To the best of our knowledge, our study is the first, which investigated the accuracy of a simple technique to estimate liver volume by the calculation of VI on the basis of routine US and CT examination data. The results of our study showed that only the calculated volume indices based on diameter measurements derived from CT are a valid approach for the estimation of liver volume, which underlines previous research15.
Determination of liver size using simple and reliable techniques is clinically warranted. Different approaches to asses liver volume in ultrasound have been presented based on the measurement of liver diameters in one21,22,23 and more than one dimensions24. In our opinion, the determination of the liver volume based on the measurement in one plane is questionable due to the high susceptibility to errors caused by possible anatomical variations or potential measurement inaccuracies. In a clinical setting, it is accepted that liver volume is enlarged if the craniocaudal distance measured by US in the midclavicular line exceeds 16 cm21. This observation must be critically questioned since no results for reliability were reported and our data clearly show a poor inter-rater agreement for diameter measurement using ultrasound. However, our own data demonstrated high inter-rater repeatability of CT diameters. Therefore, the estimation of liver volume using one dimension is possible if the diameter is reliably assessed, for example using cross-sectional imaging. This fact needs to be further explored.
The calculation of the volume index to predict liver volume is another simple approach based on the measurement of liver diameters in more than one dimension. Various formulas for estimating the liver volume using volume indices have been presented so far. An approach by Boscaini et al. used the product of measuring three diameters (length, width, height) and divided this by 2725. This approach saw the liver in the form of a cube, which explains the lack of accuracy. A further development of this approach was provided by Marchesini et al.26 and Zoli et al.27 by comparing volume calculations with CT-based volume determinations. However, the authors described difficulties in comparability and therefore proposed a calibration to eliminate them. Muggli et al.28 followed an approach to determine liver volume on the basis of diametric measurements, however, patients with changes in liver parenchyma were excluded. For this reason, the results appear to show clear limitations for clinical practice. These could also be the reason why the results of the calculated VI appear slightly better in comparison to the actual volume than in our study.
This assumption is supported by the fact that our results showed a decreased interobserver agreement in the calculated VI in patients with an increased fib-4 score, which explains a lower accuracy of VI in presents of liver parenchyma changes.
In previous research, a calibration of volume index determined by the three diameters in their maximum orientation divided by the factor 3.6 was introduced. Using this factor, the volume index and the true liver volume are comparable. However, an accurate estimation of liver volume using volume indices requires also a reliable and robust assessment of liver diameters. As shown in our study results, we found excellent inter-rater reliability if liver diameters are determined in CT images, but not for US. This is also in line with the results from Verma et al., which showed good inter-rater reliability for measurements of diameters using cross-sectional imaging like MRI and a good correlation to hepatic volume29. In addition, the result seems plausible, since in CT the measurements are based on an identical data set, whereas in sonography the measurements of the two raters are based on different images that they acquire themselves.
It must be critically noted that the variability of the differences between VI from CT and volumetry is relatively high, which can be attributed to the fact that volumetry based on segmentation is more adaptable to anatomical variabilities of the liver than the method of diameter measurement. This has to be considered in particular against the background that the highest deviations between volumetry and VI were found in patients who showed a significant enlargement of the right hepatic lobe (riedel’s lobe30) or pronounced changes of the liver parenchyma as in liver cirrhosis.
A further clear advantage of volumetry by segmentation is that it can also be performed for only partial areas of the liver, which plays an important role in preoperative planning prior to liver resection. Good results in the agreement of the volumetrically determined volume both in MRI and CT with the actual postoperative liver volume could be shown in the work of Karlo et al.31. These partial volume measurements are clearly limited when determining the liver volume on the basis of VI.
Our study results further demonstrated a worse agreement of liver diameters if they were assessed using US. We could exclude possible objective reasons for that, such as patient´s body constitution (investigated by body mass index and size of the subcutaneous fat layer) as well as variations of the liver itself such as parenchymal liver diseases (investigated by liver volume, Fib4-score, laboratory data). Using our correlation, we cannot clarify for sure the reason for the lower inter-rater reliability for measurements of diameters using US with the exception of the methodological influences such as the cooperation of the patient during inspiration or the measurement inaccuracies resulting from the different posture of the ultrasound probe.
This is supported in particular by the fact that a significantly poorer correlation can be seen in measurements within the craniocaudal plane. In addition to the fact that the ultrasound conditions were most clearly impaired in this plane due to anatomical features such as the costal arch and the resulting restricted freedom of movement of the ultrasound transducer, this can be explained in particular by the respiration-dependent changes, particularly in this plane, and the associated changes in the position of the liver, lungs and diaphragm. In the other two dimensions, there is a much better correlation between the raters, indicating that the patient’s influence on the measurements is less pronounced. In addition, a good and almost good correlation between the raters in dorsoventral and mediolateral suggests a proper training.
A further potential source of error can be suspected in cases where the diameters were not determined within one position of the ultrasonic probe but had to be calculated on the basis of two positions. In this case, despite the greatest care, it cannot be guaranteed that the second setting will exactly match the plane of the first setting, which is an additional source of error.
A possible improvement approach in the future could be the use of 3D ultrasound. This method, which is widely used in gynecological imaging32, does not yet play a clinical role in imaging the liver. However, there are already first diagnostic approaches to use it for imaging the liver surface in parenchymatous diseases33 and for monitoring during interventional procedures34,35. Regarding correlation of liver measurements with clinical parameters, we found only a correlation between the FIB-4 and the inter-rater reliability in the diameter measurement in CT. Among other things, this could be due to the change of the liver surface, which can lead to deviations in the measurements due to the difficulty in determining the exact maximum diameter in case of irregularities of the liver surface in cirrhosis.
Consistently, our study shows a significant poorer inter-rater reliability in US compared to CT and results further suggest that the volume of the liver using US-based calculated VI does not lead to a valid result. In our opinion, calculation of volume index based on measurement of liver diameters using CT is an excellent, simple approach to predict liver volume. This simple technique is ready for clinical setting.
There are some limitations in our study. Study patients underwent CT exclusively for clinical reasons. As such, a clinically healthy control group was missing for comparison. However, in previous research, the excellent accuracy of calibrated volume indices to estimate liver volume in a cohort of volunteers15 was confirmed. In this study, we intentionally focus on a patient cohort to demonstrate that our approach is applicable in clinical practice. A further limitation of the study is that among the subjects included, no one had previously undergone liver surgery. A statement regarding the value of VI, for example after hemihepatectomy, is still to be investigated. In addition, it is a known obstacle in ultrasound to obtain exact orientations within a plane. Nevertheless, despite rigorous efforts (multiple control of the measurement level) and extensive previous training, there may have been measurement inaccuracies by distortions, which may have influenced the results. Furthermore, the level of experience of the examiners, especially in such an examiner-dependent procedure as ultrasound, must be considered as a possible cause for the low inter-rater agreement.
In conclusion, the volume index is an excellent approach to estimate liver volume very fast using a routinely available data set. However, an accurate estimation of liver volume requires a robust and reliable assessment of liver diameters as provided by cross-sectional imaging, such as CT. US cannot reliably measure maximum liver diameters, wherefore the usage of US-based volume indices should not be used due to its low accuracy.