D foveal patches was trained with fourfold cross-validation. Testing was performed on an unseen dataset containing 15 contrast-enhanced CT scans of patients who were referred upon suspicion or for staging of bronchial carcinoma., Results: The algorithm achieved a good overall performance with a total detection rate of 76.9% for enlarged LNs during fourfold cross-validation in the training dataset with 10.3 false-positives per volume and of 69.9% in the unseen testing dataset. In the training dataset a better detection rate was observed for enlarged LNs compared to smaller LNs, the detection rate for LNs with a short-axis diameter (SAD) ≥ 20 mm and SAD 5-10 mm being 91.6% and 62.2% (p < 0.001), respectively. Best detection rates were obtained for LNs located in Level 4R (83.6%) and Level 7 (80.4%)., Conclusions: The proposed 3D deep learning approach achieves an overall good performance in the automatic detection and segmentation of thoracic LNs and shows reasonable generalizability, yielding the potential to facilitate detection during routine clinical work and to enable radiomics research without observer-bias.

Machine learning (ML) is making a dramatic impact on cardiovascular magnetic resonance (CMR) in many ways. This review seeks to highlight the major areas in CMR where ML, and deep learning in particular, can assist clinicians and engineers in improving imaging efficiency, quality, image analysis and interpretation, as well as patient evaluation. We discuss recent developments in the field of ML relevant to CMR in the areas of image acquisition & reconstruction, image analysis, diagnostic evaluation and derivation of prognostic information. To date, the main impact of ML in CMR has been to significantly reduce the time required for image segmentation and analysis. Accurate and reproducible fully automated quantification of left and right ventricular mass and volume is now available in commercial products. Active research areas include reduction of image acquisition and reconstruction time, improving spatial and temporal resolution, and analysis of perfusion and myocardial mapping. Although large cohort studies are providing valuable data sets for ML training, care must be taken in extending applications to specific patient groups. Since ML algorithms can fail in unpredictable ways, it is important to mitigate this by open source publication of computational processes and datasets. Furthermore, controlled trials are needed to evaluate methods across multiple centers and patient groups.

Background: The aim of this study was to evaluate the diagnostic potential of a novel cardiovascular magnetic resonance (CMR) based multiparametric imaging approach in suspected myocarditis and to compare it to traditional Lake Louise criteria (LLC)., Methods: CMR data from 67 patients with suspected acute myocarditis were retrospectively analyzed. Seventeen age- and gender-matched healthy subjects served as control. T2-mapping data were acquired using a Gradient-Spin-Echo T2-mapping sequence in short-axis orientation. T2-maps were segmented according to the 16-segments AHA-model and segmental T2 values and pixel-standard deviation (SD) were recorded. Afterwards, the parameters maxT2 (the highest segmental T2 value) and madSD (the mean absolute deviation (MAD) of the pixel-SDs) were calculated for each subject. Cine sequences in three long axes and a stack of short-axis views covering the left and right ventricle were analyzed using a dedicated feature tracking algorithm., Results: A multiparametric imaging model containing madSD and LV global circumferential strain (GCS LV ) resulted in the highest diagnostic performance in receiver operating curve analyses (area under the curve [AUC] 0.84) when compared to any model containing a single imaging parameter or to LLC (AUC 0.79). Adding late gadolinium enhancement (LGE) to the model resulted in a further increased diagnostic performance (AUC 0.93) and yielded the highest diagnostic sensitivity of 97% and specificity of 77%., Conclusions: A multiparametric CMR imaging model including the novel T2-mapping derived parameter madSD, the feature tracking derived strain parameter GCS LV and LGE yields superior diagnostic sensitivity in suspected acute myocarditis when compared to any imaging parameter alone and to LLC.

Background: The purpose of the present study was to investigate the diagnostic value of T2-mapping in acute myocarditis (ACM) and to define cut-off values for edema detection., Methods: Cardiovascular magnetic resonance (CMR) data of 31 patients with ACM were retrospectively analyzed. 30 healthy volunteers (HV) served as a control. Additionally to the routine CMR protocol, T2-mapping data were acquired at 1.5 T using a breathhold Gradient-Spin-Echo T2-mapping sequence in six short axis slices. T2-maps were segmented according to the 16-segments AHA-model and segmental T2 values as well as the segmental pixel-standard deviation (SD) were analyzed., Results: Mean differences of global myocardial T2 or pixel-SD between HV and ACM patients were only small, lying in the normal range of HV. In contrast, variation of segmental T2 values and pixel-SD was much larger in ACM patients compared to HV. In random forests and multiple logistic regression analyses, the combination of the highest segmental T2 value within each patient (maxT2) and the mean absolute deviation (MAD) of log-transformed pixel-SD (madSD) over all 16 segments within each patient proved to be the best discriminators between HV and ACM patients with an AUC of 0.85 in ROC-analysis. In classification trees, a combined cut-off of 0.22 for madSD and of 68 ms for maxT2 resulted in 83% specificity and 81% sensitivity for detection of ACM., Conclusions: The proposed cut-off values for maxT2 and madSD in the setting of ACM allow edema detection with high sensitivity and specificity and therefore have the potential to overcome the hurdles of T2-mapping for its integration into clinical routine.

Background: The aim of this study was the evaluation of a fast Gradient Spin Echo Technique (GraSE) for cardiac T2-mapping, combining a robust estimation of T2 relaxation times with short acquisition times. The sequence was compared against two previously introduced T2-mapping techniques in a phantom and in vivo., Methods: Phantom experiments were performed at 1.5 T using a commercially available cylindrical gel phantom. Three different T2-mapping techniques were compared: a Multi Echo Spin Echo (MESE; serving as a reference), a T2-prepared balanced Steady State Free Precession (T2prep) and a Gradient Spin Echo sequence. For the subsequent in vivo study, 12 healthy volunteers were examined on a clinical 1.5 T scanner. The three T2-mapping sequences were performed at three short-axis slices. Global myocardial T2 relaxation times were calculated and statistical analysis was performed. For assessment of pixel-by-pixel homogeneity, the number of segments showing an inhomogeneous T2 value distribution, as defined by a pixel SD exceeding 20 % of the corresponding observed T2 time, was counted., Results: Phantom experiments showed a greater difference of measured T2 values between T2prep and MESE than between GraSE and MESE, especially for species with low T1 values. Both, GraSE and T2prep resulted in an overestimation of T2 times compared to MESE. In vivo, significant differences between mean T2 times were observed. In general, T2prep resulted in lowest (52.4 ± 2.8 ms) and GraSE in highest T2 estimates (59.3 ± 4.0 ms). Analysis of pixel-by-pixel homogeneity revealed the least number of segments with inhomogeneous T2 distribution for GraSE-derived T2 maps., Conclusions: The GraSE sequence is a fast and robust sequence, combining advantages of both MESE and T2prep techniques, which promises to enable improved clinical applicability of T2-mapping in the future. Our study revealed significant differences of derived mean T2 values when applying different sequence designs. Therefore, a systematic comparison of different cardiac T2-mapping sequences and the establishment of dedicated reference values should be the goal of future studies.

Background: Fibrofatty degeneration of myocardium in ARVC is associated with wall motion abnormalities. The aim of this study was to examine whether Cardiovascular Magnetic Resonance (CMR) based strain analysis using feature tracking (FT) can serve as a quantifiable measure to confirm global and regional ventricular dysfunction in ARVC patients and support the early detection of ARVC., Methods: We enrolled 20 patients with ARVC, 30 with borderline ARVC and 22 subjects with a positive family history but no clinical signs of a manifest ARVC. 10 healthy volunteers (HV) served as controls. 15 ARVC patients received genotyping for Plakophilin-2 mutation (PKP-2), of which 7 were found to be positive. Cine MR datasets of all subjects were assessed for myocardial strain using FT (TomTec Diogenes Software). Global strain and strain rate in radial, circumferential and longitudinal mode were assessed for the right and left ventricle. In addition strain analysis at a segmental level was performed for the right ventricular free wall., Results: RV global longitudinal strain rates in ARVC (-0.68 ± 0.36 sec⁻¹) and borderline ARVC (-0.85 ± 0.36 sec⁻¹) were significantly reduced in comparison with HV (-1.38 ± 0.52 sec⁻¹, p ≤ 0.05). Furthermore, in ARVC patients RV global circumferential strain and strain rates at the basal level were significantly reduced compared with HV (strain: -5.1 ± 2.7 vs. -9.2 ± 3.6%; strain rate: -0.31 ± 0.13 sec(-1) vs. -0.61 ± 0.21 sec⁻¹). Even for patients with ARVC or borderline ARVC and normal RV ejection fraction (n=30) global longitudinal strain rate proved to be significantly reduced compared with HV (-0.9 ± 0.3 vs. -1.4 ± 0.5 sec(-1); p < 0.005). In ARVC patients with PKP-2 mutation there was a clear trend towards a more pronounced impairment in RV global longitudinal strain rate. On ROC analysis RV global longitudinal strain rate and circumferential strain rate at the basal level proved to be the best discriminators between ARVC patients and HV (AUC: 0.9 and 0.92, respectively)., Conclusion: CMR based strain analysis using FT is an objective and useful measure for quantification of wall motion abnormalities in ARVC. It allows differentiation between manifest or borderline ARVC and HV, even if ejection fraction is still normal.