Despite the potential of deep learning (DL)-based methods in substituting CT-based PET attenuation and scatter correction for CT-free PET imaging, a critical bottleneck is their limited capability in handling large heterogeneity of tracers and scanners of PET imaging. This study employs a simple way to integrate domain knowledge in DL for CT-free PET imaging. In contrast to conventional direct DL methods, we simplify the complex problem by a domain decomposition so that the learning of anatomy-dependent attenuation correction can be achieved robustly in a low-frequency domain while the original anatomy-independent high-frequency texture can be preserved during the processing. Even with the training from one tracer on one scanner, the effectiveness and robustness of our proposed approach are confirmed in tests of various external imaging tracers on different scanners. The robust, generalizable, and transparent DL development may enhance the potential of clinical translation.

Pre-therapy dosimetry prediction is a prerequisite for treatment planning and personalized optimization of the emerging radiopharmaceutical therapy (RPT). Physiologically based pharmacokinetic (PBPK) model, describing the intrinsic pharmacokinetics of radiopharmaceuticals, have been proposed for pre-therapy prediction of dosimetry. However, it is restricted with organ-wise prediction and the customization based on pre-therapy measurements is still challenging. On the other side, artificial intelligence (AI) has demonstrated the potential in pre-therapy dosimetry prediction. Nevertheless, it is still challenging for pure data-driven model to achieve voxel-wise prediction due to huge gap between the pre-therapy imaging and post-therapy dosimetry. This study aims to integrate the PBPK model into deep learning for voxel-wise pre-therapy dosimetry prediction. A conditional generative adversarial network (cGAN) integrated with the PBPK model as regularization was developed. For proof of concept, 120 virtual patients with 68Ga-PSMA-11 PET imaging and 177Lu-PSMA-I&T dosimetry were generated using realistic in silico simulations. In kidneys, spleen, liver and salivary glands, the proposed method achieved better accuracy and dose volume histogram than pure deep learning. The preliminary results confirmed that the proposed PBPK-adapted deep learning can improve the pre-therapy voxel-wise dosimetry prediction and may provide a practical solution to support treatment planning of heterogeneous dose distribution for personalized RPT.

Prostate-specific membrane antigen radioligand therapy (PSMA-RLT) has emerged as a promising treatment for metastatic castration-resistant prostate cancer (mCRPC). However, current patient selection methods – largely based on qualitative imaging criteria – may impede precision and efficacy of treatment. We aimed to evaluate the predictive value of quantitative imaging biomarkers derived from dual-tracer [ ⁶⁸ Ga]Ga-PSMA-11 and [ ¹⁸ F]F-FDG PET/CT, with a focus on concordant lesions. Methods: Thirty-seven mCRPC patients from two institutions underwent [ ⁶⁸ Ga]Ga-PSMA-11 and [ ¹⁸ F]F-FDG PET/CT prior to receiving at least two cycles of [ ¹⁷⁷ Lu]Lu-PSMA therapy. An automated pipeline enabled lesion segmentation, dual-tracer image fusion, and extraction of quantitative features from concordant (PSMA + /FDG +) and non-concordant lesions. A decision tree model was developed on the Vienna cohort ( n = 24) and validated on an independent cohort from Augsburg ( n = 13). SHAP analysis was used to identify key predictive features. Results: The decision tree achieved 95.8% accuracy in the training cohort and 84.6% in external validation. SUV _mean of concordant lesions was the most predictive features. Patients with SUV _mean [PSMA Concordant] ≥ 12.1 g/mL were more likely to respond. Organ-specific analysis further identified high SUV _max in bone metastases as a negative prognostic marker. Conclusions: Quantitative metrics from dual-tracer PET, particularly those characterizing concordant lesions, show promise for predicting response to PSMA-RLT. These preliminary findings highlight the potential to move beyond binary eligibility criteria toward a more nuanced, biomarker-driven approach to patient selection.

Polarization imaging has advantages to reveal objective characteristics in foggy conditions. Quality assessment of polarization analysis images in imaging is rarely reported. A quality assessment method suitable to polarization analysis images in foggy conditions is proposed. Firstly, polarization characteristics in foggy conditions are analyzed. Three types of quality factors are selected, which are contrast factor based on brightness, the moment of inertia factor based on structure, and MSCN (Mean Subtracted Contrast Normalized) factor based on Stokes parameters. Corresponding assessment model is constructed, and pooling strategies are designed to map quality score of intensity of polarization image I and degree of polarization image P. Experiments on images captured by polarization camera in simulation foggy environment and generated by fog simulation of Photoshop. Results show that the proposed method is well consistent with subjective observation and analysis.

Background: While TSH suppression is essential in patients with differentiated thyroid cancer (DTC) to reduce the risk of recurrence, it has also been linked to side effects, particularly a reduction in bone mineral density that may contribute to osteoporosis. However, previous studies investigating this association have yielded inconsistent results. This study aimed to evaluate bone density using Hounsfield units from PET/CT scans in a longitudinal analysis including both sexes. Methods: Patients with DTC under continuous TSH suppression who underwent two PET/CT scans were included. Hounsfield units were measured for each lumbar vertebra (L1–L5) in the CT by placing an elliptical region of interest (ROI) in the center of the vertebra, avoiding hyperdense edges. Laboratory parameters were also collected. Results: A total of 50 patients were included in the study (25 male, 25 female), with a mean age of 57.2 (±15.3) years at the time of the first scan. The mean duration of TSH suppression before the first scan was 3.7 ± 3.9 years, and the mean interval between both scans was 4.4 ± 4.0 years. At the follow-up scan, bone density was significantly lower compared with baseline for all lumbar vertebrae (L1–L5 combined and individually) (all p < 0.05). Subgroup analysis revealed a significant decline in women at L1, L2, L4, and L5 and for overall lumbar bone density, while men showed nonsignificant trends. Conclusions: Our study suggests a sustained reduction in vertebral bone density during TSH suppression. The results support routine monitoring in both sexes, risk stratification by age and duration of suppression, and, when oncologically appropriate, consideration of lower suppression intensity or initiation of bone-protective therapy in high-risk patients.