Search

Abstract In cirrhotic patients with hepatocellular carcinoma (HCC), right-sided radioembolization (RE) with Yttrium-90-loaded microspheres is an established palliative therapy and can be considered a “curative intention” treatment when aiming for sequential tumor resection. To become surgical candidate, hypertrophy of the left liver lobe to > 40% (future liver remnant, FLR) is mandatory, which can develop after RE. The amount of radiation-induced shrinkage of the right lobe and compensatory hypertrophy of the left lobe is difficult for clinicians to predict. This study aimed to utilize machine learning to predict left lobe liver hypertrophy in patients with HCC and cirrhosis scheduled for right lobe RE, with external validation. The results revealed that machine learning can accurately predict relative and absolute volume changes of the left liver lobe after right lobe RE. This prediction algorithm could help to estimate the chances of conversion from palliative RE to curative major hepatectomy following significant FLR hypertrophy.

Abstract Accurate determination of lymph-node (LN) metastases is a prerequisite for high precision radiotherapy. The primary aim is to characterise the performance of PET/CT-based machine-learning classifiers to predict LN-involvement by endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA) in stage-III NSCLC. Prediction models for LN-positivity based on [18F]FDG-PET/CT features were built using logistic regression and machine-learning models random forest (RF) and multilayer perceptron neural network (MLP) for stage-III NSCLC before radiochemotherapy. A total of 675 LN-stations were sampled in 180 patients. The logistic and RF models identified SUVmax, the short-axis LN-diameter and the echelon of the considered LN among the most important parameters for EBUS-positivity. Adjusting the sensitivity of machine-learning classifiers to that of the expert-rater of 94.5%, MLP (P = 0.0061) and RF models (P = 0.038) showed lower misclassification rates (MCR) than the standard-report, weighting false positives and false negatives equally. Increasing the sensitivity of classifiers from 94.5 to 99.3% resulted in increase of MCR from 13.3/14.5 to 29.8/34.2% for MLP/RF, respectively. PET/CT-based machine-learning classifiers can achieve a high sensitivity (94.5%) to detect EBUS-positive LNs at a low misclassification rate. As the specificity decreases rapidly above that level, a combined test of a PET/CT-based MLP/RF classifier and EBUS-TBNA is recommended for radiation target volume definition.

Abstract Biliary-tract-carcinomas (BTC), pancreatic-ductal-adenocarcinomas (PDAC) and adenoidcystic-carcinomas (AC) have in common that they are traditionally treated with large clinical-target-volumes (CTV). The aim of this study is to examine the impact of pretreatment-[68Ga]FAPI-PET/CT on target-volume-definition and posttreatment-[68Ga]FAPI-PET/CT-response-assessment for BTC-, PDAC- and AC-patients referred to radiation-therapy. All consecutive BTC-, PDAC-, and AC-patients who received pretreatment-[68Ga]FAPI-PET/CT±[18F]FDG-PET/CT were included from 01.01.2020 to 01.03.2022. MTV and SUVmax were separately generated based on [68Ga]FAPI- and [18F]FDG-PET/CT-images. A [68Ga]FAPI- and [18F]FDG-based-CTV was defined. Treatment-plans were compared. Treatment-response was reassessed by a second [68Ga]FAPI-PET/CT and [18F]FDG-PET/CT after treatment-completion. Intermodality comparison of lesion-to-background-ratios [SUVmax_lesion/SUVmean_background] for individual timepoints t 1 and t 2 revealed significant higher values for [68Ga]FAPI compared to [18F]FDG (t 1 , p = 0.008; t 2 , p = 0.005). Intermodality comparison of radiation-therapy-plans showed that [68Ga]FAPI-based planning resulted in D100% = 97.2% and V95% = 98.8% for the [18F]FDG-MTV. [18F]FDG-based-planning resulted in D100% = 35.9% and V95% = 78.1% for [68Ga]FAPI-MTV. [18F]FDG-based-planning resulted only in 2 patients in V95% > 95% for [68Ga]FAPI-MTV, and in 1 patient in D100% > 97% for [68Ga]FAPI-MTV. GTV-coverage in terms of V95% was 76.4% by [18F]FDG-based-planning and 99.5% by [68Ga]FAPI-based-planning. Pretreatment [68Ga]FAPI-PET/CT enhances radiation-treatment-planning in this particular group of patients. While perilesional and tumoral follow-up [18F]FDG-uptake behaved uniformly, perilesional and tumoral reaction may differ in follow-up [68Ga]FAPI-imaging. Complementary [68Ga]FAPI- and [18F]FDG-imaging enhance treatment-response-assessment.

Abstract The radioiodine isotope pair 124I/131I is used in a theranostic approach for patient-specific treatment of differentiated thyroid cancer. Lesion detectability is notably higher for 124I PET (positron emission tomography) than for 131I gamma camera imaging but can be limited for small and low uptake lesions. The recently introduced silicon-photomultiplier-based (SiPM-based) PET/CT (computed tomography) systems outperform previous-generation systems in detector sensitivity, coincidence time resolution, and spatial resolution. Hence, SiPM-based PET/CT shows an improved detectability, particularly for small lesions. In this study, we compare the size-dependant minimum detectable 124I activity (MDA) between the SiPM-based Biograph Vision and the previous-generation Biograph mCT PET/CT systems and we attempt to predict the response to 131I radioiodine therapy of lesions additionally identified on the SiPM-based system. A tumour phantom mimicking challenging conditions (derived from published patient data) was used; i.e., 6 small spheres (diameter of 3.7–9.7 mm), 9 low activity concentrations (0.25–25 kBq/mL), and a very low signal-to-background ratio (20:1). List-mode emission data (single-bed position) were divided into frames of 4, 8, 16, and 30 min. Images were reconstructed with ordinary Poisson ordered-subsets expectation maximization (OSEM), additional time-of-flight (OSEM-TOF) or TOF and point spread function modelling (OSEM-TOF+PSF). The signal-to-noise ratio and the MDA were determined. Absorbed dose estimations were performed to assess possible treatment response to high-activity 131I radioiodine therapy. The signal-to-noise ratio and the MDA were improved from the mCT to the Vision, from OSEM to OSEM-TOF and from OSEM-TOF to OSEM-TOF+PSF reconstructed images, and from shorter to longer emission times. The overall mean MDA ratio of the Vision to the mCT was 0.52 ± 0.18. The absorbed dose estimations indicate that lesions ≥ 6.5 mm with expected response to radioiodine therapy would be detectable on both systems at 4-min emission time. Additional smaller lesions of therapeutic relevance could be detected when using a SiPM-based PET system at clinically reasonable emission times. This study demonstrates that additional lesions with predicted response to 131I radioiodine therapy can be detected. Further clinical evaluation is warranted to evaluate if negative 124I PET scans on a SiPM-based system can be sufficient to preclude patients from blind radioiodine therapy.