Background and objective: Treatment planning through the diagnostic dimension of theranostics provides insights into predicting the absorbed dose of RPT, with the potential to individualize radiation doses for enhancing treatment efficacy. However, existing studies focusing on dose prediction from diagnostic data often rely on organ-level estimations, overlooking intra-organ variations. This study aims to characterize the intra-organ theranostic heterogeneity and utilize artificial intelligence techniques to localize them, i.e. to predict voxel-wise absorbed dose map based on pre-therapy PET. Methods: 23 patients with metastatic castration-resistant prostate cancer treated with [¹⁷⁷Lu]Lu-PSMA I&T RPT were retrospectively included. 48 treatment cycles with pre-treatment PET imaging and at least 3 post-therapeutic SPECT/CT imaging were selected. The distribution of PET tracer and RPT dose was compared for kidney, liver and spleen, characterizing intra-organ heterogeneity differences. Pharmacokinetic simulations were performed to enhance the understanding of the correlation. Two strategies were explored for pre-therapy voxel-wise dosimetry prediction: (1) organ-dose guided direct projection; (2) deep learning (DL)-based distribution prediction. Physical metrics, dose volume histogram (DVH) analysis, and identity plots were applied to investigate the predicted absorbed dose map. Results: Inconsistent intra-organ patterns emerged between PET imaging and dose map, with moderate correlations existing in the kidney (r = 0.77), liver (r = 0.5), and spleen (r = 0.58) (P < 0.025). Simulation results indicated the intra-organ pharmacokinetic heterogeneity might explain this inconsistency. The DL-based method achieved a lower average voxel-wise normalized root mean squared error of 0.79 ± 0.27%, regarding to ground-truth dose map, outperforming the organ-dose guided projection (1.11 ± 0.57%) (P < 0.05). DVH analysis demonstrated good prediction accuracy (R² = 0.92 for kidney). The DL model improved the mean slope of fitting lines in identity plots (199% for liver), when compared to the theoretical optimal results of the organ-dose approach. Conclusion: Our results demonstrated the intra-organ heterogeneity of pharmacokinetics may complicate pre-therapy dosimetry prediction. DL has the potential to bridge this gap for pre-therapy prediction of voxel-wise heterogeneous dose map.

Single-time-point (STP) image-based dosimetry offers a more convenient approach for clinical practice in radiopharmaceutical therapy (RPT) compared with conventional multiple-time-point image-based dosimetry. Despite numerous advancements, current STP methods are limited by the need for strict and late timing in data acquisition, posing challenges in routine clinical settings. This study introduces a new concept of instant STP (iSTP) dosimetry, achieved by predicting the effective half-life (T_eff) of organs using machine learning applied on pretherapy patient data (PET and clinical values). Methods: Data from 22 patients who underwent pretherapy ⁶⁸Ga-gallium N,N-bis[2-hydroxy-5-(carboxyethyl)benzyl]ethylenediamine-N,N-diacetic acid ([⁶⁸Ga]Ga-PSMA-11) imaging and subsequently [¹⁷⁷Lu]Lu-PSMA I&T RPT were analyzed. A machine learning model was developed for T_eff predictions for the left and right kidneys, liver, and spleen subsequently used to estimate time-integrated activity and absorbed dose. iSTP results were compared against multiple-time-point and previously proposed Hänscheid methods. Our method comprised 2 different prediction scenarios, using data before each therapy cycle and from the first cycle. Results: The iSTP method introduced early posttherapy time points (2, 20, 43, and 69 h) for the left kidney, right kidney, liver, and spleen. Dosimetry in the first scenario, aggregating 2 and 20 h, achieved mean differences in time-integrated activity below 27% for all organs. To assess the feasibility, these time points were compared with the best results from the Hänscheid method (kidneys, 69 h; liver and spleen, 20 h). At 2 h, a significant difference (P < 0.001) was found for almost all organs except for the spleen (P = 0.1370). However, at 20 h, no significant differences were found for the right kidney, liver, and spleen, apart from the left kidney (P < 0.01). In the scenario using only the initial PET/CT data to predict T_eff for subsequent cycles, iSTP dosimetry achieved no statistical significance (P > 0.05) for all cycles in comparison to results using PET data before each therapy cycle. Conclusion: Our preliminary results prove the concept for prediction of T_eff with pretherapy data and achieving STP shortly and flexibly after the RPT. The proposed method may expedite the application of dosimetry in broader contexts, such as outpatient or short-duration inpatient treatment.

Purpose: Dopamine transporter imaging is routinely used in Parkinson’s disease (PD) and atypical parkinsonian syndromes (APS) diagnosis. While [¹¹C]CFT PET is prevalent in Asia with a large APS database, Europe relies on [¹²³I]FP-CIT SPECT with limited APS data. Our aim was to develop a deep learning-based method to convert [¹¹C]CFT PET images to [¹²³I]FP-CIT SPECT images, facilitating multicenter studies and overcoming data scarcity to promote Artificial Intelligence (AI) advancements. Methods: A CycleGAN was trained on [¹¹C]CFT PET (n = 602, 72%PD) and [¹²³I]FP-CIT SPECT (n = 1152, 85%PD) images from PD and non-parkinsonian control (NC) subjects. The model generated synthetic SPECT images from a real PET test set (n = 67, 75%PD). Synthetic images were quantitatively and visually evaluated. Results: Fréchet Inception Distance indicated higher similarity between synthetic and real SPECT than between synthetic SPECT and real PET. A deep learning classification model trained on synthetic SPECT achieved sensitivity of 97.2% and specificity of 90.0% on real SPECT images. Striatal specific binding ratios of synthetic SPECT were not significantly different from real SPECT. The striatal left-right differences and putamen binding ratio were significantly different only in the PD cohort. Real PET and real SPECT had higher contrast-to-noise ratio compared to synthetic SPECT. Visual grading analysis scores showed no significant differences between real and synthetic SPECT, although reduced diagnostic performance on synthetic images was observed. Conclusion: CycleGAN generated synthetic SPECT images visually indistinguishable from real ones and retained disease-specific information, demonstrating the feasibility of translating [¹¹C]CFT PET to [¹²³I]FP-CIT SPECT. This cross-modality synthesis could enhance further AI classification accuracy, supporting the diagnosis of PD and APS.

Background: Bone marrow (BM) is the dominant dose‐limiting organ in [ ¹⁷⁷ Lu]Lu‐PSMA‐617 therapy for patients with metastasized castration resistant prostate cancer, where BM dosimetry is challenging due to segmentation. Purpose: We aim to develop an automatic image‐based segmentation method on peri‐therapeutic sequential [ ¹⁷⁷ Lu]Lu‐PSMA‐617 images for personalized BM dosimetry. Methods: Quantitative SPECT/CT imaging at 2, 20, 40 and 60 ( n = 14)/200 ( n = 16) h post [ ¹⁷⁷ Lu]Lu‐PSMA‐617 injection were analyzed for 10 patients with 30 treatment cycles. X‐means clustering was applied on the deep learning‐based segmented lumbar spines CT images to classify the BM region. A single threshold method, two empirical segmentation methods (one sphere and five spheres), and gold standard manual segmentation were also implemented. The Dice similarity coefficient between BM masks of the X‐means clustering and single threshold method was calculated as compared to the gold standard. BM mean absorbed dose ( D _mean ) was obtained for different segmentation methods. Absolute errors and Bland–Altman analysis were also evaluated for BM D _mean derived from evaluated segmentation methods compared with the gold standard. Wilcoxon signed‐rank test was performed for statistical evaluation. BM D _mean was correlated with the change of platelets and white blood cells (WBC) pre‐ and post‐treatment using Pearson correlation analysis. Results: In 30 cycles of 10 patients, the average Dice is 0.76 ± 0.18 for the X‐means clustering method, as compared to 0.61 ± 0.19 for the single threshold method. The gold standard yields mean BM D _mean of 0.46 ± 0.69 Gy. The X‐means clustering method exhibits significantly ( p < 0.01) lower mean absolute BM D _mean (25.34 ± 64.48%) errors, followed by the single threshold (32.46 ± 69.49%), one sphere (50.53 ± 35.40%), and five spheres (72.73 ± 115.97%) methods. Bland–Altman analysis reveals that the X‐means clustering method has a smaller D _mean difference (0.0330 Gy) compared to the single threshold (0.0512 Gy), one sphere (−0.1903 Gy), and five spheres (0.2108 Gy) methods. Stronger correlations (r ≤ −0.65) are found between platelets/WBC changes and BM D _mean from both the gold standard and X‐means clustering methods than other methods. Conclusions: X‐means clustering is feasible to segment the BM based on the CT images of peri‐therapy SPECT/CT and shows advantages compared with the single threshold and empirical sphere segmentation methods.

Objective: Interpretability and reproducibility remain major challenges in applying deep neural network (DNN) to neuroimaging-based diagnosis. This study proposes a radiomics-guided dual-channel deep neural network (RDDNN) to improve feature transparency and enhance clinical understanding in the classification of Parkinsonian syndromes. Methods: In this bi-centric study, we analysed two independent cohorts comprising 1,275 patients with idiopathic Parkinson’s disease (IPD), multiple system atrophy (MSA), and progressive supranuclear palsy (PSP), alongside 223 healthy controls from Huashan Hospital and 90 patients with IPD, MSA, and PSP (34IPD, 17MSA, 39PSP) from the University Hospital Munich. It is a re-analysis of well-studied Chinese and German cohorts of ¹⁸ F-fluorodeoxyglucose Positron emission tomography (FDG-PET) imaging of parkinsonian patients and the FDG scans were of 10-min static acquisition at 60 min post FDG injection and normalized against whole brain activity. The RDDNN model combines local features extracted via dilated convolutional networks and global features derived from Transformer-based self-attention networks. Model performance was evaluated using classification metrics and compared to radiomics and DNN approaches. The model’s outputs were also compared with nuclear medicine specialists’ visual assessments to assess interpretability and time efficiency. Furthermore, SHapley Additive Explanations (SHAP), Layer-wise Class Activation Mapping (Layer-CAM), and Rollout Attention Map (RAM) were employed to evaluate which features played the most critical roles in the model’s final classification decisions after supervised training, and to examine how both networks spatially corresponded to known brain connectivity regions. Results: In the internal blind-test cohort, the RDDNN achieved high accuracy (AUC = 0.99, accuracy = 0.98). SHAP and correlation analyses jointly indicated complementary information across channels, some of which were clinically interpretable. In the external cohort, the model maintained robust performance (AUC = 0.94, accuracy = 0.81), with consistent feature patterns across populations. The model significantly reduced evaluation time compared to nuclear medicine specialists’ readings ( p < 0.001), and the heatmaps showed disease-specific activation in anatomically relevant regions for IPD, MSA, and PSP. Conclusion: The RDDNN framework provides a clinically interpretable and reproducible DNN solution for classifying Parkinsonian disorders. By integrating radiomics and attention-based modeling, it enhances lesion localization, supports clinical decision-making, and offers performance comparable to human specialists—while substantially improving diagnostic efficiency.

Medical imaging has been intensively employed in screening, diagnosis and monitoring during the COVID-19 pandemic. With the improvement of RT–PCR and rapid inspection technologies, the diagnostic references have shifted. Current recommendations tend to limit the application of medical imaging in the acute setting. Nevertheless, efficient and complementary values of medical imaging have been recognized at the beginning of the pandemic when facing unknown infectious diseases and a lack of sufficient diagnostic tools. Optimizing medical imaging for pandemics may still have encouraging implications for future public health, especially for long-lasting post-COVID-19 syndrome theranostics. A critical concern for the application of medical imaging is the increased radiation burden, particularly when medical imaging is used for screening and rapid containment purposes. Emerging artificial intelligence (AI) technology provides the opportunity to reduce the radiation burden while maintaining diagnostic quality. This review summarizes the current AI research on dose reduction for medical imaging, and the retrospective identification of their potential in COVID-19 may still have positive implications for future public health.