Semiautomated segmentation of hepatocellular carcinoma tumors with MRI using convolutional neural networks.

OBJECTIVE: To assess the performance of convolutional neural networks (CNNs) for semiautomated segmentation of hepatocellular carcinoma (HCC) tumors on MRI. METHODS: This retrospective single-center study included 292 patients (237 M/55F, mean age 61 years) with pathologically confirmed HCC between 08/2015 and 06/2019 and who underwent MRI before surgery. The dataset was randomly divided into training (n = 195), validation (n = 66), and test sets (n = 31). Volumes of interest (VOIs) were manually placed on index lesions by 3 independent radiologists on different sequences (T2-weighted imaging [WI], T1WI pre-and post-contrast on arterial [AP], portal venous [PVP], delayed [DP, 3 min post-contrast] and hepatobiliary phases [HBP, when using gadoxetate], and diffusion-weighted imaging [DWI]). Manual segmentation was used as ground truth to train and validate a CNN-based pipeline. For semiautomated segmentation of tumors, we selected a random pixel inside the VOI, and the CNN provided two outputs: single slice and volumetric outputs. Segmentation performance and inter-observer agreement were analyzed using the 3D Dice similarity coefficient (DSC). RESULTS: A total of 261 HCCs were segmented on the training/validation sets, and 31 on the test set. The median lesion size was 3.0 cm (IQR 2.0-5.2 cm). Mean DSC (test set) varied depending on the MRI sequence with a range between 0.442 (ADC) and 0.778 (high b-value DWI) for single-slice segmentation; and between 0.305 (ADC) and 0.667 (T1WI pre) for volumetric-segmentation. Comparison between the two models showed better performance in single-slice segmentation, with statistical significance on T2WI, T1WI-PVP, DWI, and ADC. Inter-observer reproducibility of segmentation analysis showed a mean DSC of 0.71 in lesions between 1 and 2 cm, 0.85 in lesions between 2 and 5 cm, and 0.82 in lesions > 5 cm. CONCLUSION: CNN models have fair to good performance for semiautomated HCC segmentation, depending on the sequence and tumor size, with better performance for the single-slice approach. Refinement of volumetric approaches is needed in future studies. KEY POINTS: • Semiautomated single-slice and volumetric segmentation using convolutional neural networks (CNNs) models provided fair to good performance for hepatocellular carcinoma segmentation on MRI. • CNN models' performance for HCC segmentation accuracy depends on the MRI sequence and tumor size, with the best results on diffusion-weighted imaging and T1-weighted imaging pre-contrast, and for larger lesions.

MR elastography outperforms shear wave elastography for the diagnosis of clinically significant portal hypertension.

OBJECTIVES: Portal hypertension (PH) is associated with complications such as ascites and esophageal varices and is typically diagnosed through invasive hepatic venous pressure gradient (HVPG) measurement, which is not widely available. In this study, we aim to assess the diagnostic performance of 2D/3D MR elastography (MRE) and shear wave elastography (SWE) measures of liver and spleen stiffness (LS and SS) and spleen volume, to noninvasively diagnose clinically significant portal hypertension (CSPH) using HVPG measurement as the reference. METHODS: In this prospective study, patients with liver disease underwent 2D/3D MRE and SWE of the liver and spleen, as well as HVPG measurement. The correlation between MRE/SWE measures of LS/SS and spleen volume with HVPG was assessed. ROC analysis was used to determine the utility of MRE, SWE, and spleen volume for diagnosing CSPH. RESULTS: Thirty-six patients (M/F 22/14, mean age 55 ± 14 years) were included. Of the evaluated parameters, 3D MRE SS had the strongest correlation with HVPG (r = 0.686, p < 0.001), followed by 2D MRE SS (r = 0.476, p = 0.004). 3D MRE SS displayed the best performance for diagnosis of CSPH (AUC = 0.911) followed by 2D MRE SS (AUC = 0.845) and 3D MRE LS (AUC = 0.804). SWE SS showed poor performance for diagnosis of CSPH (AUC = 0.583) while spleen volume was a fair predictor (AUC = 0.738). 3D MRE SS was significantly superior to SWE LS/SS (p ≤ 0.021) for the diagnosis of CSPH. CONCLUSION: SS measured with 3D MRE outperforms SWE for the diagnosis of CSPH. SS appears to be a useful biomarker for assessing PH severity. These results need further validation. KEY POINTS: • Spleen stiffness measured with 2D and 3D MR elastography correlates significantly with hepatic venous pressure gradient measurement. • Spleen stiffness measured with 3D MR elastography demonstrates excellent performance for the diagnosis of clinically significant portal hypertension (AUC 0.911). • Spleen stiffness measured with 3D MR elastography outperforms liver and spleen stiffness measured with shear wave elastography for diagnosis of clinically significant portal hypertension.

Precision of MRI radiomics features in the liver and hepatocellular carcinoma.

OBJECTIVES: To assess the precision of MRI radiomics features in hepatocellular carcinoma (HCC) tumors and liver parenchyma. METHODS: The study population consisted of 55 patients, including 16 with untreated HCCs, who underwent two repeat contrast-enhanced abdominal MRI exams within 1 month to evaluate: (1) test-retest repeatability using the same MRI system (n = 28, 10 HCCs); (2) inter-platform reproducibility between different MRI systems (n = 27, 6 HCCs); (3) inter-observer reproducibility (n = 16, 16 HCCs). Shape and 1st- and 2nd-order radiomics features were quantified on pre-contrast T1-weighted imaging (WI), T1WI portal venous phase (pvp), T2WI, and ADC (apparent diffusion coefficient), on liver regions of interest (ROIs) and HCC volumes of interest (VOIs). Precision was assessed by calculating intraclass correlation coefficient (ICC), concordance correlation coefficient (CCC), and coefficient of variation (CV). RESULTS: There was moderate to excellent test-retest repeatability of shape and 1st- and 2nd-order features for all sequences in HCCs (ICC: 0.53-0.99; CV: 3-29%), and moderate to good test-retest repeatability of 1st- and 2nd-order features for T1WI sequences, and 2nd-order features for T2WI in the liver (ICC: 0.53-0.73; CV: 12-19%). There was poor inter-platform reproducibility for all features and sequences, except for shape and 1st-order features on T1WI in HCCs (CCC: 0.58-0.99; CV: 3-15%). Good to excellent inter-observer reproducibility was found for all features and sequences in HCCs (CCC: 0.80-0.99; CV: 4-15%) and moderate to good for liver (CCC: 0.45-0.86; CV: 6-25%). CONCLUSIONS: MRI radiomics features have acceptable repeatability in the liver and HCC when using the same MRI system and across readers but have low reproducibility across MR systems, except for shape and 1st-order features on T1WI. Data must be interpreted with caution when performing multiplatform radiomics studies. KEY POINTS: • MRI radiomics features have acceptable repeatability when using the same MRI system but less reproducible when using different MRI platforms. • MRI radiomics features extracted from T1 weighted-imaging show greater stability across exams than T2 weighted-imaging and ADC. • Inter-observer reproducibility of MRI radiomics features was found to be good in HCC tumors and acceptable in liver parenchyma.

MRI is the most commonly used imaging modality for HCC screening at a tertiary care transplant center.

PURPOSE: In this study, we describe the patterns of hepatocellular carcinoma (HCC) screening with imaging and factors associated with imaging modality selection in a tertiary care transplant center. METHODS: This was a retrospective study where all adult patients with cirrhosis and/or chronic hepatitis B virus infection referred for HCC screening with ultrasound (US), CT or MRI were identified during 2017. The association between imaging methods, demographic/clinical data were analyzed by uni- and multivariate analysis. RESULTS: A total of 1437 patients were included (median age 61y, 59% male, median BMI 27.5 kg/m2, median AFP 3.4 ng/mL, 37% with HCV and 87% with cirrhosis). Index screening imaging method utilization included MRI (51%), US (33%) and CT (16%). Use of US as the index imaging modality for screening was significantly associated with race/ethnicity [Odds Ratio (OR) 1.71-2.01, all p < 0.05] in multivariate analysis. Presence of cirrhosis (OR 0.29, p < 0.001) and referral by a hepatologist (OR 0.23, p < 0.001) were associated with screening with MRI in the multivariate analysis; while gender, age, BMI, etiology and income at ZIP code of residence were not significantly associated with imaging modality selection. HCC was observed in 62 patients (prevalence 4.3%). Rate of HCC detection was significantly higher with MRI vs US (5.9% vs. 1.5%, p = 0.001). CONCLUSION: MRI was the most frequently used modality (> 50%) for HCC screening in our tertiary care center, in contrast with the current practice guidelines. Race/ethnicity, cirrhosis and referral by a hepatologist were associated with the imaging method used for HCC screening.

Dynamic contrast-enhanced MRI perfusion quantification in hepatocellular carcinoma: comparison of gadoxetate disodium and gadobenate dimeglumine.

OBJECTIVES: (1) To assess the quality of the arterial input function (AIF) during dynamic contrast-enhanced (DCE) MRI of the liver and (2) to quantify perfusion parameters of hepatocellular carcinoma (HCC) and liver parenchyma during the first 3 min post-contrast injection with DCE-MRI using gadoxetate disodium compared to gadobenate dimeglumine (Gd-BOPTA) in different patient populations. METHODS: In this prospective study, we evaluated 66 patients with 83 HCCs who underwent DCE-MRI, using gadoxetate disodium (group 1, n = 28) or Gd-BOPTA (group 2, n = 38). AIF qualitative and quantitative features were assessed. Perfusion parameters (based on the initial 3 min post-contrast) were extracted in tumours and liver parenchyma, including model-free parameters (time-to-peak enhancement (TTP), time-to-washout) and modelled parameters (arterial flow (Fa), portal venous flow (Fp), total flow (Ft), arterial fraction, mean transit time (MTT), distribution volume (DV)). In addition, lesion-to-liver contrast ratios (LLCRs) were measured. Fisher's exact tests and Mann-Whitney U tests were used to compare the two groups. RESULTS: AIF quality, modelled and model-free perfusion parameters in HCC were similar between the 2 groups (p = 0.054-0.932). Liver parenchymal flow was lower and liver enhancement occurred later in group 1 vs group 2 (Fp, p = 0.002; Ft, p = 0.001; TTP, MTT, all p < 0.001), while there were no significant differences in tumour LLCR (max. positive LLCR, p = 0.230; max. negative LLCR, p = 0.317). CONCLUSION: Gadoxetate disodium provides comparable AIF quality and HCC perfusion parameters compared to Gd-BOPTA during dynamic phases. Despite delayed and decreased liver enhancement with gadoxetate disodium, LLCRs were equivalent between contrast agents, indicating similar tumour conspicuity. KEY POINTS: • Arterial input function quality, modelled, and model-free dynamic parameters measured in hepatocellular carcinoma are similar in patients receiving gadoxetate disodium or gadobenate dimeglumine during the first 3 min post injection. • Gadoxetate disodium and gadobenate dimeglumine show similar lesion-to-liver contrast ratios during dynamic phases in patients with HCC. • There is lower portal and lower total hepatic flow and longer hepatic mean transit time and time-to-peak with gadoxetate disodium compared to gadobenate dimeglumine.

4D flow MRI for the assessment of renal transplant dysfunction: initial results.

OBJECTIVES: (1) Determine inter-observer reproducibility and test-retest repeatability of 4D flow parameters in renal allograft vessels; (2) determine if 4D flow measurements in the renal artery (RA) and renal vein (RV) can distinguish between functional and dysfunctional allografts; (3) correlate haemodynamic parameters with estimated glomerular filtration rate (eGFR), perfusion measured with dynamic contrast-enhanced MRI (DCE-MRI) and histopathology. METHODS: Twenty-five prospectively recruited renal transplant patients (stable function/chronic renal allograft dysfunction, 12/13) underwent 4D flow MRI at 1.5 T. 4D flow coronal oblique acquisitions were performed in the transplant renal artery (RA) (velocity encoding parameter, VENC = 120 cm/s) and renal vein (RV) (VENC = 45 cm/s). Test-retest repeatability (n = 3) and inter-observer reproducibility (n = 10) were assessed by Cohen's kappa, coefficient of variation (CoV) and Bland-Altman statistics. Haemodynamic parameters were compared between patients and correlated to the estimated glomerular filtration rate, DCE-MRI parameters (n = 10) and histopathology from allograft biopsies (n = 15). RESULTS: For inter-observer reproducibility, kappa was > 0.99 and 0.62 and CoV of flow was 12.6% and 7.8% for RA and RV, respectively. For test-retest repeatability, kappa was > 0.99 and 0.5 and CoV of flow was 27.3% and 59.4%, for RA and RV, respectively. RA (p = 0.039) and RV (p = 0.019) flow were both significantly reduced in dysfunctional allografts. Both identified chronic allograft dysfunction with good diagnostic performance (RA: AUC = 0.76, p = 0.036; RV: AUC = 0.8, p = 0.018). RA flow correlated negatively with histopathologic interstitial fibrosis score ci (ρ = - 0.6, p = 0.03). CONCLUSIONS: 4D flow parameters had better repeatability in the RA than in the RV. RA and RV flow can identify chronic renal allograft dysfunction, with RA flow correlating with histopathologic interstitial fibrosis score. KEY POINTS: • Inter-observer reproducibility of 4D flow measurements was acceptable in both the transplant renal artery and vein, but test-retest repeatability was better in the renal artery than in the renal vein. • Blood flow measurements obtained with 4D flow MRI in the renal artery and renal vein are significantly reduced in dysfunctional renal transplants. • Renal transplant artery flow correlated negatively with histopathologic interstitial fibrosis score.

Assessment of Hepatocellular Carcinoma Response to 90Y Radioembolization Using Dynamic Contrast Material-enhanced MRI and Intravoxel Incoherent Motion Diffusion-weighted Imaging.

Purpose: To quantify diffusion and perfusion changes in hepatocellular carcinoma (HCC) induced by yttrium 90 (90Y) radioembolization and to assess the value of dynamic contrast material-enhanced (DCE) MRI and intravoxel incoherent motion (IVIM) diffusion-weighted imaging (DWI) for predicting HCC response. Materials and Methods: Institutional review board approval was obtained for this prospective study (clinical trial registry NCT01871545). Twenty-four participants with HCC (mean age, 69 years ± 9 [standard deviation], 18 men) underwent multiparametric MRI, including IVIM DWI and gadoxetic acid DCE MRI before (n = 24) and 6 weeks (n = 21) after radioembolization. IVIM DWI and DCE MRI histogram parameters were quantified in HCCs and liver parenchyma. HCC response was assessed by using modified Response Evaluation Criteria in Solid Tumors at 6 weeks and 6-12 months after radioembolization. Logistic regression analysis was used to evaluate the diagnostic performance of baseline MRI and clinical parameters for prediction of response. Results: Twenty-five HCCs were analyzed (mean size, 3.6 cm ± 1.9). Radioembolization resulted in significantly decreased perfusion (DCE MRI arterial flow, P = .002; IVIM pseudodiffusion coefficient [D*], P = .014). Multivariate logistic regression selected combined serum α-fetoprotein and portal flow (F p ) skewness (area under the curve [AUC] = 0.924) and combined D* standard deviation and F p kurtosis (AUC = 0.916) for prediction of objective and complete response at 6 weeks, respectively. Standard deviation of DCE MRI parameter arterial fraction was selected as the optimal predictor for complete response at 6-12 months (AUC = 0.857). Conclusion: Diffusion and perfusion MRI can be used to evaluate the response of HCC to radioembolization. Pretreatment DCE MRI histogram parameters may be useful for radioembolization treatment stratification. Supplemental material is available for this article. © RSNA, 2020.

Gadoxetate-enhanced abbreviated MRI is highly accurate for hepatocellular carcinoma screening.

OBJECTIVES: The primary objective was to compare the performance of 3 different abbreviated MRI (AMRI) sets extracted from a complete gadoxetate-enhanced MRI obtained for hepatocellular carcinoma (HCC) screening. Secondary objective was to perform a preliminary cost-effectiveness analysis, comparing each AMRI set to published ultrasound performance for HCC screening in the USA. METHODS: This retrospective study included 237 consecutive patients (M/F, 146/91; mean age, 58 years) with chronic liver disease who underwent a complete gadoxetate-enhanced MRI for HCC screening in 2017 in a single institution. Two radiologists independently reviewed 3 AMRI sets extracted from the complete exam: non-contrast (NC-AMRI: T2-weighted imaging (T2wi)+diffusion-weighted imaging (DWI)), dynamic-AMRI (Dyn-AMRI: T2wi+DWI+dynamic T1wi), and hepatobiliary phase AMRI (HBP-AMRI: T2wi+DWI+T1wi during the HBP). Each patient was classified as HCC-positive/HCC-negative based on the reference standard, which consisted in all available patient data. Diagnostic performance for HCC detection was compared between sets. Estimated set characteristics, including historical ultrasound data, were incorporated into a microsimulation model for cost-effectiveness analysis. RESULTS: The reference standard identified 13/237 patients with HCC (prevalence, 5.5%; mean size, 33.7 ± 30 mm). Pooled sensitivities were 61.5% for NC-AMRI (95% confidence intervals, 34.4-83%), 84.6% for Dyn-AMRI (60.8-95.1%), and 80.8% for HBP-AMRI (53.6-93.9%), without difference between sets (p range, 0.06-0.16). Pooled specificities were 95.5% (92.4-97.4%), 99.8% (98.4-100%), and 94.9% (91.6-96.9%), respectively, with a significant difference between Dyn-AMRI and the other sets (p < 0.01). All AMRI methods were effective compared with ultrasound, with life-year gain of 3-12 months against incremental costs of US$ < 12,000. CONCLUSIONS: NC-AMRI has limited sensitivity for HCC detection, while HBP-AMRI and Dyn-AMRI showed excellent sensitivity and specificity, the latter being slightly higher for Dyn-AMRI. Cost-effectiveness estimates showed that AMRI is effective compared with ultrasound. KEY POINTS: • Comparison of different abbreviated MRI (AMRI) sets reconstructed from a complete gadoxetate MRI demonstrated that non-contrast AMRI has low sensitivity (61.5%) compared with contrast-enhanced AMRI (80.8% for hepatobiliary phase AMRI and 84.6% for dynamic AMRI), with all sets having high specificity. • Non-contrast and hepatobiliary phase AMRI can be performed in less than 14 min (including set-up time), while dynamic AMRI can be performed in less than 17 min. • All AMRI sets were cost-effective for HCC screening in at-risk population in comparison with ultrasound.

Characterization of solid renal neoplasms using MRI-based quantitative radiomics features.

PURPOSE: To assess the diagnostic value of magnetic resonance imaging (MRI)-based radiomics features using machine learning (ML) models in characterizing solid renal neoplasms, in comparison/combination with qualitative radiologic evaluation. METHODS: Retrospective analysis of 125 patients (mean age 59 years, 67% males) with solid renal neoplasms that underwent MRI before surgery. Qualitative (signal and enhancement characteristics) and quantitative radiomics analyses (histogram and texture features) were performed on T2-weighted imaging (WI), T1-WI pre- and post-contrast, and DWI. Mann-Whitney U test and receiver-operating characteristic analysis were used in a training set (n = 88) to evaluate diagnostic performance of qualitative and radiomics features for differentiation of renal cell carcinomas (RCCs) from benign lesions, and characterization of RCC subtypes (clear cell RCC [ccRCC] and papillary RCC [pRCC]). Random forest ML models were developed for discrimination between tumor types on the training set, and validated on an independent set (n = 37). RESULTS: We assessed 104 RCCs (51 ccRCC, 29 pRCC, and 24 other subtypes) and 21 benign lesions in 125 patients. Significant qualitative and quantitative radiomics features (area under the curve [AUC] between 0.62 and 0.90) were included for ML analysis. Models with best diagnostic performance on validation sets showed AUC of 0.73 (confidence interval [CI] 0.5-0.96) for differentiating RCC from benign lesions (using combination of qualitative and radiomics features); AUC of 0.77 (CI 0.62-0.92) for diagnosing ccRCC (using radiomics features), and AUC of 0.74 (CI 0.53-0.95) for diagnosing pRCC (using qualitative features). CONCLUSION: ML models incorporating MRI-based radiomics features and qualitative radiologic assessment can help characterize renal masses.

MRI radiomics features predict immuno-oncological characteristics of hepatocellular carcinoma.

OBJECTIVE: To assess the value of qualitative and quantitative MRI radiomics features for noninvasive prediction of immuno-oncologic characteristics and outcomes of hepatocellular carcinoma (HCC). METHODS: This retrospective, IRB-approved study included 48 patients with HCC (M/F 35/13, mean age 60y) who underwent hepatic resection or transplant within 4 months of abdominal MRI. Qualitative imaging traits, quantitative nontexture related and texture features were assessed in index lesions on contrast-enhanced T1-weighted and diffusion-weighted images. The association of imaging features with immunoprofiling and genomics features was assessed using binary logistic regression and correlation analyses. Binary logistic regression analysis was also employed to analyse the association of radiomics, histopathologic and genomics features with radiological early recurrence of HCC at 12 months. RESULTS: Qualitative (r = - 0.41-0.40, p < 0.042) and quantitative (r = - 0.52-0.45, p < 0.049) radiomics features correlated with immunohistochemical cell type markers for T-cells (CD3), macrophages (CD68) and endothelial cells (CD31). Radiomics features also correlated with expression of immunotherapy targets PD-L1 at protein level (r = 0.41-0.47, p < 0.029) as well as PD1 and CTLA4 at mRNA expression level (r = - 0.48-0.47, p < 0.037). Finally, radiomics features, including tumour size, showed significant diagnostic performance for assessment of early HCC recurrence (AUC 0.76-0.80, p < 0.043), while immunoprofiling and genomic features did not (p = 0.098-0929). CONCLUSIONS: MRI radiomics features may serve as noninvasive predictors of HCC immuno-oncological characteristics and tumour recurrence and may aid in treatment stratification of HCC patients. These results need prospective validation. KEY POINTS: • MRI radiomics features showed significant associations with immunophenotyping and genomics characteristics of hepatocellular carcinoma. • Radiomics features, including tumour size, showed significant associations with early hepatocellular carcinoma recurrence after resection.

Luminal Water Imaging: Comparison With Diffusion-Weighted Imaging (DWI) and PI-RADS for Characterization of Prostate Cancer Aggressiveness.

BACKGROUND: Luminal water imaging (LWI), a multicomponent T2 mapping technique, has shown promise for prostate cancer (PCa) detection and characterization. PURPOSE: To 1) quantify LWI parameters and apparent diffusion coefficient (ADC) in PCa and benign peripheral zone (PZ) tissues; and 2) evaluate the diagnostic performance of LWI, ADC, and PI-RADS parameters for differentiation between low- and high-grade PCa lesions. STUDY TYPE: Prospective. SUBJECTS: Twenty-six PCa patients undergoing prostatectomy (mean age 59 years, range 46-72 years). FIELD STRENGTH/SEQUENCE: Multiparametric MRI at 3.0T, including diffusion-weighted imaging (DWI) and LWI T2 mapping. ASSESSMENT: LWI parameters and ADC were quantified in index PCa lesions and benign PZ. STATISTICAL TESTS: Differences in MRI parameters between PCa and benign PZ were assessed using Wilcoxon signed tests. Spearman correlation of pathological grade group (GG) with LWI parameters, ADC, and PI-RADS was evaluated. The utility of each of the parameters for differentiation between low-grade (GG ≤2) and high-grade (GG ≥3) PCa was determined by Mann-Whitney U tests and ROC analyses. RESULTS: Twenty-six index lesions were analyzed (mean size 1.7 ± 0.8 cm, GG: 1 [n = 1; 4%], 2 [n = 14, 54%], 3 [n = 8, 31%], 5 [n = 3, 12%]). LWI parameters and ADC both showed high diagnostic performance for differentiation between benign PZ and PCa (highest area under the curve [AUC] for LWI parameter T2,short [AUC = 0.98, P < 0.001]). The LWI parameters luminal water fraction (LWF) and amplitude of long T2 component Along significantly correlated with GG (r = -0.441, P = 0.024 and r = -0.414, P = 0.036, respectively), while PI-RADS, ADC, and the other LWI parameters did not (P = 0.132-0.869). LWF and Along also showed significant differences between low-grade and high-grade PCa (AUC = 0.776, P = 0.008 and AUC = 0.758, P = 0.027, respectively). Maximum diagnostic performance for discrimination of high-grade PCa was found with combined LWI parameters (AUC 0.891, P = 0.001). DATA CONCLUSION: LWI parameters, in particular in combination, showed superior diagnostic performance for differentiation between low-grade and high-grade PCa compared to ADC and PI-RADS assessment. J. Magn. Reson. Imaging 2020;52:271-279.