RPS 1001c - Liver imaging and beyond: giving answers to clinical questions

Purpose:

To investigate the inter-rater reliability and validity of distance measurements in the mid-clavicular line (MCL) that are frequently reported as an estimate of liver size in cross-sectional imaging on a large dataset.

Methods and materials:

We identified 275 consecutive abdominal CT examinations acquired at our institution in December 2018. The corresponding reports contain measurements of liver height in MCL. Firstly, to investigate inter-rater reliability, MCL distance measurements were performed independently by three radiology residents on a subset of 55 exams (coronal plane). ICC and the corresponding 95% confidence interval (CI) were calculated (absolute-agreement, two-way-random-effects model). Secondly, to assess validity, we calculated the Pearson correlation coefficient for actual liver volumes and the MCL distance measurements extracted from the corresponding clinically-approved reports. Liver volumes of all 275 datasets were determined by an AI-based organ volumetry software (NeuronX, Siemens Healthineers), whose validity had been previously proven. All measurements were performed on the portal-venous phase series (slice-thickness: 5 mm).

Results:

Mean liver volume was 1775 ml (95% CI: 1693 ml–1856 ml). ICC for the MCL measurements of the three raters was 0.56 (95% CI: 0.39-0.70). The Pearson correlation coefficient for distance measurements in MCL and liver volumes was 0.58 (p<0.001).

Conclusion:

The poor to moderate inter-rater reliability of MCL distance measurements of the liver as well as its only moderate correlation with actual liver volumes suggest that we should stop using them as an estimate for liver size. Instead, we should promote the integration of reliable automated liver volumetric measurements into our workflows.

Limitations:

Only three readers. Measurement of liver volumes only in the coronal plane.

Ethics committee approval

Written informed consent was waived by the local ethics committee.

Funding:

No funding was received for this work.

Purpose:

Hepatocellular adenomas (HCAs) are rare benign liver tumours. Guidelines recommend continued surveillance for patients diagnosed with HCAs, but these recommendations are mainly based on small series or experts’ opinion. The aims of this study were to analyse the long-term evolution of HCAs, including solitary and multiple lesions, and to identify predictive features of progression.

Methods and materials:

In a retrospective cohort study, we included 118 patients (mean age 40±10 years old) with pathology-proven HCAs: 41 had solitary HCAs and 77 patients multiple HCAs. ß-catenin mutated HCAs and HCAs with foci of malignancy were defined at-risk of progression. MR exams were analysed and tumour evolution was evaluated using RECISTv1.1.

Results:

In a median follow-up of 5.0 years, 37/41 (90%) patients with solitary HCAs and 55/77 (71%) patients with multiple HCAs showed stable or regressive disease. After resection of solitary HCAs, new lesions appeared only in 2/29 (7%) patients, both with HCAs at-risk of progression. Among patients with multiple HCAs, HNF-1α inactivated HCAs showed a higher rate of progression compared to inflammatory HCAs (11/26[42.3%] vs. 7/37[18.9%], p=0.043), and lower use and lesser duration of oral contraceptives intake (28/32 [87.5%] vs. 45/45 (100%), p=0.027, and mean 12.0±7.5 years vs. 19.2±9.2 years, p= 0.001, respectively).

Conclusion:

78% of HCAs showed long-term stability or regression. After resection of solitary HCAs, new lesions occurred only in HCAs at-risk of progression. Patients with multiple HCAs were more likely to show progressive disease, with HNF-1 α inactivated HCAs being the most common subtype showing progression.

Limitations:

Retrospective design. A low number of unclassified, sonic hedgehog, or ß-catenin mutated in exon 7-8 HCAs. A lack of serial MRI exams. An adoption of RECIST 1.1 may have underestimated size changes.

Ethics committee approval

IRB approved; waiver of informed consent obtained.

Funding:

No funding was received for this work.

Purpose:

The histopathological growth pattern (HGP) is a prognostic factor in patients with colorectal liver metastases (CRLM). Currently, HGPs can only be obtained postoperatively through resection. We present a non-invasive, preoperative alternative using CT and radiomics.

Methods and materials:

As a proof-of-concept, we aimed to distinguish two extrema: pure (i.e. 100%) desmoplastic HGP (dHGP) and pure replacement HGP (rHGP). The dataset consisted of CT scans from 76 patients with 93 CRLM (45 dHGP; 48 rHGP). Three clinicians manually delineated the lesions; one repeated process. Observer segmentation variability was determined using the dice similarity coefficient (DSC). From each segmentation, 410 radiomics features were extracted. Decision models were created for the segmentations of each observer separately and by training and testing on different observers. The models were created through an automated search amongst a variety of machine learning algorithms to find the combination that maximizes performance. The evaluation was implemented through a 100x random-split cross-validation, using 80% for training and model optimisation and 20% for testing.

Results:

The mean inter-rater DSC was 0.69, indicating poor observer agreement. The single-observer radiomics models resulted in mean area under the curves (mAUCs) of 0.67, 0.61, and 0.70. The multiobserver radiomics models, trained on two observers and tested on the third, resulted in mAUCs of 0.72, 0.70, and 0.64. Thus, no substantial model differences were found among the segmentations of the observers.

Conclusion:

Our radiomics model is able to associate quantitative CT imaging features with pure HGPs and generalises well to segmentations of an unseen observer. Pending further optimisation and validation, radiomics based on CT imaging may have potential as a non-invasive, preoperative surrogate for postoperative HGP assessment.

Limitations:

Relatively small dataset.

Ethics committee approval

Erasmus MC IRB (MEC-2017-479).

Funding:

NWO #14929-14930.

Purpose:

Recent evidence suggests that gadolinium-ethoxybenzyl-diethylenetriamine penta-acetic acid-enhanced magnetic resonance imaging (Gd-EOB-DTPA MRI) may be used to evaluate liver function. We assessed whether the signal intensity of Gd-EOB-DTPA MRI might correlate with common use liver disease clinical score systems and posthepatectomy liver failure (PHLF) in patients undergoing hepatectomy for liver tumours.

Methods and materials:

We retrospectively analysed 137 preoperative Gd-EOB-DTPA MRIs of patients undergoing hepatectomy from November 1^st, 2015, and July 31^st, 2018.

Mean signal intensity of liver (L₂₀) and spleen (S₂₀) were measured on T1-weighted single-breath-hold 3D fat-saturated gradient-echo sequences acquired 20 minutes after Gd-EOB-DTPA administration.

The hepatocellular uptake index (HUI) of liver volume (V_L) was calculated with the following formula V_L[(L₂₀/S₂₀)-1] and was tested with several clinical score systems for liver disease and to the occurrence of PHLF.

Results:

Patients with unhealthy livers had significantly lower values of HUI in comparison with those with normal function. This was found for MELD score ≤9 vs. >9 (p=0.0488), BILCHE score ≤2 vs. >2 (p=0.0208), ALBI grades (p=0.0357), and Humanitas score ≤6 vs. >6 (p=0.0311). HUI was significantly lower in those patients with PHLF (p=0.001). Receiver operating characteristics curve analysis revealed valuable HUI ability in predicting PHLF (AUC=0.84; 95%CI=0.71-0.92; p<0.001), with a cutoff value of 574.33 (98% sensitivity; 83% specificity).

Conclusion:

HUI measured on preoperative Gd-EOB-DTPA MRI identifies patients with unhealthy liver and predicts PHLF. This index could be used to discriminate those patients at higher risk of complications after hepatectomy.

Limitations:

Patients with severely diseased liver were not included since they weren't considered candidates for hepatectomy. Other proposed methods of Gd-EOB-DTPA MRI signal intensity measurements weren't analysed for our cohort.

Ethics committee approval

Informed consent was obtained from each considered patient.

Funding:

No funding was received for this work.

Purpose:

To evaluate whether multiple arterial phases (MAP) using DISCO acquisition would improve the display of hepatic arteries when comparing to single arterial phase (SAP) and computed tomographic angiography (CTA).

Methods and materials:

A total of 130 patients were enrolled. In part I of the study, 50 patients underwent MRI with MAP image and 50 patients with SAP images. In part II of the study, 30 patients underwent both MRI with MAP and CTA. Two readers independently assessed the hepatic arterial display on a four-point scale in terms of image quality and visualisation of hepatic arterial branches. The kappa test was used to evaluate the agreement between the two readers. Kruskal-Wallis test was used to compare the difference of arterial display and Bonferroni correction was used for further multiple comparisons of arteries.

Results:

Moderate to excellent interobserver agreement was obtained for the arterial phase timing and degree of motion artifact (all kappa value > 0.65). For part I, the mean arterial display score obtained with MAP was higher than SAP imaging in the common hepatic artery (CHA, 3.68±0.47vs3.14±0.5), proper hepatic artery (PHA, 3.40±0.50vs2.88±0.44), left hepatic artery (LHA, 2.78±0.42vs2.40±0.57), right hepatic artery (RHA, 3.04±0.45vs2.72±0.54), left gastric artery (LGA, 3.10±0.42vs2.62±0.67), and gastroduodenal artery (GDA, 2.80±0.61vs2.42±0.76) (all p<0.01). For part II, MAP and CTA acquisition showed comparable image quality and arterial display score in CHA (3.50±0.51vs3.47±0.51, p=0.798), PHA (3.30±0.47vs3.30±0.53, p=0.935), LHA (2.83±0.59vs3.03±0.41, p=0.122), RHA (3.10±0.48vs3.13±0.35, p=0.809), LGA (3.03±0.32vs3.00±0.37, p=0.710), and GDA(2.80±0.61vs2.80±0.48, p=0.659).

Conclusion:

MAP using DISCO acquisition is superior than SAP and is comparable with CTA in the display of hepatic arteries.

Limitations:

The diagnostic performance of these imaging methods in lesion detection and conspicuity was not evaluated.

Ethics committee approval

Local IRB approved this study and written informed consent were obtained.

Funding:

No funding was received for this work.

Purpose:

ALPPS is a two-stage liver resection: the splitting in-situ and portal vein ligation in the 1^st step allow for rapid liver hypertrophy and completion of resection after 7-15 days (2^nd step). Accurate patient selection is mandatory to avoid periprocedural liver failure.

The aim of this work is to evaluate the correlation of preoperative CT textural features with liver hypertrophy on virtual resection on CT after ALPPS 1^st step.

Methods and materials:

All ALPPS performed between January 2013 and January 2015 were retrospectively included. All patients were <70 yo with no-known cirrhosis or severe steatosis.

All CT were obtained with a 64-row CT (LighSpeed VCT, GE Healthcare) and performed at baseline and at 7^th postoperative day after ALPPS 1^st step with a tri-phasic CT protocol. The remnant liver volume (mL) was calculated with liver analysis on Syngo.via (Siemens Healthineers). The liver regeneration index (RI) was calculated as: [(RLV_7day−RLV_baseline)∕RLV _baseline]×100. The texture analysis was performed on baseline CT with radiomics prototype (Syngo.via Frontier, Siemens Healthineers) on the segmented RLV at baseline. The correlation between RI>70% and texture features were calculated with univariate and multivariate analysis.

Results:

14 patients were included (10 F/4M) with a median age of 47 (25-75p: 41-62 yo) with a RI of 69% (25-75p: 52%-81%). Among texture parameters, gray level co-occurrence matrix (GLCM) contrast was an independent predictor of RI>70%.

Conclusion:

The preliminary data on CT texture analysis in patients undergoing ALPPS show potential for prediction of liver hypertrophy after ALPPS 1^st Step.

Limitations:

This is a single-center study with a limited number of patients.

Ethics committee approval

n/a

Funding:

No funding was received for this work.