In late 1895, Wilhelm Roentgen ignited a disease diagnostics revolution with the discovery of x-rays, which opened a new era for medical imaging. Imaging is swiftly evolving from being primarily a disease diagnostic tool to also having an indispensable role in the context of personalized management. Nowadays, liver imaging has been routinely performed in clinical practice, and emerging relevant techniques further accelerate its application in precision medicine.
Liver imaging for determining disease etiology, and for staging advanced fibrosis and cirrhosis, was comprehensively discussed in a New England Journal of Medicine paper in December 2017. Although imaging approaches have not replaced liver biopsy yet, they have sharply reduced the demand for invasive procedures. This is also true for risk stratification of portal hypertension in compensated advanced chronic liver disease. Imaging evidence of portocollateral circulation or a flow reversal is sufficient to diagnose clinically significant portal hypertension. However, the low sensitivity of liver imaging makes it difficult for the early diagnosis and primary prophylaxis of portal hypertension bleeding. Hepatic venous pressure gradient (HVPG) is currently the only validated approach to stratify the risk of portal hypertension, but its invasiveness, cost, and requirement for specific equipment have hindered the routine and wider use of HVPG in practice. The computed tomography imaging-based three-dimensional (3D) modeling combined with computational fluid dynamics in the CHESS1601 trial have made it possible to develop a noninvasive surrogate of HVPG. Additionally, magnetic resonance hemodynamic parameters have also showed a satisfactory correlation with HVPG. Taken together, imaging techniques have greatly improved our ability to precisely diagnose liver diseases. However, heterogeneity across studies of imaging tools generated from different etiologies and populations cannot be ignored. We would like to particularly underline the need for reliable liver imaging techniques that are able to monitor HVPG changes, because this is an unmet need with major clinical implications.
Liver imaging also contributes to a novel paradigm of “precision liver surgery”, featuring certain surgical interventions-guiding precise management. Precision liver surgery integrates advanced imaging techniques, eg, imaging-based 3D visualization, and equips the surgeon with real anatomy of the liver including the location of lesions, the traverse and territory of vessels, and the spatial relationship between lesions and vessels. The Western current best practices introduce the mixed reality technique for simulating 3D images and reduce the offset between visualization and working space allowing for improved spatial-visual approximation of image and patient. A suitable standard for segmenting image masks and texture mapping of hepatocellular carcinoma and vessels has been developed. The Chinese Society of Digital Medicine has also achieved consensus on the 3D visualization technology for precision liver surgery of complicated hepatocellular carcinoma, hilar cholangiocarcinoma, and hepatolithiasis. Moreover, imaging-based surgical planning systems also allow the comparison of various strategies of “virtual resection”, which provides useful information to help the surgeon select the optimal procedure. This tool is valuable in evaluating and predicting the resectability of a liver lesion, especially when lesions involve important anatomic structures and require complex major hepatectomy. In addition, one critical issue regarding the safety limit of liver resection is the functional capacity of the liver remnant. Combined with multimodality imaging-based volumetric measuring, surgeons can accurately assess and plan the functional volume of liver remnant. Currently, the imaging technological foundation of precision liver surgery has been established, but the wide applicability of this approach in clinical practice is yet to come. We would like to underline the critical need of novel imaging-based multidisciplinary technologies, such as 3D quantitative regional assessment of liver function, intraoperative real-time liver surgical navigation, and digital-assisted decision-making system. The mixed reality technique remains challenging, as the liver undergo major deformations due to respiratory motion, manipulation, and the interaction with the surgical instruments. The challenges also include sensor and display technologies as well as object-recognition algorithms and biomechanical modeling. These are the unmet needs of precision liver surgery.
Artificial intelligence-based high-throughput computing now enables the extraction of innumerable quantitative features from liver imaging. The conversion of digital images into mineable high-dimensional data, termed radiomics, is motivated by the concept that standard-of-care images contain information that reflects underlying pathophysiology. This information can be harnessed through quantitative image analyses and leveraged via clinical-decision support systems to improve precision medicine. In a recent paper, published in May 2018 in the journal Gut, a deep learning, convolutional neural network, radiomics method of shear wave elastography images (DLRE) was developed for staging liver fibrosis. Accordingly, DLRE showed the best performance in predicting fibrosis stages in chronic hepatitis B patients in the multicenter study. In addition, a novel radiomics signature as surrogate of HVPG was developed and prospectively validated in the CHESS1701 trial as an accurate detection of clinically significant portal hypertension and variceal hemorrhage risk in advanced chronic liver disease. Recently, a noninvasive nomogram integrated clinical factors, radiological characteristics and radiomics features, termed ModelCRR, was established to assess the progression-free survival difference between liver resection and transarterial chemoembolization in hepatocellular carcinoma. Currently, the field of radiomics lacks standardized evaluation of both the scientific integrity and the clinical relevance. Rigorous evaluation criteria need to be followed. Besides, there is an urgent need for homogeneous assessment and optimal collection of diverse sources of liver images in a quantitative manner. More importantly, radiomics approaches should incorporate reproducibility assessments that accurately and robustly predicts the reference standard and subsequently supports precision medicine.
EBioMedicine aims to form a community that spans the interface and creates an opportunity for translational research in specialty subjects of hepatology, radiology and bioengineering, and to move the field of liver imaging in precision medicine forward.
Disclosure
The authors declare no conflicts of interest.