Reply to the Letter to the Editor: “Does quantitative assessment of arterial phase hyperenhancement and washout improve LI-RADS v2018–based classification of liver lesions?”
by Daniel Stocker, Anton S. Becker, Borna K. Barth, Stephan Skawran, Malwina Kaniewska, Michael A. Fischer, Olivio Donati, Caecilia S. Reiner (Caecilia.Reiner@usz.ch)Does quantitative assessment of arterial phase hyperenhancement and washout improve LI-RADS v2018–based classification of liver lesions?
With great interest we have received the “Letter to the Editor” by Mulé and Luciani regarding our recent publication “Does quantitative assessment of arterial phase hyperenhancement and washout improve LI-RADS v2018–based classification of liver lesions?” .
We agree that the terminology we used for nonrim arterial phase hyperenhancement (APHE) and nonperipheral washout in the manuscript is imprecise and differs from the LI-RADS v2018 lexicon terms . However, readers were instructed to assess nonrim APHE and nonperipheral washout based on LI-RADS v2018 as explicitly stated in the Materials and Methods section. Therefore, only nonrim APHE and nonperipheral washout was classified as APHE and washout, respectively. Consequently, only lesions with nonrim APHE and/or nonperipheral washout were classified as hepatocellular carcinoma (HCC). Thus, our evaluation of these features and the stratification of HCC probability is accurately following LI-RADS v2018. We furthermore agree that the spatial distribution of APHE and washout is crucial for the differentiation of HCC and non-HCC malignancies. However, the aim of our study was to evaluate the inter-reader agreement and diagnostic performance of quantitative versus qualitative assessment of LI-RDAS criteria based on previously implemented quantitative measures of APHE and washout. We see a need for a less subjective interpretation of LI-RADS v2018 primary imaging criteria in order to achieve a reduction of inter-reader variability, which is essential for clinical adoption of LI-RADS. Accordingly, we appreciate the current discussion on further improvements and encourage further studies to improve the quantitative assessment of APHE as well as washout. Furthermore, we agree with the authors that a 3D segmentation of liver lesions – preferably in a (semi-) automatic fashion – has the potential to further improve diagnostic accuracy as well as inter-reader agreement by reduction of false placement of a region of interest (ROI).
In our study, the ROI was placed in the part of the lesion with the largest and strongest area of arterial enhancement (if present) and in the same position on pre-contrast, portal venous and delayed phase images. We have chosen this approach to introduce consistency throughout all lesions (HCC and benign lesions) and reduce subjective reader decisions resulting in lower inter-reader agreement, which is typically lower for washout than APHE [3, 4]. In this study design, the only subjective reader decision for quantitative measurements was to choose the part with the largest and strongest area of APHE.
We agree that APHE and washout do not necessarily need to coincide in the same area of the lesion. Choosing a 3D ROI approach would account for this problem and would further decrease the subjective reader decisions by covering the whole lesion rather than choosing a small area within the lesion. However, the 3D whole lesion approach would include necrotic, less or non-vascularized tumor areas, which would result in decreased mean values of APHE and washout and thus potential feature underestimation. In this setting, higher order statistics would have to be obtained. For now, the added value of 3D ROI placement compared to 2D ROI placement remains unknown and calls for evaluation in future studies.
Improvement in inter-reader agreement of the feature enhancing “capsule” would be desirable. We did not include quantitative “capsule” measurements, because we believe that quantitative assessment of enhancing “capsule” is very challenging. ROI placement in such a narrow field of interest underlies reader bias even more than placing larger ROIs and may introduce a lot of measured noise and yield less reliable results. Furthermore, the time factor would have to be weighed against potential increase in diagnostic accuracy, if such a meticulous measurement is introduced in clinical practice.
Currently there are no recommendations regarding optimal cut-off values for the assessment of APHE and washout in focal liver lesions in the literature. We chose the cut-off value of 10% signal intensity increase for quantitative APHE and no specific absolute numeric cut-off value for quantitative washout (any form of quantitative hypointensity compared with the surrounding liver parenchyma in the portal venous or delayed phase) to increase sensitivity. The exact definition of washout by LI-RADS is: “Nonperipheral visually assessed temporal reduction in enhancement in whole or in part relative to composite liver tissue from earlier to later phase resulting in hypoenhancement in the extracellular phase (portal venous or delayed phase if extracellular contrast agents or gadobenate is given; portal venous phase if gadoxetate is given)” . The definition of quantitative washout in our manuscript tries to reflect this visual assessment by taking the lesion enhancement in the arterial and in the portal venous or delayed phase into account. In our study quantitative washout was only considered when the lesion was iso- or hyperintense compared with the liver in the arterial phase and had a negative lesion-to-liver contrast ratio in the portal venous or delayed phase, meaning that the lesion was hypointense compared with the surrounding liver. Therefore, the so-called “dynamic part” termed by Mulé and Luciani was considered for the assessment of quantitative washout in our study.
Finally, we would like to underline that we tried to prospectively implement previously defined quantitative APHE and washout criteria in order to assess and compare diagnostic accuracy of quantitative versus qualitative read-out. ROC analysis is useful to determine optimal cut-off values but needs a separate validation group as results cannot be applied to the same study group resulting in a falsely increased diagnostic accuracy of quantitative LI-RADS. However, we agree that the optimal cut-off value for quantitative assessment of APHE and washout to diagnose HCCs remains unknown as of now and may differ from our proposed cut-off values.