Long-term liver lesion tracking in contrast-enhanced ultrasound videos via a siamese network with temporal motion attention.

Propose: Contrast-enhanced ultrasound has shown great promises for diagnosis and monitoring in a wide range of clinical conditions. Meanwhile, to obtain accurate and effective location of lesion in contrast-enhanced ultrasound videos is the basis for subsequent diagnosis and qualitative treatment, which is a challenging task nowadays. Methods: We propose to upgrade a siamese architecture-based neural network for robust and accurate landmark tracking in contrast-enhanced ultrasound videos. Due to few researches on it, the general inherent assumptions of the constant position model and the missing motion model remain unaddressed limitations. In our proposed model, we overcome these limitations by introducing two modules into the original architecture. We use a temporal motion attention based on Lucas Kanade optic flow and Karman filter to model the regular movement and better instruct location prediction. Moreover, we design a pipeline of template update to ensure timely adaptation to feature changes. Results: Eventually, the whole framework was performed on our collected datasets. It has achieved the average mean IoU values of 86.43% on 33 labeled videos with a total of 37,549 frames. In terms of tracking stability, our model has smaller TE of 19.2 pixels and RMSE of 27.6 with the FPS of 8.36 ± 3.23 compared to other classical tracking models. Conclusion: We designed and implemented a pipeline for tracking focal areas in contrast-enhanced ultrasound videos, which takes the siamese network as the backbone and uses optical flow and Kalman filter algorithm to provide position prior information. It turns out that these two additional modules are helpful for the analysis of CEUS videos. We hope that our work can provide an idea for the analysis of CEUS videos.

Liver cirrhosis and tumor location can affect the range of intrahepatic microwave ablation zone.

BACKGROUND: The range of an ablation zone (AZ) plays a crucial role in the treatment effect of microwave ablation (MWA). The aim of this study was to analyze the factors influencing the AZ range. METHODS: Fourteen factors in four areas were included: patient-related factors (sex, age), disease-related factors (tumor location, liver cirrhosis), serological factors (ALT, AST, total protein, albumin, total bilirubin, direct bilirubin, and platelets), and MWA parameters (ablation time, power, and needle type). Multiple sequence MRI was used to delineate AZ by three radiologists using 3D Slicer. MATLAB was used to calculate the AZ length, width, and area of the largest section. Linear regression analysis was used to analyze influencing factors. Moreover, a subgroup analysis was conducted for patients with viral hepatitis. RESULT: 220 patients with 290 tumors were included between 2010-2021. In addition to MWA parameters, cirrhosis and tumor location were significant factors that influenced AZ (p < 0.001). The standardized coefficient (beta) of cirrhosis (cirrhosis vs. non-cirrhosis) was positive, which meant cirrhosis would lead to a decrease in AZ range. The beta of tumor location (near the hepatic hilar zone, intermediate zone, and periphery zone) was negative, indicating that AZ range decreased as the tumor location approached the hepatic hilum. For viral hepatitis patients, Fibrosis 4 (FIB4) score was a significant factor influencing AZ (p < 0.001), and the beta was negative, indicating that AZ range decreased as FIB4 increased. CONCLUSION: Liver cirrhosis, tumor location, and FIB4 affect the AZ range and should be considered when planning MWA parameters.

Peripheral immune factors aiding clinical parameter for better early recurrence prediction of hepatocellular carcinoma after thermal ablation.

OBJECTIVES: Current predictors are largely unsatisfied for early recurrence (ER) of hepatocellular carcinoma (HCC) after thermal ablation. We aimed to explore the prognostic value of peripheral immune factors (PIFs) for better ER prediction of HCC after thermal ablation. METHODS: Patients who received peripheral blood mononuclear cells (PBMCs) tests before thermal ablation were included. Clinical parameters and 18 PIFs were selected to construct ModelClin, ModelPIFs and the hybrid ModelPIFs-Clin. Model performances were evaluated using area under the curve (AUC), and recurrence-free survival (RFS) were analyzed by Kaplan-Meier analysis and log-rank tests. RESULTS: 244 patients were included and were randomly divided in 3:1 ratio to discovery and validation cohorts. Clinical parameters including tumor size and AFP, and PIFs including neutrophils, platelets, CD3+CD16+CD56+ NKT and CD8+CD28- T lymphocytes were selected. The ModelPIFs-Clin showed increase in predictive performance compared with ModelClin, with the AUC improved from 0.664 (95%CI:0.588-0.740) to 0.801 (95%CI:0.734-0.867) in discovery cohort (p < 0.0001), and from 0.645 (95%CI:0.510-0.781) to 0.737(95%CI:0.608-0.865) in validation cohort (p = 0.1006). ModelPIFs-Clin enabled ER risk stratification of patients. Patients predicted in ModelPIFs-Clin high-risk subgroup had a poor RFS compared with those predicted as ModelPIFs-Clin low-risk subgroup, with the median RFS was 18.00 month versus 100.78 month in discovery cohort (p < 0.0001); and 24.00 month versus 60.35 month in validation cohort (p = 0.288). Patients in different risk subgroups exhibited distinct peripheral immune contexture. CONCLUSIONS: Peripheral immune cells aiding clinical parameters boosted the prediction ability for ER of HCC after thermal ablation, which be helpful for pre-ablation ER risk stratification.

The tumor ghost on MRI after microwave ablation for hepatocellular carcinoma: A new modality to assess the ablative margin.

PURPOSE: The ablated tumor ghost can be visually distinguished on MR images after ablation. This retrospective study aimed to assess the performance of tumor ghost on post-ablation contrast-enhanced MRI with excellular contrast agent gadolinium-DTPA in evaluating the ablative margin of hepatocellular carcinoma (HCC) after microwave ablation (MWA). METHOD: 315 HCC lesions less than 5 cm in 287 patients completely treated by MWA were enrolled in the study. The tumor ghost was characterized as a lower signal intensity area than the surrounding tissues of the ablation zone on T1WI imaging. The ablation margin (AM) status was classified into AM0 (＞5mm) and AM1 (＜5mm) according to the minimum distance between the tumor ghost and ablated zone. Inter-observer agreement between two radiologists on the AM assessment was analyzed using the Cohen κ coefficient. Multivariate analysis using Cox proportional hazard model was performed to investigate independent risk factors for LTP. RESULTS: 175 and 140 tumors were evaluated as AM0 and AM1 through tumor ghost. The inter-observer agreement level between two radiologists for assessment of AM was good (κ coefficient = 0.752, 95 % confidence interval: 0.679-0.825, p < 0.001). The mediate follow-up period was 32.2 months (range 3.0-60.8 months). The incidence of LTP in the AM0 lesions and AM1 lesions was 6.3 % (11/175) and 20.0 % (28/140), respectively. AM status was identified as an independent prognostic factor for LTP (HR 3.057, 95 % CI, 1.445-6.470, p = 0.003). CONCLUSIONS: The assessment of the AM by tumor ghost on post-ablation MRI is an accurate and efficiently method for evaluating the completeness of microwave ablation for hepatocellular carcinoma.

A Hybrid Machine Learning Model Based on Semantic Information Can Optimize Treatment Decision for Naïve Single 3-5-cm HCC Patients.

Background: Tumor recurrence is an abomination for hepatocellular carcinoma (HCC) patients receiving local treatment. Purpose: The aim of the study was to build a hybrid machine learning model to recommend optimized first treatment (laparoscopic hepatectomy [LH] or microwave ablation [MWA]) for naïve single 3-5-cm HCC patients based on early recurrence (ER, ≤2 years) probability. Methods: This retrospective study collected 20 semantic variables of 582 patients (LH: 300, MWA: 282) from 13 hospitals with at least 24 months follow-up. Both groups were divided into training, validation, and test set, respectively. Five algorithms (logistics regression, random forest, neural network, stochastic gradient boosting, and eXtreme Gradient Boosting [XGB]) were used for model building. A model with highest area under the receiver operating characteristic curve (AUC) in a validation set of LH and MWA was selected to connect as a hybrid model which made decision based on ER probability. Model testing was performed in a comprehensive set comprising LH and MWA test sets. Results: Four variables in each group were selected to build LH and MWA models, respectively. LH-XGB model (AUC = 0.744) and MWA-stochastic gradient method (AUC = 0.750) model were selected for model building. In the comprehensive set, a treatment confusion matrix was established based on recommended and actual treatment. The predicted ER probabilities were comparable with the actual ER rates for various types of patients in matrix (p > 0.05). ER rate of patients whose actual treatment consistent with recommendation was lower than that of inconsistent patients (LH: 21.2% vs. 46.2%, p = 0.042; MWA: 26.3% vs. 54.1%, p = 0.048). By recommending optimal treatment, the hybrid model can significantly reduce ER probability from 38.2% to 25.6% for overall patients (p < 0.001). Conclusions: The hybrid model can accurately predict ER probability of different treatments and thereby provide reliable evidence to make optimal treatment decision for patients with single 3-5-cm HCC.

Are all local tumour progressions of HCC related to thermal ablation? A study of the causes and classification of local tumour progression.

OBJECTIVES: Local tumour progression (LTP) is believed to be a negative consequence of imperfect thermal ablation, but we wondered if all LTP is truly due to imperfect ablation. METHODS: This study included 185 LTPs occurring within 1 cm of the ablation zone (AZ) after clinical curative thermal ablation for ≤ 5 cm hepatocellular carcinoma between 2010 and 2019. The AZ was divided into 8 quadrants by coronal, sagittal, and horizontal planes. Two methods, visual assessment through pre- and post-MRI (VA) and tumour mapping for 3D visualisation pre- and post-MRI fusion (MF), were used to assess which AZ quadrant included the shortest ablation margin (AM) by three doctors. LTP subclassification was based on whether LTP contacted the AZ margin (contacted LTP and dissociated-type LTP) and occurrence at different time points (12, 18, and 24 months). RESULTS: Fleiss's Kappa of VA and MF was 0.769 and 0.886, respectively. Cohen's Kappa coefficient between VA and MF was 0.830. For all LTPs, 98/185 (53.0%) occurred in the shortest AM quadrant, which showed a significant central tendency (p < 0.001). However, only 8/51 (15.7%) dissociated - type LTPs and 6/39 (15.4%) LTPs after 24 months occurred in the shortest AM quadrant, which showed no evenly distributed difference (p = 0.360 and 0.303). CONCLUSIONS: MF is an accurate and convenient method to assess the shortest AM quadrant. LTP is a central tendency in the shortest AM quadrant, but dissociated-type and LTPs after 24 months are not, and these LTP types could be considered nonablation-related LTPs. KEY POINTS: • LTPs are not evenly distributed around the AZ. More than half of LTPs occur in the shortest AM quadrant. • Subgroup analysis showed that the occurrence of contacted-type LTPs (tumour margin has direct contact with the AZ) within 24 months after ablation indeed had a high proportion in the shortest AM quadrant, and they could be called ablation-related LTPs. • However, the dissociated-type LTPs (tumour margin adjacent to but not in contact with the AZ) or LTPs occurring beyond 24 months after ablation were evenly distributed around the AZ, and they could be called nonablation-related LTPs.

Radiomics analysis of ultrasound to predict recurrence of hepatocellular carcinoma after microwave ablation.

OBJECTIVE: To develop and validate an ultrasonic radiomics model for predicting the recurrence and differentiation of hepatocellular carcinoma (HCC). Convolutional neural network (CNN) ResNet 18 and Pyradiomics were used to analyze gray-scale-ultrasonic images to predict the prognosis and degree of differentiation of HCC. METHODS: This retrospective study enrolled 513 patients with HCC who underwent preoperative grayscale-ultrasonic imaging, and their clinical characteristics were observed. Patients were randomly divided into training (n = 413) and validation (n = 100) cohorts. CNN ResNet 18 and Pyradiomics were used to analyze ultrasonic images of HCC and peritumoral images to develop a prognostic and differentiation model. Clinical characteristics were integrated into the radiomics model and patients were stratified into high- and low-risk groups. The predictive effect was evaluated using the C-index and receiver operating characteristic (ROC) curve. RESULTS: The model combined with ResNet 18 and clinical characteristics achieved a good predictive ability. The C-indices of early recurrence (ER), late recurrence (LR), and recurrence-free survival (RFS) were 0.695 (0.561-0.789), 0.715 (0.623-0.800) and 0.721 (0.647-0.795), respectively, in the validation cohort, which was superior to the clinical model and ultrasonic semantic model. The model could stratify patients into high- and low-risk groups, which showed significant differences (p < 0.001) in ER, LR, and RFS. The area under the curve for predicting the degree of HCC differentiation was 0.855 and 0.709 in the training and validation cohorts, respectively. CONCLUSION: We developed and validated a radiomics model to predict HCC recurrence and HCC differentiation, which could also acquire pathological information in a noninvasive manner.KEY RESULTSA hepatocellular carcinoma (HCC) prognostic prediction model was developed and validated by convolutional neural network (CNN) ResNet 18-based gray-scale ultrasound (US).A differentiation of HCC prediction model was developed for preoperative prediction avoiding invasive operation.Compared with Pyradiomics, CNN ResNet was more suitable for extracting information from US images.

Percutaneous microwave ablation versus robot-assisted hepatectomy for early hepatocellular carcinoma: A real-world single-center study.

BACKGROUND: Both microwave ablation and robot-assisted hepatectomy are representative minimally invasive treatments for early hepatocellular carcinoma. Our study compares the practicability and medium-term therapeutic efficacy between them. METHODS: Patients with early HCC treated by MWA or RH from 2013 to 2019 were included. Propensity score matching (PSM) and inverse probability of treatment weight (IPTW) were used to minimize baseline imbalance. Operation trauma, postoperative recovery, complications, cost and oncological efficacy were compared. RESULTS: 401 patients with a median follow-up of 28 months were included (MWA n = 240; RH n = 161). After PSM, 3-year recurrence-free survival (RFS), overall survival (OS) and cancer-specific survival (CSS) of MWA group and RH group were 52.2% vs 65.8%, 91.5% vs 91.3% and 91.5% vs 91.3%, respectively. OS and CSS were comparable (p = 0.44 and 0.96), while RFS of MWA was slightly lower but not significant (p = 0.097). The above results after IPTW followed the same trend. After PSM, MWA showed advantages in operation time and blood loss, while RH performed better in postoperative liver function. There was no significant difference in incidence of severe complications between two groups. CONCLUSIONS: For early HCC parents, both treatments can achieve good, safe and comparable medium-term therapeutic effects. MWA is more minimally invasive, while RH has better accuracy and causes less damage to liver parenchyma.