SS 1801 - Technical innovations in liver imaging
Clinical feasibility of multiple arterial phases in free-breathing at gadoxetic acid-enhanced liver MRI
Purpose: To investigate clinical feasibility of free-breathing multi-arterial phase using Cartesian sampling and added value of motion resolved reconstruction in gadoxetic acid-enhanced liver magnetic resonance imaging (MRI).
Methods and Materials: This retrospective study was approved by IRB and the requirement of informed consent was waved. A total of 37 patients were included who underwent iterative three-dimensional gradient echo sequence for precontrast, early arterial, late arterial and portal venous phases of T1-weighted image (T1WI) in free-breathing manner. Timing, motion artifact, image quality were evaluation by four radiologists on a 4-point scale. In addition, respiratory motion-resolved reconstruction was performed for patients with below average score (score <3) of late arterial phase and reviewed in the same manner.
Results: Overall image quality of free-breathing T1WI was 3.30±0.59 on precontrast, 2.68±0.70, 2.93±0.65 and 3.30±0.49 on early arterial, late arterial and portal venous phases, respectively. Image quality of early and late arterial phase was significantly lower than precontrast and portal venous phase (P<0.0001~0.001). Eleven patients showed late arterial phase with below than average image quality (<3), but image quality of late arterial phase was significantly improved on motion-resolved reconstruction images using the same MRI data (P=0.007). In addition, none of the patients missed adequate arterial phase timing on multiarterial phase acquired in free-breathing manner.
Conclusion: Dynamic phase with acceptable image quality at gadoxetic acid-enhanced liver MRI can be achieved using Cartesian reconstruction in free-breathing manner.
Virtual monochromatic image of detector-based spectral CT: improved image quality as compared with that obtained with conventional CT
Purpose: To compare image quality obtained in phantom with virtual monochromatic image of detector-based spectral CT (VMS) with that performed with conventional CT on various body size and dose right index (DRI) values.
Methods and Materials: Three different size of body phantoms with 8 FLLs were scanned with two conventional mode (80 and 120kVp) by detector-based spectral CT with iDose (level 4) reconstruction. The VMS is reconstructed from 120kVp iDose (level 4) image basis and displayed on vendor specific workstation. The noise and CNR in different body size, and subjective image quality and FLL conspicuity in small phantom were analysed with different DRI values (16, 19, 22 and 25).
Results: Image noise on VMS is always lower than that of 80kVp with higher CNR in low KeV ranges (from 40 to 69~73KeV) regardless of DRI values. Also comparing 120kVp, CNR of VMS is improved in low KeV ranges (from 40 to 75~81KeV) on each DRI values. On the qualitative analysis, diagnostic acceptability (p=0.015) and subjective noise (p<0.01) is significantly higher in VMS. Lesion conspicuity of low-contrast FLLs are significantly improved in VMS (all p<0.05) than 120kVp CT. Also, VMS present lesser stiff noise increment with higher CNR than those of 120 kVp as body size increases.
Conclusion: VMS of detector-based spectral CT, the up-to-date VMS technique, is presenting improved CNR, noise and subjective image quality, and it is especially valuable in large phantom and low-contrast FLLs.
Respiratory motion artefacts during arterial phase imaging with gadoxetic acid: how can we minimise this drawback?
Purpose: To determine which of three gadoxetic-acid injection techniques, performed serially in the same patient, could produce liver MRI with fewer contrast-related arterial-phase motion artefacts.
Methods and Materials: This IRB-approved, retrospective study included a cohort of 78 consecutive patients. Each patient had serial gadoxetic acid-enhanced MRI of the liver performed with at least two of three injection techniques: M1-test bolus undiluted, power-injected; M2-test bolus, diluted 50% with saline, power-injected; M3-fixed delay, undiluted, manually injected. The injection dose was 0.025mmol/kg body weight in all three methods. Three readers independently rated the images for motion artifacts, timing, and lesion detectability based upon a four-point Likert scale.
Results: Respiratory artefacts related to gadoxetic-acid arterial phase images obtained by M3 were superior to M1 (p=0.0001). M2 was rated significantly better than M1 (p=0.012). The difference between M3 and M2 scores was not statistically significant (p=0.49). Arterial-phase timing was significantly better for M1 compared to M3 (p<0.0001). The AUC was 0.59-0.68. M2 significantly outperformed M3 in arterial phase timing (p<0.0001). There was no significant difference between M1 and M2 (p=0.35).With regard to lesion detectability, there was no significant difference in the ratings for all three-injection techniques (p>0.05).Inter-reader agreement was moderate to substantial (0.41-0.62).
Conclusion: There were significantly fewer contrast-related-motion-artefacts in the arterial phase when gadoxetic acid was administered undiluted manually (M3), or diluted power-injected (M2), compared to undiluted power injection (M1).The arterial phase timing was significantly better with test-bolus, power injection of contrast media, however the diagnostic performance for all three methods was not significantly different.
Free-breathing undersampled radial VIBE as a salvage strategy for liver dynamics in patients unable to suspend respiration
Purpose: Respiratory motion artifacts are a frequent source for image degradation in abdominal MRI. The introduced free-breathing sequence can be used as a salvage strategy in patients with limited breath-holding capacity for dynamic liver imaging.
Methods and Materials: Twenty-seven consecutive patients with known liver metastases underwent dynamic liver MR imaging using a free-breathing undersampled radial sequence. Patients were eligible for the free-breathing protocol due to severe respiratory artifacts at the precontrast sequences. The introduced free-breathing sequence was compared to a standard dynamic breathhold VIBE. Overall image quality, liver edge sharpness, liver lesion conspicuity, hepatic vessel clarity, respiratory motion artifacts and streaking artifacts were independently, retrospectively and blindly scored by four experienced readers using a 5-point Likert scale. Contrast-enhancement ratio (CER) between native and arterial phases was measured to assess temporal resolution.
Results: Free-breathing VIBE received lower image quality scores than breathhold VIBE but presented with generally good image quality. Regarding patients with severe respiratory artifacts already at breathhold VIBE, free-breathing VIBE received significantly higher scores for overall image quality (4.1 vs 3.8, p=0.01) and hepatic lesion conspicuity (4.2 vs 3.8, p=0.02). There were generally more respiratory artifacts in breathhold VIBE (p<0.0001), whereas streaking was characteristic for the undersampled radial acquisition (p<0.0001). Temporal resolution assessed with CER showed higher values for free-breathing VIBE sequence (p=0.028).
Conclusion: The introduced free-breathing sequence is a promising alternative for liver dynamics in challenging patients. Accurate temporal resolution, low motion artifact susceptibility and good image quality qualify this sequence for daily routine.
Accuracy of new real-time shear wave elastography for assessing liver fibrosis in chronic viral hepatitis patients
Purpose: Recently, new real-time shear wave elastography (SWE, Aplio500®, Toshiba) has been released. In this study, the diagnostic performance of SWE and TE (transient elastography) in assessing liver fibrosis in chronic viral hepatitis was investigated. Differences in diagnostic performance of SWE were evaluated according to the obesity.
Methods and Materials: 53 chronic viral hepatitis patients who underwent liver biopsy were prospectively enrolled (23 with HBV, 30 with HCV). For each patient, liver stiffness was assessed by SWE and liver biopsy was performed in the same session (METAVIR score). TE was performed within two weeks of liver biopsy. AUROCs were performed and compared for each degree of liver fibrosis. Differences in AUROCs were investigated according to body mass index (BMI ≥25, and BMI≥ 30).
Results: Among 52 patients, 16 belonged to F0-F1 stage, 12 were F2, 16 were F3 and 9 were F4. TE was significantly more accurate for prediction of F ≥2 than SWE (AUC: SWE=0.630, TE = 0.810, P=0.02). However, there were no significant differences between SWE and TE in prediction of F≥3(AUC: SWE = 0.800, TE = 0.883, P = 0.20). In overweight patients(BMI ≥25), SWE showed good diagnostic performance (AUC:0.808 with cutoff value 18.9kPa with 69.23% sensitivity and 87.5% specificity) in prediction of F≥3. However, in obese patients (BMI ≥30), SWE showed poor diagnostic performance (AUC:0.667).
Conclusion: New SWE seems to be good method for assessing advanced liver fibrosis with similar predictive value to TE. SWE showed poor diagnostic performance for assessing advanced liver fibrosis in obese patients.
Evaluation of virtual monoenergetic images computed by new dual-layer CT in assessing hypervascularised liver lesions
Purpose: Evaluation of both objective and subjective image quality of virtual monoenergetic (vME) images for assessment of arterially hypervascularized liver lesions in dual-layer CT (e.g. haemangiomas, metastases).
Methods and Materials: 50 liver lesions in 20 contrast enhanced CT examinations, acquired at dual-layer CT (IQon, Philips Healthcare) in arterial contrast phase, were identified and retrospectively analysed in vME (40-200 keV; 10 keV intervals) and in conventional images, respectively. Image noise was represented by standard deviation of liver parenchyma; signal- (SNR) and contrast-to-noise ratio (CNR) of lesions and liver parenchyma were calculated. To assess subjective image quality, vME-images (40-110 keV) were rated by two readers with regard to lesion delineation and overall image quality, compared to conventional images, on a 5-point-scale.
Results: SNR of liver parenchyma was significantly higher in 40- and 50-keV-vME-images as compared to conventional images (3.8, 2.8 vs 1.3; p≤0.001). SNR and CNR of lesions were significantly higher in 40/50/60-keV-vME-images, compared to conventional images (SNR: 17.8, 12.4, 8.8 vs 5.0; CNR: 14.2, 10.5, 7.7 vs 4.4; p≤0.001). Image noise was significantly lower in vME-images (50-200keV), while 40-keV-vME-images exhibited similar image noise as conventional images (12.1 vs 13.8). Subjective lesion delineation received best rating at 50-keV-vME-images, while overall image quality was rated inferior to equal at all vME-intervals, compared to conventional reconstructions.
Conclusion: Both objective image quality and subjective assessment of hypervascularised liver lesions are superior at vME images of 40-60 keV. However, improved CNR at low-keV-vME-images impairs overall image quality.Thus, vME-images of low-keV may be useful as screening reconstructions.
Comparison between the conventional Couinaud method vs a semiautomatic software (Syngo.CT. liver analysis, Siemens) in the precise localisation of liver metastases
Purpose: To analyse the variability in the location of CT detected liver metastases using the conventional method of Couinaud (CM) or a semiautomatic software (SAS) for liver segmentation (Syngo.CT-liver Analysis©, Siemens Healthcare, Germany).
Methods and Materials: Two expert radiologists in abdominal imaging consecutively studied 25 oncologic patients with a total of 70 liver metastases detected on abdominal CT. Lesions were distributed by segments, following either the Couinaud method based on anatomical landmarks or the semiautomatic software liver Analysis©, which relies on the ramifications of portal branches. The global agreement between the two methods was calculated, as well as the Kappa agreement values for each segment.
Results: The kappa values for each segment were: I(1.0), II(0.79), III(0.9), IV(0.68), V(0.42), VI(0.65), VII(0.49), VIII(0.4). The global agreement between methods was of 62%. There agreement was only moderate for lesions localised on the segments V, VII and VIII. With CM it is difficult to differentiate the V-VI boundaries because the right hepatic vein is not well recognised in the inferior segments. Between segments VII-VIII at the dome the middle hepatic vein trajectory is not well depicted.
Conclusion: The Couinaud method alone, based on indirect anatomical landmarks is not reliable enough for the precise localisation of hepatic metastases, especially for the lesions located below the liver dome (segments V-VI, VII-VII) Software tools that are able to demarcate the liver segments using a more anatomical approach, based on the portal ramifications, are essential for an accurate localisation of liver metastases, especially in patients undergoing surgery.
Differentiation of malignant thrombus from bland thrombus of the portal vein in patients with cirrhosis: application of intravoxel incoherent motion diffusion-weighted MR imaging
Purpose: Our purpose of this study was to demonstrate the presence of blood flow within the malignant thrombus using intravoxel incoherent motion (IVIM) diffusion-weighted (DW) MR imaging and investigate the utility of IVIM in distinguishing bland thrombus from malignant tumour thrombus of the PV in patients with cirrhosis or hepatocellular carcinoma.
Methods and Materials: Forty three patients with PV thrombosis (malignant thrombus n=29 and bland thrombus n=14) examined with gadoxetic acid-enhanced MR imaging including IVIM were enrolled. IVIM DW imaging was acquired with free-breathing axial single-shot echo-planar two-dimensional imaging sequence and the following eight b values: 0,25,50,75,100,200, 500 and 800 sec/mm2. Diffusion coefficient (D), pseudo-diffusion coefficient (D*), and perfusion fraction (f) were calculated with bi-exponential model using Matlab software and were compared between malignant and bland thrombi using unpaired t-test.
Results: D* of malignant thrombus (mean=67.51 x 10-3mm2/sec) was significantly higher than that of bland thrombus (9.31 x 10-3mm2/sec, p < 0.001). However, there was no significant difference in f between malignant (18.47 %) and bland thrombi (18.46 %). D of bland thrombus (1.35 x 10-3mm2/sec) was significantly higher than that of malignant thrombus (1.03 x 10-3mm2/sec, p=0.014). However, the mean D difference between the two thrombi was relatively small.
Conclusion: Increased pseudo-diffusion coefficient (D*) of malignant thrombus suggests higher intra-thrombus micro-perfusion, which might be due to blood flow by arterial neovascularisation within the malignant thrombus. Therefore, IVIM DW imaging appears to be a promising method for the discrimination between bland and malignant PV thrombi.