Reduction of ADC bias in diffusion MRI with deep learning-based acceleration: A phantom validation study at 3.0 T

PURPOSE

Further acceleration of DWI in diagnostic radiology is desired but challenging mainly due to low SNR in high b-value images and associated bias in quantitative ADC values. Deep learning-based reconstruction and denoising may provide a solution to address this challenge.

METHODS

The effects of SNR reduction on ADC bias and variability were investigated using a commercial diffusion phantom and numerical simulations. In the phantom, performance of different reconstruction methods, including conventional parallel (SENSE) imaging, compressed sensing (C-SENSE), and compressed SENSE acceleration with an artificial intelligence deep learning-based technique (C-SENSE AI), was compared at different acceleration factors and flip angles using ROI-based analysis. ADC bias was assessed by Lin's Concordance correlation coefficient (CCC) followed by bootstrapping to calculate confidence intervals (CI). ADC random measurement error (RME) was assessed by the mean coefficient of variation (CV¯) and non-parametric statistical tests.

RESULTS

The simulations predicted increasingly negative bias and loss of precision towards lower SNR. These effects were confirmed in phantom measurements of increasing acceleration, for which CCC decreased from 0.947 to 0.279 and CV¯ increased from 0.043 to 0.439, and of decreasing flip angle, for which CCC decreased from 0.990 to 0.063 and CV¯ increased from 0.037 to 0.508. At high acceleration and low flip angle, C-SENSE AI reconstruction yielded best denoised ADC maps. For the lowest investigated flip angle, CCC = {0.630, 0.771 and 0.987} and CV¯={0.508, 0.426 and 0.254} were obtained for {SENSE, C-SENSE, C-SENSE AI}, the improvement by C-SENSE AI being significant as compared to the other methods (CV: p = 0.033 for C-SENSE AI vs. C-SENSE and p < 0.001 for C-SENSE AI vs. SENSE; CCC: non-overlapping CI between reconstruction methods). For the highest investigated acceleration factor, CCC = {0.479,0.926,0.960} and CV¯={0.519,0.119,0.118} were found, confirming the reduction of bias and RME by C-SENSE AI as compared to C-SENSE (by trend) and to SENSE (CV: p < 0.001; CCC: non-overlapping CI).

CONCLUSION

ADC bias and random measurement error in DWI at low SNR, typically associated with scan acceleration, can be effectively reduced by deep-learning based C-SENSE AI reconstruction.

Managing hepatocellular carcinoma across the stages: efficacy and outcomes of stereotactic body radiotherapy : A retrospective study

PURPOSE

Hepatocellular carcinoma (HCC) poses a unique challenge due to its predilection for developing on compromised livers, often limiting surgical options. Stereotactic body radiotherapy (SBRT) has emerged as a promising local treatment modality for HCC. This study aims to assess the effectiveness of SBRT in HCC patients not suitable for surgery, focusing on local control, optimal radiation dosing, and prognostic factors.

METHODS

In this retrospective analysis, 52 HCC patients treated with SBRT were examined. The study assessed local control, progression-free survival (PFS), and overall survival (OS) while conducting dosimetric analyses. The relationship between mean liver dose and Child-Pugh score (CPS) progression was also explored.

RESULTS

SBRT demonstrated 93.4% freedom from local progression (FFLP) at 12 months. Notably, a near minimum dose (D98%) below 61 Gy as an equivalent dose in 2‑Gy fractions with α/β 10 Gy (EQD2α/β10) was associated with reduced FFLP (p-value 0.034). Logistic regression analysis revealed a dose-response relationship for FFLP and D98% with 95% and 98% probability of FFLP at a dose of 56.9 and 73.1 Gy, respectively. The study observed OS rates of 63.7% at 1 year and 34.3% at 3 years. Patients with portal vein tumor thrombus (PVTT) and larger tumors (≥ 37 cm3) experienced decreased PFS and OS. Multivariate analysis identified PVTT, larger tumor volume, and performance status as independent predictors of reduced OS. Notably, classical radiation-induced disease (cRILD) was absent, but nonclassical (nc) RILD occurred in 7.7% of patients. Regression analysis linked a mean EQD2α/β3-8 dose to the liver (12.8-12.6) with a 10% likelihood of ncRILD.

CONCLUSION

SBRT offers a compelling option for achieving high local control and promising survival outcomes in HCC. The study supports a radiation dose range of 61-73.1 Gy, coupled with a mean liver dose under 12.6-12.8 Gy as EQD2, to achieve favorable FFLP rates, with acceptable toxicity rates.

Liver diffusion-weighted MR imaging with L1-regularized iterative sensitivity encoding reconstruction based on single-shot echo-planar imaging: initial clinical experience

To investigate whether combining L1-regularized iterative sensitivity encoding (SENSE) reconstruction and single-shot echo planar imaging (EPI) is useful in hepatic DWI. Single-shot EPI-DWI with L1-regularized iterative SENSE reconstruction (L1-DWI) and conventional parallel imaging-based reconstruction (conv-DWI) in liver MRI were compared in volunteers and patients. For the patient cohort, 75 subjects (60 ± 13 years) with 349 focal liver lesions (FLL) were included. Patient groups A and B were used to reduce acquisition time or improve spatial resolution, respectively. Image parameters were rated on a 5-point scale. The number of FLLs was recorded; in case of discrepancy, the reason for non-detectability was analyzed. In volunteers, higher signal-to-noise ratio (24.4 ± 5.6 vs. 12.2 ± 2.3, p < 0.001 at b = 0; 19.3 ± 2.8 vs. 9.8 ± 1.6, p < 0.001 at b = 800) and lower standard deviation of the apparent diffusion coefficient-values (0.17 vs. 0.20 mm²/s, p < 0.05) were found on L1-DWI compared to conv-DWI. In patients, image ratings were similar for all parameters except for "conspicuity of FLLs" which was rated significantly lower on L1-DWI vs. conv-DWI (4.7 ± 0.6 vs. 4.2 ± 0.9, p < 0.05) in group A. In five patients, 11/349 FLLs were not detectable on L1-DWI, but on conv-DWI. L1-regularized iterative reconstruction of single-shot EPI DWI can accelerate image acquisition or improve spatial resolution. However, our finding that FLLs were non-detectable on L1-DWI warrants further research.

Irreversible Electroporation for Hepatic Tumors: Protocol Standardization Using the Modified Delphi Technique

PURPOSE

A consensus study of panelists was performed to provide a uniform protocol regarding (contra) indications, procedural parameters, perioperative care, and follow-up of irreversible electroporation (IRE) for the treatment of hepatic malignancies.

MATERIALS AND METHODS

Interventional radiologists who had 2 or more publications on IRE, reporting at least 1 patient cohort in the field of hepatobiliary IRE, were recruited. The 8 panelists were asked to anonymously complete 3 iterative rounds of IRE-focused questionnaires to collect data according to a modified Delphi technique. Consensus was defined as having reached 80% or greater agreement.

RESULTS

Panel members' response rates were 88%, 75%, and 88% in rounds 1, 2, and 3, respectively; consensus was reached on 124 of 136 items (91%). Percutaneous or intraoperative hepatic IRE should be considered for unresectable primary and secondary malignancies that are truly unsuitable for thermal ablation because of proximity to critical structures. Absolute contraindications are ventricular arrhythmias, cardiac stimulation devices, and congestive heart failure of New York Heart Association class 3 or higher. A metal stent outside the ablation zone should not be considered a contraindication. For the only commercially available IRE device, the recommended settings are an inter-electrode distance of 10-20 mm and an exposure length of 20 mm. After 10 test pulses, 90 treatment pulses of 1500 V/cm should be delivered continuously, with a pulse length of 70-90 μs. The first post-procedural follow-up should take place 1 month after IRE and thereafter every 3 months, using cross-sectional imaging plus tumor marker assessment.

CONCLUSIONS

This article provides recommendations, created by a modified Delphi consensus study, regarding patient selection, workup, procedure, and follow-up of IRE treatment for hepatic malignancies.

Signal changes in T2-weighted MRI of liver metastases under bevacizumab-A practical imaging biomarker?

OBJECTIVE

The purpose of this study was to investigate signal changes in T2-weighted magnetic resonance imaging of liver metastases under treatment with and without bevacizumab-containing chemotherapy and to compare these signal changes to tumor contrast enhancement.

MATERIALS AND METHODS

Retrospective analysis of 44 patients, aged 36-84 years, who underwent liver magnetic resonance imaging including T2-weighted and dynamic contrast enhancement sequences. Patients received bevacizumab-containing (n = 22) or conventional cytotoxic chemotherapy (n = 22). Magnetic resonance imaging was obtained at baseline and at three follow-ups (on average 3, 6 and 9 months after initial treatment). Three independent readers rated the T2 signal intensity and the relative contrast enhancement of the metastases on a 5-point scale.

RESULTS

T2 signal intensity of metastases treated with bevacizumab showed a significant (p<0.001) decrease in T2 signal intensity after initial treatment and exhibit compared to conventionally treated metastases significantly (p<0.001 for each follow-up) hypointense (bevacizumab: 0.70 ± 0.83 before vs. -1.55 ± 0.61, -1.91 ± 0.62, and -1.97 ± 0.52; cytotoxic: 0.73 ± 0.79 before vs. -0.69 ± 0.81, -0.71 ± 0.68, and -0.75 ± 0.65 after 3, 6, and 9 months, respectively). T2 signal intensity was strongly correlated with tumor contrast enhancement (r = 0.71; p<0.001). Intra-observer agreement for T2-signal intensity was substantial (κ = 0.75). The agreement for tumoral contrast enhancement between the readers was considerably lower (κ = 0.39).

CONCLUSION

Liver metastases exhibit considerably hypointense in T2-weighted imaging after treatment with bevacizumab, in contrast to conventionally treated liver metastases. Therefore, T2-weighted imaging seems to reflect the effect of bevacizumab.

Toward Transpulmonary Chemoembolization with Degradable Starch Microspheres: Systematic Analysis of Local and Systemic Effects in a Porcine Model

PURPOSE

To investigate local and systemic effects of transpulmonary chemoembolization (TPCE) with degradable starch microspheres (DSM) and doxorubicin. The long-term goal is to establish DSM-TPCE as a treatment option for pulmonary malignancies.

MATERIALS AND METHODS

Nine pigs underwent TPCE of either the right or left lower lobe pulmonary artery (LLPA) and bland embolization (TPE) of the contralateral LLPA. Before the procedures, macroaggregated albumin (MAA) particles were injected into both LLPAs, to exclude systemic shunting. Pulmonary arterial pressure, heart rate and oxygenation were recorded immediately before and at 1, 3, 5 and 10 min after treatment. To investigate possible nontarget embolization, animals underwent cerebral MRI (cMRI). We killed the animals after a contrast-enhanced chest computed tomography (CT) and performed a pathologic examination at 12 h (3), 24 h (3) and 72 h (3) after treatment.

RESULTS

All experiments were technically successful. Mean injected DSM dose until stasis was similar in TPCE and TPE (4.3 ± 1.4 vs. 4.0 ± 1.4 mL). Pulmonary arterial pressure increased significantly 3 min after treatment (TPE: 17 ± 5 vs. 27 ± 7 mmHg; TPCE: 22 ± 6 vs. 36 ± 8 mmHg). No significant changes in heart rate or peripheral oxygenation levels occurred. We observed no evidence of structural lung damage or permanent perfusion disruption on CT. MAA test injection and cMRI revealed no shunting or nontarget embolization. The pathologic assessment revealed nonspecific local inflammation of the lung parenchyma.

CONCLUSION

In this large-animal model, TPCE and TPE appear feasible and safe. We observed a mild increase in pulmonary arterial pressure. Nontarget embolization did not occur. TPCE, as well as TPE, did not cause structural damage to the normal lung parenchyma.

Electroporation of the Liver: More Than 2 Concurrently Active, Curved Electrodes Allow New Concepts for Irreversible Electroporation and Electrochemotherapy

Irreversible electroporation and electrochemotherapy are 2 innovative electroporation-based minimally invasive therapies for the treatment of cancer. Combining nonthermal effects of irreversible electroporation with local application of chemotherapy, electrochemotherapy is an established treatment modality for skin malignancies. Since the application of electrochemotherapy in solid organs is a promising approach, this article describes a novel electrode configuration and field generating method. For the treatment of hepatic malignancies, the shape of the electric field should resemble a spherical 3-dimensional geometry around the target tissue inside the liver. To adapt the actual shape of the field, the probe is designed in computer-aided design with a live link to a computer simulation software: Changes in design can be revalued quickly, regarding different quality criteria for field strength inside and outside the tumor. To rate these criteria, a set of formulas with weighting coefficients has been included. As a result of this design process, a needle-shaped prototype applicator has been built, designed for an intracorporal electroporation-based treatment. It can be used as percutaneous, image-guided, minimally invasive treatment option for malignant liver tumors. The shaft of the probe is used as central electrode and fitted with additional 4 expandable electrodes. These satellite electrodes are hollow, thus serving as injectors for chemotherapeutic agents within the area of the electric field. This configuration can be used for electrochemotherapy as well as irreversible electroporation. By placing 5 electrodes with just one needle, the procedure duration as well as the radiation dose can be reduced tremendously. Additionally, the probe offers an option to adapt the field geometry to the tumor geometry by connecting the 5 electrodes to 5 individually chosen electric potentials: By fine-tuning the ablation zone via the potentials instead of adjusting the location of the electrode(s), the procedure duration as well as the radiation dose will decrease further.

Diffusion-weighted MRI Is Superior to PET/CT in Predicting Survival of Patients Undergoing 90Y Radioembolization of Hepatic Metastases

Purpose To determine the relationship between diffusion-weighted (DW) liver MR images obtained 4-6 weeks after lobar yttrium 90 (⁹⁰Y) treatment and overall survival in comparison with PET/CT or established oncologic factors known to affect survival. Materials and Methods The institutional review board approved this prospective intraindividual comparative study in 36 consecutive patients (25 women) with liver-dominant metastases (20 colorectal, 14 breast, two other) (mean age, 60 years ± 10 [standard deviation]) who underwent fluorine 18 (¹⁸F) fluorodeoxyglucose (FDG) PET/CT and DW MRI before and 4-6 weeks after ⁹⁰Y radioembolization. DW MRI response was defined as a mean minimal apparent diffusion coefficient increase of more than 30%; PET/CT response was defined as a mean maximal standardized uptake value decrease of more than 30%. Kaplan-Meier curves, log-rank test, and multivariable Cox regression analyses were used to compare patient survival as a function of imaging and Response Evaluation Criteria in Solid Tumors (RECIST) response, pretreatment Eastern Cooperative Oncology Group (ECOG) performance status (PS) (0 vs 1), hepatic tumor load (<25% vs ≥25%), and presence versus absence of extrahepatic disease. Results Thirty-five of the 36 patients were observed until death (median survival, 36 weeks). Response was observed with PET/CT in 18 of 36 patients (50%). Median survival was 39 weeks in patients who responded to PET/CT versus 27 weeks in those who did not (P = .60). Response was observed with DW MRI in 24 of 36 patients (67%). Median survival was 53 weeks in DW MRI responders versus 20 weeks in nonresponders (P = .01). At multivariable analysis, DW MRI response was the only independent predictor of survival (P < .01). Response based on RECIST parameters, ECOG PS, hepatic tumor load, and presence of extrahepatic metastases did not correlate with survival. Conclusion In patients with hepatic metastases undergoing ⁹⁰Y radioembolization, prediction of response to therapy with DW MRI was superior to that with PET/CT and established oncologic factors.

Midterm Safety and Efficacy of Irreversible Electroporation of Malignant Liver Tumors Located Close to Major Portal or Hepatic Veins

Purpose To investigate the efficacy and safety of irreversible electroporation (IRE) in the treatment of hepatic tumors not suitable for thermal ablation (radiofrequency ablation [RFA] or microwave ablation). Materials and Methods This was an institutional review board-approved prospective study in 29 patients (15 men, 14 women; mean age, 63 years ± 12 [standard deviation]) with 43 primary (n = 8) or secondary (n = 35) malignant liver tumors who underwent computed tomography (CT)-guided IRE. All target tumors were located immediately adjacent to major hepatic veins, portal veins, or both; thus, they were not considered suitable for RFA or microwave ablation. Patients underwent postinterventional CT and magnetic resonance (MR) imaging. Systematic follow-up MR imaging was performed for 24 months on average to assess complete ablation, intrahepatic tumor recurrence, and complications. The 95% confidence intervals (CIs) were determined for the rate of bile duct strictures, incomplete ablation, and tumor recurrence. Results Complete ablation was achieved in 40 (93%; 95% CI: 85, 100) of 43 target tumors, with a safety margin of 5-10 mm, and was confirmed at immediate postinterventional CT and MR imaging. In 13 (33%; 95% CI: 18, 47) of 40 completely ablated tumors, intrahepatic tumor recurrence was observed at 2-18 months. However, only two (15%; 95% CI: 0, 35) of these 13 tumors were observed within the ablation zone. In the remaining 11 (85%; 95% CI: 65, 100), tumor growth was observed alongside the needle tract. None of the two true local recurrences occurred at the site of the vessel. All adjacent vessels remained perfused at follow-up. Five (24%; 95% CI: 5, 39) of 21 patients with target tumors adjacent to portal veins developed mild to moderate cholestasis 2-6 weeks after IRE. Conclusion IRE is useful to avoid incomplete ablation secondary to heat-sink effects and damage to major blood vessels; however, needle tract seeding is observed in 26% of treated tumors, and IRE induces sufficient local heating to bile ducts in 24% of ablations. ^© RSNA, 2017.

Target lesion selection as a source of variability of response classification by RECIST 1.1

11077

Background: Response classification in RECIST is based on manual uni-dimensional quantification of changes of target lesion size. In RECIST 1.1, the maximum number of target lesions was reduced to 5, with a maximum of 2 per organ. We analyzed the importance of different factors (manual vs. automated size measurement, uni- vs. three-dimensional size assessment, and between-reader-variability of target lesion selection) on response categorization. Methods: 41 female patients (58.1±13.2 y) with metastatic breast cancer underwent contrast-enhanced thoraco-abdominal CT for initial staging and first follow-up after systemic chemotherapy. Data were independently and prospectively interpreted by three radiologists. In addition, response was evaluated by a CAD system that allowed automated uni- and three-dimensional assessment of target lesions. Results: Response classification differed between readers in 19/41 patients (46%). In 25/41 patients, readers chose the same target lesions. In 6 of these 25 patients (24%), readers disagreed with regards to response classification. In 16/41 patients, readers chose different target lesions. In 13 of these 16 patients, readers disagreed (81%) (p < 0.001). When dichotomizing response classification according to its therapeutic implication into progressive vs. non-progressive disease, readers disagreed in 11/41 patients (27%). In 9 of these 11 patients, readers had chosen different target lesions. Classification by manual vs. automated uni-dimensional measurement differed in 11/41 of patients (27%). Classification by uni-dimensional vs. volumetric measurements differed in 6/41 of patients (15%). Conclusions: Response classification by RECIST suffers from substantial between-reader variability. Major source of variability is not the manual or uni-dimensional measurement, but the variable choice of target lesions between readers. [Table: see text]