Sequence-Type Classification of Brain MRI for Acute Stroke Using a Self-Supervised Machine Learning Algorithm.

We propose a self-supervised machine learning (ML) algorithm for sequence-type classification of brain MRI using a supervisory signal from DICOM metadata (i.e., a rule-based virtual label). A total of 1787 brain MRI datasets were constructed, including 1531 from hospitals and 256 from multi-center trial datasets. The ground truth (GT) was generated by two experienced image analysts and checked by a radiologist. An ML framework called ImageSort-net was developed using various features related to MRI acquisition parameters and used for training virtual labels and ML algorithms derived from rule-based labeling systems that act as labels for supervised learning. For the performance evaluation of ImageSort-net (MLvirtual), we compare and analyze the performances of models trained with human expert labels (MLhumans), using as a test set blank data that the rule-based labeling system failed to infer from each dataset. The performance of ImageSort-net (MLvirtual) was comparable to that of MLhuman (98.5% and 99%, respectively) in terms of overall accuracy when trained with hospital datasets. When trained with a relatively small multi-center trial dataset, the overall accuracy was relatively lower than that of MLhuman (95.6% and 99.4%, respectively). After integrating the two datasets and re-training them, MLvirtual showed higher accuracy than MLvirtual trained only on multi-center datasets (95.6% and 99.7%, respectively). Additionally, the multi-center dataset inference performances after the re-training of MLvirtual and MLhumans were identical (99.7%). Training of ML algorithms based on rule-based virtual labels achieved high accuracy for sequence-type classification of brain MRI and enabled us to build a sustainable self-learning system.

Transient severe motion artifacts on gadoxetic acid-enhanced MRI: risk factor analysis in 2230 patients.

OBJECTIVES: To determine risk factors for transient severe motion (TSM) artifact on arterial phase of gadoxetic acid-enhanced MRI using a large cohort. METHODS: A total of 2230 patients who underwent gadoxetic acid-enhanced MRI was consecutively included. Two readers evaluated respiratory motion artifact on arterial phase images using a 5-point grading scale. Clinical factors including demographic data, underlying disease, laboratory data, presence of ascites and pleural effusion, and previous experience of gadoxetic acid-enhanced MRI were investigated. Univariable and multivariable logistic regression analyses were performed to determine significant risk factors for TSM. Predictive value of TSM was calculated according to the number of significant risk factors. RESULTS: Overall incidence of TSM was 5.0% (111/2230). In the multivariable analysis, old age (≥ 65 years; odds ratio [OR] = 2.01 [95% CI, 1.31-3.07]), high body mass index (≥ 25 kg/m2; OR = 1.76 [1.18-2.63]), chronic obstructive pulmonary disease (OR = 6.11 [2.32-16.04]), and moderate to severe pleural effusion (OR = 3.55 [1.65-7.65]) were independent significant risk factors for TSM. Presence of hepatitis B (OR = 0.66 [0.43-0.99]) and previous experience of gadoxetic acid-enhanced MRI (OR = 0.52 [0.33-0.83]) were negative risk factors for TSM. When at least one of the significant factors was present, the predictive risk was 5.7% (109/1916), whereas it was 16.3% (17/104) when at least four factors were present. CONCLUSION: Knowing risk factors for transient severe motion artifact on gadoxetic acid-enhanced MRI can be clinically useful for providing diagnostic strategies more tailored to individual patients. KEY POINTS: • Old age, high body mass index, chronic obstructive pulmonary disease, and moderate to severe pleural effusion were independent risk factors for transient severe motion artifact on gadoxetic acid-enhanced MRI. • Patients with hepatitis B or previous experience of gadoxetic acid-enhanced MRI were less likely to show transient severe motion artifact. • As the number of risk factors for transient severe motion artifact increased, the predicted risk for it also showed a tendency to increase.

Development and validation of an ensemble artificial intelligence model for comprehensive imaging quality check to classify body parts and contrast enhancement.

BACKGROUND: Despite the dramatic increase in the use of medical imaging in various therapeutic fields of clinical trials, the first step of image quality check (image QC), which aims to check whether images are uploaded appropriately according to the predefined rules, is still performed manually by image analysts, which requires a lot of manpower and time. METHODS: In this retrospective study, 1669 computed tomography (CT) images with five specific anatomical locations were collected from Asan Medical Center and Kangdong Sacred Heart Hospital. To generate the ground truth, two radiologists reviewed the anatomical locations and presence of contrast enhancement using the collected data. The individual deep learning model is developed through InceptionResNetv2 and transfer learning, and we propose Image Quality Check-Net (Image QC-Net), an ensemble AI model that utilizes it. To evaluate their clinical effectiveness, the overall accuracy and time spent on image quality check of a conventional model and ImageQC-net were compared. RESULTS: ImageQC-net body part classification showed excellent performance in both internal (precision, 100%; recall, 100% accuracy, 100%) and external verification sets (precision, 99.8%; recovery rate, 99.8%, accuracy, 99.8%). In addition, contrast enhancement classification performance achieved 100% precision, recall, and accuracy in the internal verification set and achieved (precision, 100%; recall, 100%; accuracy 100%) in the external dataset. In the case of clinical effects, the reduction of time by checking the quality of artificial intelligence (AI) support by analysts 1 and 2 (49.7% and 48.3%, respectively) was statistically significant (p < 0.001). CONCLUSIONS: Comprehensive AI techniques to identify body parts and contrast enhancement on CT images are highly accurate and can significantly reduce the time spent on image quality checks.

Automated segmentation of the fractured vertebrae on CT and its applicability in a radiomics model to predict fracture malignancy.

Although CT radiomics has shown promising results in the evaluation of vertebral fractures, the need for manual segmentation of fractured vertebrae limited the routine clinical implementation of radiomics. Therefore, automated segmentation of fractured vertebrae is needed for successful clinical use of radiomics. In this study, we aimed to develop and validate an automated algorithm for segmentation of fractured vertebral bodies on CT, and to evaluate the applicability of the algorithm in a radiomics prediction model to differentiate benign and malignant fractures. A convolutional neural network was trained to perform automated segmentation of fractured vertebral bodies using 341 vertebrae with benign or malignant fractures from 158 patients, and was validated on independent test sets (internal test, 86 vertebrae [59 patients]; external test, 102 vertebrae [59 patients]). Then, a radiomics model predicting fracture malignancy on CT was constructed, and the prediction performance was compared between automated and human expert segmentations. The algorithm achieved good agreement with human expert segmentation at testing (Dice similarity coefficient, 0.93-0.94; cross-sectional area error, 2.66-2.97%; average surface distance, 0.40-0.54 mm). The radiomics model demonstrated good performance in the training set (AUC, 0.93). In the test sets, automated and human expert segmentations showed comparable prediction performances (AUC, internal test, 0.80 vs 0.87, p = 0.044; external test, 0.83 vs 0.80, p = 0.37). In summary, we developed and validated an automated segmentation algorithm that showed comparable performance to human expert segmentation in a CT radiomics model to predict fracture malignancy, which may enable more practical clinical utilization of radiomics.

Prediction of the Acuity of Vertebral Compression Fractures on CT Using Radiologic and Radiomic Features.

RATIONALE AND OBJECTIVES: To develop and validate prediction models to differentiate acute and chronic vertebral compression fractures based on radiologic and radiomic features on CT. MATERIALS AND METHODS: This study included acute and chronic compression fractures in patients who underwent both spine CT and MRI examinations. For each fractured vertebra, three CT findings ([1] cortical disruption, [2] hypoattenuating cleft or sclerotic line, and [3] relative bone marrow attenuation) were assessed by two radiologists. A radiomic score was built from 280 radiomic features extracted from non-contrast-enhanced CT images. Weighted multivariable logistic regression analysis was performed to build a radiologic model based on CT findings and an integrated model combining the radiomic score and CT findings. Model performance was evaluated and compared. Models were externally validated using an independent test cohort. RESULTS: A total to 238 fractures (159 acute and 79 chronic) in 122 patients and 58 fractures (39 acute and 19 chronic) in 32 patients were included in the training and test cohorts, respectively. The AUC of the radiomic score was 0.95 in the training and 0.93 in the test cohorts. The AUC of the radiologic model was 0.89 in the training and 0.83 in the test cohorts. The discriminatory performance of the integrated model was significantly higher than the radiologic model in both the training (AUC, 0.97; p<0.01) and the test (AUC, 0.95; p=0.01) cohorts. CONCLUSION: Combining radiomics with radiologic findings significantly improved the performance of CT in determining the acuity of vertebral compression fractures.

Development of a fully automatic deep learning system for L3 selection and body composition assessment on computed tomography.

As sarcopenia research has been gaining emphasis, the need for quantification of abdominal muscle on computed tomography (CT) is increasing. Thus, a fully automated system to select L3 slice and segment muscle in an end-to-end manner is demanded. We aimed to develop a deep learning model (DLM) to select the L3 slice with consideration of anatomic variations and to segment cross-sectional areas (CSAs) of abdominal muscle and fat. Our DLM, named L3SEG-net, was composed of a YOLOv3-based algorithm for selecting the L3 slice and a fully convolutional network (FCN)-based algorithm for segmentation. The YOLOv3-based algorithm was developed via supervised learning using a training dataset (n = 922), and the FCN-based algorithm was transferred from prior work. Our L3SEG-net was validated with internal (n = 496) and external validation (n = 586) datasets. Ground truth L3 level CT slice and anatomic variation were identified by a board-certified radiologist. L3 slice selection accuracy was evaluated by the distance difference between ground truths and DLM-derived results. Technical success for L3 slice selection was defined when the distance difference was < 10 mm. Overall segmentation accuracy was evaluated by CSA error and DSC value. The influence of anatomic variations on DLM performance was evaluated. In the internal and external validation datasets, the accuracy of automatic L3 slice selection was high, with mean distance differences of 3.7 ± 8.4 mm and 4.1 ± 8.3 mm, respectively, and with technical success rates of 93.1% and 92.3%, respectively. However, in the subgroup analysis of anatomic variations, the L3 slice selection accuracy decreased, with distance differences of 12.4 ± 15.4 mm and 12.1 ± 14.6 mm, respectively, and with technical success rates of 67.2% and 67.9%, respectively. The overall segmentation accuracy of abdominal muscle areas was excellent regardless of anatomic variation, with CSA errors of 1.38-3.10 cm2. A fully automatic system was developed for the selection of an exact axial CT slice at the L3 vertebral level and the segmentation of abdominal muscle areas.

Combined radiomics-clinical model to predict malignancy of vertebral compression fractures on CT.

OBJECTIVES: To develop and validate a combined radiomics-clinical model to predict malignancy of vertebral compression fractures on CT. METHODS: One hundred sixty-five patients with vertebral compression fractures were allocated to training (n = 110 [62 acute benign and 48 malignant fractures]) and validation (n = 55 [30 acute benign and 25 malignant fractures]) cohorts. Radiomics features (n = 144) were extracted from non-contrast-enhanced CT images. Radiomics score was constructed by applying least absolute shrinkage and selection operator regression to reproducible features. A combined radiomics-clinical model was constructed by integrating significant clinical parameters with radiomics score using multivariate logistic regression analysis. Model performance was quantified in terms of discrimination and calibration. The model was internally validated on the independent data set. RESULTS: The combined radiomics-clinical model, composed of two significant clinical predictors (age and history of malignancy) and the radiomics score, showed good calibration (Hosmer-Lemeshow test, p > 0.05) and discrimination in both training (AUC, 0.970) and validation (AUC, 0.948) cohorts. Discrimination performance of the combined model was higher than that of either the radiomics score (AUC, 0.941 in training cohort and 0.852 in validation cohort) or the clinical predictor model (AUC, 0.924 in training cohort and 0.849 in validation cohort). The model stratified patients into groups with low and high risk of malignant fracture with an accuracy of 98.2% in the training cohort and 90.9% in the validation cohort. CONCLUSIONS: The combined radiomics-clinical model integrating clinical parameters with radiomics score could predict malignancy in vertebral compression fractures on CT with high discriminatory ability. KEY POINTS: • A combined radiomics-clinical model was constructed to predict malignancy of vertebral compression fractures on CT by combining clinical parameters and radiomics features. • The model showed good calibration and discrimination in both training and validation cohorts. • The model showed high accuracy in the stratification of patients into groups with low and high risk of malignant vertebral compression fractures.

Test-retest repeatability of ultrasonographic shear wave elastography in a rat liver fibrosis model: toward a quantitative biomarker for preclinical trials.

PURPOSE: This study evaluated the test-retest repeatability and measurement variability of ultrasonographic shear wave elastography (SWE) for liver stiffness in a rat liver fibrosis model. METHODS: In 31 Sprague-Dawley rats divided into three groups (high-dose, low-dose, and control), liver fibrosis was induced by intraperitoneal administration of thioacetamide for 8 weeks. A dedicated radiographer performed SWE to measure liver stiffness in kilopascals in two sessions at a 3-day interval. We calculated correlations between liver stiffness and histopathologic results, measurement variability in each session using coefficients of variation (CoVs) and interquartile/median (IQR/M), and test-retest repeatability between both sessions using the repeatability coefficient. RESULTS: Different levels of liver fibrosis in each group were successfully induced in the animal model. The mean liver stiffness values were 8.88±1.48 kPa in the control group, 11.62±1.70 kPa in the low-dose group, and 11.91±1.73 kPa in the high-dose group. The correlation between collagen areas and liver stiffness values was moderate (r=0.6). In all groups, the second session yielded lower CoVs (i.e., more reliable results) for liver stiffness than the first session, suggesting a training effect for the operator. The mean IQR/M values were also lower in the second session than in the first session, which had four outliers (0.21 vs. 0.12, P<0.001). The test-retest repeatability coefficient was 3.75 kPa and decreased to 2.82 kPa after removing the four outliers. CONCLUSION: The use of ultrasonographic SWE was confirmed to be feasible and repeatable for evaluating liver fibrosis in preclinical trials. Operator training might reduce variability in liver stiffness measurements.

Tract ablation after radiofrequency ablation to prevent viable tumor cell adhesion to the needle electrode.

PURPOSE: To evaluate whether the additive needle tract ablation (TA) can reduce adherent cells on the needle tract after radiofrequency ablation (RFA) in a preclinical HCC mouse model. METHODS: Hep3B-Luc cells were engrafted in the Balb/c-nude mice. Nineteen mice were randomly assigned into three groups: the needle only group (needle placement only without performing RFA), the RFA only group (needle placement with active RFA treatment), and the RFA-TA group (needle placement with active RFA treatment and additive tract ablation). The 17-gauge needle with a 10-mm active tip was used. After RFA and TA, the viability of adherent tumor cells on the RFA needle was evaluated with bioluminescence imaging (BLI) and live-cell counting. RESULTS: We observed that RFA-TA group had the lowest BLI values compared with other groups (needle only group, 11.2 ± 6.4 million; RFA only group, 13.6 ± 9.1 million; RFA-TA group, 1.11 ± 0.8 million, p = 0.001). Live cell counting with acridine orange/propidium iodide staining also confirmed that the counted viable cell numbers in RFA-TA group were lowest compared to the other groups (needle only group, 14.8 ± 4.5; RFA only group, 643.8 ± 131.9; RFA-TA group, 1.5 ± 0.9, p < 0.001). CONCLUSIONS: The additive tract ablation can significantly reduce the number of viable tumor cells adherent to the RFA needle, which can prevent needle tract seeding after RFA procedure.

Gadoxetate-enhanced dynamic contrast-enhanced MRI for evaluation of liver function and liver fibrosis in preclinical trials.

BACKGROUND: To facilitate translational drug development for liver fibrosis, preclinical trials need to be run in parallel with clinical research. Liver function estimation by gadoxetate-enhanced dynamic contrast-enhanced MRI (DCE-MRI) is being established in clinical research, but still rarely used in preclinical trials. We aimed to evaluate feasibility of DCE-MRI indices as translatable biomarkers in a liver fibrosis animal model. METHODS: Liver fibrosis was induced in Sprague-Dawley rats by thioacetamide (200 mg, 150 mg, and saline for the high-dose, low-dose, and control groups, respectively). Subsequently, DCE-MRI was performed to measure: relative liver enhancement at 3-min (RLE-3), RLE-15, initial area-under-the-curve until 3-min (iAUC-3), iAUC-15, and maximum-enhancement (Emax). The correlation coefficients between these MRI indices and the histologic collagen area, indocyanine green retention at 15-min (ICG-R15), and shear wave elastography (SWE) were calculated. Diagnostic performance to diagnose liver fibrosis was also evaluated by receiver-operating-characteristic (ROC) analysis. RESULTS: Animal model was successful in that the collagen area of the liver was the largest in the high-dose group, followed by the low-dose group and control group. The correlation between the DCE-MRI indices and collagen area was high for iAUC-15, Emax, iAUC-3, and RLE-3 but moderate for RLE-15 (r, - 0.81, - 0.81, - 0.78, - 0.80, and - 0.51, respectively). The DCE-MRI indices showed moderate correlation with ICG-R15: the highest for iAUC-15, followed by iAUC-3, RLE-3, Emax, and RLE-15 (r, - 0.65, - 0.63, - 0.62, - 0.58, and - 0.56, respectively). The correlation coefficients between DCE-MRI indices and SWE ranged from - 0.59 to - 0.28. The diagnostic accuracy of RLE-3, iAUC-3, iAUC-15, and Emax was 100% (AUROC 1.000), whereas those of RLE-15 and SWE were relatively low (AUROC 0.777, 0.848, respectively). CONCLUSION: Among the gadoxetate-enhanced DCE-MRI indices, iAUC-15 and iAUC-3 might be bidirectional translatable biomarkers between preclinical and clinical research for evaluating histopathologic liver fibrosis and physiologic liver functions in a non-invasive manner.

Cerebral mapping of glutamate using chemical exchange saturation transfer imaging in a rat model of stress-induced sleep disturbance at 7.0T.

BACKGROUND: Glutamate chemical exchange saturation transfer (GluCEST) imaging has been widely used in brain psychiatric disorders. Glutamate signal changes may help to evaluate the sleep-related disorders, and could be useful in diagnosis. PURPOSE: To evaluate signal changes in the hippocampus and cortex of a rat model of stress-induced sleep disturbance using GluCEST. STUDY TYPE: Prospective animal study. ANIMAL MODEL: Fourteen male Sprague-Dawley rats. FIELD STRENGTH/SEQUENCE: 7.0T small bore MRI / fat-suppressed, turbo-rapid acquisition with relaxation enhancement (RARE) for CEST, and spin-echo, point-resolved proton MR spectroscopy (1 H MRS). ASSESSMENT: Rats were divided into two groups: the stress-induced sleep-disturbance group (SSD, n = 7) and the control group (CTRL, n = 7), to evaluate and compare the cerebral glutamate signal changes. GluCEST data were quantified using a conventional magnetization transfer ratio asymmetry in the left- and right-side hippocampus and cortex. The correlation between GluCEST signal and glutamate concentrations, derived from 1 H MRS, was evaluated. STATISTICAL ANALYSIS: Wilcoxon rank-sum test between CEST signals and multiparametric MR signals, Wilcoxon signed-rank test between CEST signals on the left and right hemispheres, and a correlation test between CEST signals and glutamate concentrations derived from 1 H MRS. RESULTS: Measured GluCEST signals showed significant differences between the two groups (left hippocampus; 4.23 ± 0.27% / 5.27 ± 0.42% [SSD / CTRL, P = 0.002], right hippocampus; 4.50 ± 0.44% / 5.04 ± 0.34% [P = 0.035], left cortex; 2.81 ± 0.38% / 3.56 ± 0.41% [P = 0.004], and right cortex; 2.95 ± 0.47% / 3.82 ± 0.26% [P = 0.003]). GluCEST signals showed positive correlation with glutamate concentrations (R2 = 0.312; P = 0.038). DATA CONCLUSION: GluCEST allowed the visualization of cerebral glutamate changes in rats subjected to sleep disturbance, and may yield valuable insights for interpreting alterations in cerebral biochemical information. LEVEL OF EVIDENCE: 2 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2019;50:1866-1872.

Evaluation of drug mechanism and efficacy of a novel anti-angiogenic agent, TTAC-0001, using multi-modality bioimaging in a mouse breast cancer orthotopic model.

PURPOSE: Targeting of vascular endothelial growth factor receptors (VEGFRs) has potential anti-angiogenic effects because VEGFR-2 is the major signaling regulator of VEGF/VEGFR pathways. We aimed to elucidate the drug mechanism and anti-tumor efficacy of TTAC-0001, a novel, fully human anti-VEGFR-2/KDR monoclonal antibody, in mouse orthotopic breast cancer model using multi-modal bioimaging. MATERIALS AND METHODS: We used orthotopic xenograft tumor model in which human breast cancer cells (MDA-MB-231) were injected into the right mammary fat pad of Balb/c nude mice. We investigated its biodistribution using serial fluorescence imaging after injecting fluorescent-labelled-drug and mode of action using Matrigel plug angiogenesis assays. The anti-tumor efficacy of drug was assessed using ultrasonography and bioluminescence imaging. Histopathologic analyses, including hematoxylin and eosin staining and immunohistochemistry with anti-CD31 and anti-Ki-67 antibodies, were performed. Each experiment had four groups: control, bevacizumab 10 mg/kg (BVZ-10 group), TTAC-0001 2 mg/kg (TTAC-2 group), and TTAC-0001 10 mg/kg (TTAC-10 group). RESULTS: The TTAC-10 group showed good tumor targeting that lasted for at least 6 days and had a good anti-angiogenic effect with decreased hemoglobin content and fewer CD31-positive cells in the Matrigel plug. Compared with BVZ-10 and TTAC-2 groups, the TTAC-10 group showed the strongest anti-tumor efficacy, inhibiting tumor growth as detected by ultrasonography and bioluminescence imaging. The TTAC-10 group also showed the lowest viable tumor and micro-vessel areas and the lowest Ki-67 index in histopathologic analyses. CONCLUSION: We firstly demonstrated that TTAC-0001 effectively inhibited tumor growth and neovascularization in mouse orthotopic breast cancer model. It may provide a future treatment option for breast cancer.

Enhanced efficacy of radiofrequency ablation for hepatocellular carcinoma using a novel vascular disrupting agent, CKD-516.

BACKGROUND: CKD-516 is a novel vascular disrupting agent that shuts down intratumoral blood flow. We therefore hypothesized that concomitant administration of CKD-516 would enhance the therapeutic efficacy of radiofrequency ablation (RFA) by reducing heat sink effects. We assessed the effects of the combination of CKD-516 and RFA in a rat orthotopic hepatocellular carcinoma (HCC) model. METHODS: Rat HCC cells (N1-S1) were engrafted into the hepatic lobe of Sprague-Dawley (SD) rats. Mice were randomly divided into two groups: RFA-only and CKD-RFA. In the CKD-RFA group, CKD-516 was administered by intraperitoneal injection 2 h before RFA. Ablation zone size was measured on triphenyltetrazolium chloride-stained specimens. Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining was performed to evaluate the area of apoptosis/necrosis in the ablation zone. Immunohistochemistry with anti-CD31 antibody was performed to evaluate the effect of CKD-516 on tumor vessels. RESULTS: Ablation zone size was significantly larger in the CKD-RFA group than in the RFA-only group (243.10 ± 74.39 versus 123.30 ± 28.17 mm2, p < 0.001). On TUNEL staining, the area of apoptosis/necrosis was also significantly larger in the CKD-RFA group than in the RFA-only group (274.44 ± 140.78 versus 143.74 ± 90.13 mm2; p = 0.006). Immunohistochemistry with anti-CD31 antibody revealed patent tumor vessels in the RFA-only group, while collapsed vessels were seen in the CKD-RFA group, indicating a vascular shutdown effect of CKD-516. CONCLUSION: Concomitant administration of CKD-516 during RFA can increase the ablation zone of tumors due to its vascular disrupting effect.

Feasibility of a Low-Power Radiofrequency Ablation Protocol to Delay Steam Popping.

PURPOSE: Steam popping frequently occurs during conventional high-power radiofrequency (RF) ablation, increasing the risk of tumor spread. The aim of this study was to evaluate the effect of a low-power RF ablation protocol on the intensity and timing of steam popping in ex vivo bovine liver. MATERIALS AND METHODS: High-power (maximum 200 W; group 1) and low-power (maximum 70 W; group 2) RF ablation protocols were established. In the first phase, RF ablation was conducted for 12 min. Ablation volume, intensity, and timing of maximal popping sounds and total energy generated for RF ablation were compared between groups 1 and 2. In the second phase, RF ablation was conducted until maximal popping occurred, and ablation zones on histologic specimens were compared. RESULTS: Relative to group 1, maximal popping occurred at significantly delayed timing in group 2 (50 s ± 11 vs 397 s ± 117; P < .001), but without a difference in intensity (ratios vs reference sound of 0.70 ± 0.18 vs 0.83 ± 0.26; P = .138). The ablation volume after 12 min of RF ablation did not differ between groups 1 and 2 (18.46 cm(3) ± 1.35 vs 15.78 cm(3) ± 0.64; P = .086). However, in the histologic specimens obtained when maximal popping occurred, the area of complete coagulative necrosis was significantly larger in group 2 (P < .05). CONCLUSIONS: Low-power RF ablation delays steam popping while providing comparable therapeutic effects to high-power RF ablation. Delaying maximal popping may prevent tumor cell dispersion because maximal popping occurs after an adequate ablation zone has been achieved.