Technology Insight: advances in liver imaging

Automated graded prognostic assessment for patients with hepatocellular carcinoma using machine learning

BACKGROUND

Accurate mortality risk quantification is crucial for the management of hepatocellular carcinoma (HCC); however, most scoring systems are subjective.

PURPOSE

To develop and independently validate a machine learning mortality risk quantification method for HCC patients using standard-of-care clinical data and liver radiomics on baseline magnetic resonance imaging (MRI).

METHODS

This retrospective study included all patients with multiphasic contrast-enhanced MRI at the time of diagnosis treated at our institution. Patients were censored at their last date of follow-up, end-of-observation, or liver transplantation date. The data were randomly sampled into independent cohorts, with 85% for development and 15% for independent validation. An automated liver segmentation framework was adopted for radiomic feature extraction. A random survival forest combined clinical and radiomic variables to predict overall survival (OS), and performance was evaluated using Harrell's C-index.

RESULTS

A total of 555 treatment-naïve HCC patients (mean age, 63.8 years ± 8.9 [standard deviation]; 118 females) with MRI at the time of diagnosis were included, of which 287 (51.7%) died after a median time of 14.40 (interquartile range, 22.23) months, and had median followed up of 32.47 (interquartile range, 61.5) months. The developed risk prediction framework required 1.11 min on average and yielded C-indices of 0.8503 and 0.8234 in the development and independent validation cohorts, respectively, outperforming conventional clinical staging systems. Predicted risk scores were significantly associated with OS (p < .00001 in both cohorts).

CONCLUSIONS

Machine learning reliably, rapidly, and reproducibly predicts mortality risk in patients with hepatocellular carcinoma from data routinely acquired in clinical practice.

CLINICAL RELEVANCE STATEMENT

Precision mortality risk prediction using routinely available standard-of-care clinical data and automated MRI radiomic features could enable personalized follow-up strategies, guide management decisions, and improve clinical workflow efficiency in tumor boards.

KEY POINTS

• Machine learning enables hepatocellular carcinoma mortality risk prediction using standard-of-care clinical data and automated radiomic features from multiphasic contrast-enhanced MRI. • Automated mortality risk prediction achieved state-of-the-art performances for mortality risk quantification and outperformed conventional clinical staging systems. • Patients were stratified into low, intermediate, and high-risk groups with significantly different survival times, generalizable to an independent evaluation cohort.

LiverHccSeg: A publicly available multiphasic MRI dataset with liver and HCC tumor segmentations and inter-rater agreement analysis

Accurate segmentation of liver and tumor regions in medical imaging is crucial for the diagnosis, treatment, and monitoring of hepatocellular carcinoma (HCC) patients. However, manual segmentation is time-consuming and subject to inter- and intra-rater variability. Therefore, automated methods are necessary but require rigorous validation of high-quality segmentations based on a consensus of raters. To address the need for reliable and comprehensive data in this domain, we present LiverHccSeg, a dataset that provides liver and tumor segmentations on multiphasic contrast-enhanced magnetic resonance imaging from two board-approved abdominal radiologists, along with an analysis of inter-rater agreement. LiverHccSeg provides a curated resource for liver and HCC tumor segmentation tasks. The dataset includes a scientific reading and co-registered contrast-enhanced multiphasic magnetic resonance imaging (MRI) scans with corresponding manual segmentations by two board-approved abdominal radiologists and relevant metadata and offers researchers a comprehensive foundation for external validation, and benchmarking of liver and tumor segmentation algorithms. The dataset also provides an analysis of the agreement between the two sets of liver and tumor segmentations. Through the calculation of appropriate segmentation metrics, we provide insights into the consistency and variability in liver and tumor segmentations among the radiologists. A total of 17 cases were included for liver segmentation and 14 cases for HCC tumor segmentation. Liver segmentations demonstrates high segmentation agreement (mean Dice, 0.95 ± 0.01 [standard deviation]) and HCC tumor segmentations showed higher variation (mean Dice, 0.85 ± 0.16 [standard deviation]). The applications of LiverHccSeg can be manifold, ranging from testing machine learning algorithms on public external data to radiomic feature analyses. Leveraging the inter-rater agreement analysis within the dataset, researchers can investigate the impact of variability on segmentation performance and explore methods to enhance the accuracy and robustness of liver and tumor segmentation algorithms in HCC patients. By making this dataset publicly available, LiverHccSeg aims to foster collaborations, facilitate innovative solutions, and ultimately improve patient outcomes in the diagnosis and treatment of HCC.

Improved performance and consistency of deep learning 3D liver segmentation with heterogeneous cancer stages in magnetic resonance imaging

PURPOSE

Accurate liver segmentation is key for volumetry assessment to guide treatment decisions. Moreover, it is an important pre-processing step for cancer detection algorithms. Liver segmentation can be especially challenging in patients with cancer-related tissue changes and shape deformation. The aim of this study was to assess the ability of state-of-the-art deep learning 3D liver segmentation algorithms to generalize across all different Barcelona Clinic Liver Cancer (BCLC) liver cancer stages.

METHODS

This retrospective study, included patients from an institutional database that had arterial-phase T1-weighted magnetic resonance images with corresponding manual liver segmentations. The data was split into 70/15/15% for training/validation/testing each proportionally equal across BCLC stages. Two 3D convolutional neural networks were trained using identical U-net-derived architectures with equal sized training datasets: one spanning all BCLC stages ("All-Stage-Net": AS-Net), and one limited to early and intermediate BCLC stages ("Early-Intermediate-Stage-Net": EIS-Net). Segmentation accuracy was evaluated by the Dice Similarity Coefficient (DSC) on a dataset spanning all BCLC stages and a Wilcoxon signed-rank test was used for pairwise comparisons.

RESULTS

219 subjects met the inclusion criteria (170 males, 49 females, 62.8±9.1 years) from all BCLC stages. Both networks were trained using 129 subjects: AS-Net training comprised 19, 74, 18, 8, and 10 BCLC 0, A, B, C, and D patients, respectively; EIS-Net training comprised 21, 86, and 22 BCLC 0, A, and B patients, respectively. DSCs (mean±SD) were 0.954±0.018 and 0.946±0.032 for AS-Net and EIS-Net (p<0.001), respectively. The AS-Net 0.956±0.014 significantly outperformed the EIS-Net 0.941±0.038 on advanced BCLC stages (p<0.001) and yielded similarly good segmentation performance on early and intermediate stages (AS-Net: 0.952±0.021; EIS-Net: 0.949±0.027; p = 0.107).

CONCLUSION

To ensure robust segmentation performance across cancer stages that is independent of liver shape deformation and tumor burden, it is critical to train deep learning models on heterogeneous imaging data spanning all BCLC stages.

Motion model ultrasound localization microscopy for preclinical and clinical multiparametric tumor characterization

Super-resolution imaging methods promote tissue characterization beyond the spatial resolution limits of the devices and bridge the gap between histopathological analysis and non-invasive imaging. Here, we introduce motion model ultrasound localization microscopy (mULM) as an easily applicable and robust new tool to morphologically and functionally characterize fine vascular networks in tumors at super-resolution. In tumor-bearing mice and for the first time in patients, we demonstrate that within less than 1 min scan time mULM can be realized using conventional preclinical and clinical ultrasound devices. In this context, next to highly detailed images of tumor microvascularization and the reliable quantification of relative blood volume and perfusion, mULM provides multiple new functional and morphological parameters that discriminate tumors with different vascular phenotypes. Furthermore, our initial patient data indicate that mULM can be applied in a clinical ultrasound setting opening avenues for the multiparametric characterization of tumors and the assessment of therapy response.

Motion model ultrasound localization microscopy for preclinical and clinical multiparametric tumor characterization

Super-resolution imaging methods promote tissue characterization beyond the spatial resolution limits of the devices and bridge the gap between histopathological analysis and non-invasive imaging. Here, we introduce motion model ultrasound localization microscopy (mULM) as an easily applicable and robust new tool to morphologically and functionally characterize fine vascular networks in tumors at super-resolution. In tumor-bearing mice and for the first time in patients, we demonstrate that within less than 1 min scan time mULM can be realized using conventional preclinical and clinical ultrasound devices. In this context, next to highly detailed images of tumor microvascularization and the reliable quantification of relative blood volume and perfusion, mULM provides multiple new functional and morphological parameters that discriminate tumors with different vascular phenotypes. Furthermore, our initial patient data indicate that mULM can be applied in a clinical ultrasound setting opening avenues for the multiparametric characterization of tumors and the assessment of therapy response.

The vascular structure of tumors impacts diagnosis, prognosis and drug response; however, imaging methods to analyse this important feature have been hindered by spatial resolution limitations. Here the authors present a tool called motion model ultrasound localization microscopy to morphologically and functionally characterize fine vascular networks in tumors at super-resolution.

Steatosis and hepatitis C

Hepatitis C virus (HCV) infection is a common liver disease worldwide with a high rate of chronicity (75–80%) in infected individuals. The chronic form of HCV leads to steatosis, cirrhosis and hepatocellualr carcinoma. Steatosis is prevalent in HCV patients (55%) due to a combination of viral factors (effect of viral proteins on some of the intracellular pathways) and host factors (overweight, insulin resistance, diabetes mellitus, and alcohol consumption). The response rates to treatment of chronic HCV with pegylated interferon (PEG-IFN) and (in the case of genotype-1 HCV, the most common infecting genotype in the USA) ribavirin (RBV) is low, with a sustained viral response rate ≤ 40%. Adding direct-acting antiviral agents—recently approved by the FDA—to the standard protocol has increased the response rate; however HCV-related end-stage liver disease is still the primary indication for liver transplantation in the USA. The focus of this article is on the interrelation between HCV, steatosis and metabolic syndrome.

T₁ and T₂ dual-mode MRI contrast agent for enhancing accuracy by engineered nanomaterials

One of the holy grails in biomedical imaging technology is to achieve accurate imaging of biological targets. The development of sophisticated instrumentation and the use of contrast agents have improved the accuracy of biomedical imaging. However, the issue of false imaging remains a problem. Here, we developed a dual-mode artifact filtering nanoparticle imaging agent (AFIA) that comprises a combination of paramagnetic and superparamagnetic nanomaterials. This AFIA has the ability to perform "AND logic gate" algorithm to eliminate false errors (artifacts) from the raw images to enhance accuracy of the MRI. We confirm the artifact filtering capability of AFIA in MRI phantoms and further demonstrate that artifact-free imaging of stem cell migration is possible in vivo.

Changes in the management of benign liver tumours: an analysis of 285 patients

OBJECTIVES

Benign liver tumours (BLTs) are common and their management remains controversial. This study assesses the safety of a selective management approach.

METHODS

Patients with BLT were identified from an institutional database. Patients with simple cysts or an incidental BLT in the setting of metastasis or concomitant malignancy were excluded.

RESULTS

A total of 285 patients presenting during the period from January 1992 to December 2009 with haemangioma (53.0%), focal nodular hyperplasia (23.9%), adenoma (10.2%) or indeterminate/other lesions (13.0%) were evaluated. Of these, 117 patients (41.1%) underwent immediate resection and 168 patients (58.9%) were followed with serial imaging (median follow-up: 30 months). During observation, eight patients (4.8%) underwent resection for tumour growth, inability to exclude malignancy or symptoms; no patients demonstrated malignant transformation or tumour-related complications. During the study period, the number of BLTs evaluated and the proportion of patients observed increased from 129 BLTs of which 36.4% were observed in 1992-2002 to 156 BLTs of which 71.2% were observed in 2003-2009 (P < 0.001). Diagnostic uncertainty led to resection in 29.5% of patients during the earlier period, but in only 13.4% during the more recent 7 years (P < 0.05).

CONCLUSIONS

Asymptomatic BLTs without concern for malignancy or adenoma can be safely observed with minimal risk for misdiagnosis. Patients selected for observation rarely require resection or develop tumour-related complications.

Recent advances in imaging hepatic fibrosis and steatosis

Liver disease is an increasing cause of morbidity and mortality worldwide. Currently, the gold standard for diagnosis and assessment of parenchymal disease is histopathological assessment of a percutaneous or transjugular liver biopsy. The risks and limitations of this technique are well recognized and as a result, significant effort has gone into the development of novel noninvasive methods of diagnosis and longitudinal assessment. Imaging techniques have improved significantly over the past decade and new technologies are beginning to enter clinical practice. Ultrasound, computed tomography and MRI are the main modalities currently used, but novel MRI-based techniques will have an increasing role. While there has been extensive research into the imaging of focal liver disease, the evidence base for imaging in diffuse disease has also undergone recent rapid development, particularly in the assessment of fibrosis and steatosis. Both of these abnormalities of the parenchyma can lead to cirrhosis and/or hepatocellular carcinoma and represent an important opportunity for detection of early liver disease. We discuss the recent advances in liver imaging techniques and their role in the diagnosis and monitoring of diffuse liver disease, with a focus on their current and potential clinical relevance and whether they may replace or augment liver biopsy. We also discuss techniques currently under development and their potential clinical applications in the future.

Differentiation of focal liver lesions using three-dimensional ultrasonography: Retrospective and prospective studies

AIM: To differentiate focal liver lesions based on enhancement patterns using three-dimensional ultrasonography (3D US) with perflubutane-based contrast agent.

METHODS: Two hundred and eighty two patients with focal liver lesions, including 168 hepatocellular carcinomas (HCCs), 63 metastases, 40 hemangiomas and 11 focal nodular hyperplasias (FNHs), were examined by 3D US with perflubutane-based contrast agent. Tomographic ultrasound images and sonographic angiograms were reconstructed. Among 282 lesions, enhancement patterns of 163 lesions between January 2007 and October 2007 were analyzed retrospectively. Then from November 2007 to May 2008, compared with contrast-enhanced (CE) 2D US, CE 3D US was performed on 119 lesions for prospective differential diagnosis. Sensitivity, specificity, area under receiver operating characteristic curve (A_z) and inter-reader agreement were assessed.

RESULTS: With the tridimensional view, dominant enhancement patterns were revealed as diffuse enhancement or peripheral ring-like enhancement, followed with washout change for HCCs or metastases, respectively, and peripheral nodular enhancement or diffuse enhancement with spoke-wheel arteries, followed by persistent enhancement for hemangiomas or FNHs, respectively. At CE 3D US, the prospective differentiation of lesions showed sensitivity 92% (mean for two readers), specificity 91% and A_z value 0.95 for HCCs, 84%, 97%, and 0.95 for metastases, 91%, 98%, and 0.98 for hemangiomas and 80%, 99%, and 0.99 for FNHs, respectively, while good to excellent inter-reader agreement was achieved. No significant difference exists between prospective diagnosis accuracy at CE 3D US and that at CE 2D US.

CONCLUSION: CE 3D US provides a spatial perspective for liver tumor enhancement, and could help in differentiating focal liver lesions.

Optimized multidetector computed tomographic protocol for the diagnosis of active obscure gastrointestinal bleeding: a feasibility study

OBJECTIVE

The purpose of this feasibility study was to prospectively evaluate an optimized multidetector computed tomographic protocol for the diagnosis of active obscure gastrointestinal bleeding (OGIB).

METHODS

Between October 2006 and February 2008, patients admitted for active OGIB were included in this prospective unicenter study. Water was administered orally and rectally as neutral luminal contrast material. A contrast-enhanced 16-row multidetector computed tomography (MDCT) was performed in the arterial and venous phases. Mesenteric digital subtraction angiography was carried out immediately after MDCT as standard of reference.

RESULTS

Six patients were included in this study. Multidetector computed tomography identified the bleeding site and source in 5 (83%) of the patients. Digital subtraction angiography was performed in 4 patients, and the result was positive in 1 (25%) of the patients. Multidetector computed tomography detected the site and source of bleeding in 2 patients whose digital subtraction angiographic result was negative.

CONCLUSIONS

The results of this feasibility study indicate that optimized MDCT is an excellent diagnostic tool for the diagnosis of active OGIB.

Can imaging modalities diagnose and stage hepatic fibrosis and cirrhosis accurately?

The accurate diagnosis and staging of hepatic fibrosis is crucial for prognosis and treatment of liver disease. The current gold standard, liver biopsy, cannot be used for population-based screening, and has well known drawbacks if used for monitoring of disease progression or treatment success. Our objective was to assess performance and promise of radiologic modalities and techniques as alternative, noninvasive assessment of hepatic fibrosis. A systematic review was conducted. Six hundred twenty-eight studies were identified via electronic search. One hundred fifty-three papers were reviewed. Most described techniques that could differentiate between cirrhosis or severe fibrosis and normal liver. Accurate staging of fibrosis or diagnosis of mild fibrosis was often not achievable. Ultrasonography is the most common modality used in the diagnosis and staging of hepatic fibrosis. Elastographic measurements, either ultrasonography-based or magnetic resonance-based, and magnetic resonance diffusion weighted imaging, show the most promise for accurate staging of hepatic fibrosis. Most currently available imaging techniques can detect cirrhosis or significant fibrosis reasonably accurately. However, to date only magnetic resonance elastography has been able to stage fibrosis or diagnose mild disease. Utrasonographic elastography and magnetic resonance diffusion weighted appear next most promising.