Hepatobiliary phase hypointensity predicts progression to hepatocellular carcinoma for intermediate-high risk observations, but not time to progression

Choi

et al. 2022

Liver International

Background and Aims We assessed the imaging outcomes of Liver Imaging Reporting and Data System (LI‐RADS) v2018 categories in prospective hepatocellular carcinoma (HCC) surveillance cohort and determined imaging features significantly predictive of progression to a malignant LI‐RADS category. Methods The imaging outcomes of 120 patients (162 observations) prospectively enrolled between November 2011 and August 2012 were analysed according to LI‐RADS v2018. Cumulative incidences for progression to a malignant category (LR‐5 or LR‐M) and LR‐4 or higher were calculated for each baseline category and compared using log‐rank tests. Clinical variables and imaging features significantly predictive of progression to a malignant category were evaluated using Cox proportional hazards modelling. Results The 162 observations were initially categorized into 60 LR‐2, 75 LR‐3 and 27 LR‐4. For LR‐4 observations, the 1‐year, 3‐year and 5‐year cumulative incidences of progression to a malignant category were 18.5% (95% confidence interval, 6.6–35.2%), 43.0% (23.1–61.5%) and 52.5% (25.9–73.5%), which were significantly higher than those of LR‐2 and LR‐3 (p < .001). For LR‐3, the 1‐year, 3‐year and 5‐year cumulative incidences of progression to LR‐4 or higher were 4.1% (1.1–10.4%), 13.9% (6.7–23.6%) and 23.1% (12.7–35.4%), which were significantly higher than that of LR‐2 (p = .009). In multivariable analysis, size ≥1.0 cm (hazard ratio [HR] = 2.58, 1.04–6.40) and nonrim arterial‐phase hyperenhancement (HR = 2.45, 1.11–5.42) were significantly independently associated with progression to a malignant category. Conclusion Long‐term imaging outcomes differed significantly according to LI‐RADS category. Size ≥1.0 cm and nonrim arterial‐phase hyperenhancement were imaging features significantly predictive of progression to a malignant category.

Section: Discussioncontrasting

confidence: 91%

Liver Imaging Reporting and Data System categories: Long‐term imaging outcomes in a prospective surveillance cohort

Choi

et al. 2022

Liver International

“…Morimoto et al 35 described non‐hypervascular hypointense nodules on MRI as the only factor correlated with subsequent HCC development after DAA, while Toyoda et al 36 analysed the incidence of hypervascularization in non‐hypervascular hypointense nodules on gadoxetate disodium‐MRI demonstrating no difference in the development of arterial phase hyperenhancement in patient with eradicated HCV, compared to patient with persistent HCV infection. In clinical practice, indeterminate liver nodules may carry a higher risk of malignancy progression according to the combination of specific imaging features 19,37 . The LI‐RADS algorithm may predict the outcome and time to progression based on the initial LI‐RADS category and help identify the subset of observations that require a more aggressive management 20,38 .…”

Section: Discussionmentioning

confidence: 99%

“…Other studies have reported similar trends regarding the risk of development of de novo HCC, with most evidences suggesting that DAA does not increase, but even reduces, the incidence of HCC in cirrhotic patients. 9,10,11,12,29,30 To our knowledge this is the first study evaluating the progression of 19,37 The LI-RADS algorithm may predict the outcome and time to progression based on the initial LI-RADS category and help identify the subset of observations that require a more aggressive management. 20,38 In our study the overall cumulative incidence of progression to LR-5 or LR-M was significantly higher in observations were already advanced in the hepatocarcinogenesis process before DAA therapy.…”

Section: Discussionmentioning

confidence: 99%

“…These data were limited by heterogeneous definition of indeterminate liver nodules and lack of information about the outcome of indeterminate nodules itself. [33][34][35][36] 19,37 The LI-RADS algorithm may predict the outcome and time to progression based on the initial LI-RADS category and help identify the subset of observations that require a more aggressive management. 20,38 In our study the overall cumulative incidence of progression to LR-5 or LR-M was significantly higher in Interestingly, progression to LR-5 after DAA therapy most commonly occurred within 6 mo (67% of progressed observations in our study).…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Long‐term evolution of LI‐RADS observations in HCV‐related cirrhosis treated with direct‐acting antivirals

Cannella

Vernuccio²,

Celsa

et al. 2021

Liver International

Self Cite

Background & Aims The risk of progression of indeterminate observations to hepatocellular carcinoma (HCC) after direct‐acting antivirals (DAA) is still undetermined. To assess whether DAA therapy changes the risk of progression of observations with low (LR‐2), intermediate (LR‐3) and high (LR‐4) probability for HCC in cirrhotic patients and to identify predictors of progression. Methods This retrospective study included cirrhotic patients treated with DAA who achieved sustained virological response between 2015 and 2019. A total of 68 patients had pre‐DAA indeterminate observations and at least six months CT/MRI follow‐up before and after DAA. Two radiologists reviewed CT/MRI studies to categorize observations according to the LI‐RADSv2018 and assess the evolution on subsequent follow‐ups. Predictors of evolutions were evaluated by using the Cox proportional hazard model, Kaplan‐Meier method and log‐rank test. Results A total of 109 untreated observations were evaluated, including 31 (28.4%) LR‐2, 67 (61.5%) LR‐3 and 11 (10.1%) LR‐4. During a median follow‐up of 41 months, 17.4% and 13.3% of observations evolved to LR‐5 or LR‐M and LR‐5, before and after DAA respectively (P = .428). There was no difference in rate of progression of neither LR‐2 (P = 1.000), LR‐3 (P = .833) or LR‐4 (P = .505). At multivariate analysis, only initial LI‐RADS category was an independent predictor of progression to LR‐5 or LR‐M for all observations (hazard ratio 6.75, P < .001), and of progression to LR‐5 after DAA (hazard ratio 4.34, P = .047). Conclusions DAA therapy does not increase progression of indeterminate observations to malignant categories. The initial LI‐RADS category is an independent predictor of observations upgrade.

“…However, imaging criteria are still limited in terms of sensitivity because of the need to maintain a high specificity for the diagnosis [ 5 , 9 , 10 ]. As such, many recent studies have been published thus far to improve the diagnosis of liver cancer, especially in the setting of cirrhosis, and even to predict the progression of a liver lesion into cancer [ 11 , 12 , 13 , 14 ]. Despite these improvements, many liver lesions, especially in the setting of cirrhosis, are still deemed indeterminate [ 9 ].…”

Section: Introductionmentioning

confidence: 99%

A Multi-Scale and Multi-Level Fusion Approach for Deep Learning-Based Liver Lesion Diagnosis in Magnetic Resonance Images with Visual Explanation

Wan

Zheng

Liu³

et al. 2021

Life

Many computer-aided diagnosis methods, especially ones with deep learning strategies, of liver cancers based on medical images have been proposed. However, most of such methods analyze the images under only one scale, and the deep learning models are always unexplainable. In this paper, we propose a deep learning-based multi-scale and multi-level fusing approach of CNNs for liver lesion diagnosis on magnetic resonance images, termed as MMF-CNN. We introduce a multi-scale representation strategy to encode both the local and semi-local complementary information of the images. To take advantage of the complementary information of multi-scale representations, we propose a multi-level fusion method to combine the information of both the feature level and the decision level hierarchically and generate a robust diagnostic classifier based on deep learning. We further explore the explanation of the diagnosis decision of the deep neural network through visualizing the areas of interest of the network. A new scoring method is designed to evaluate whether the attention maps can highlight the relevant radiological features. The explanation and visualization make the decision-making process of the deep neural network transparent for the clinicians. We apply our proposed approach to various state-of-the-art deep learning architectures. The experimental results demonstrate the effectiveness of our approach.