Qian Liu, Shangyan Wei, Huijuan Peng, Yuna Tang, Xinyuan Chen, Jianrong Dai, Kuo Men
{"title":"A cascade deep learning network for simultaneous prediction of dose and linear energy transfer in proton therapy of prostate cancer.","authors":"Qian Liu, Shangyan Wei, Huijuan Peng, Yuna Tang, Xinyuan Chen, Jianrong Dai, Kuo Men","doi":"10.1088/1361-6560/aeba33","DOIUrl":"10.1088/1361-6560/aeba33","url":null,"abstract":"<p><p><i>Objective.</i>Prediction of dose-averaged linear energy transfer (LETd) may facilitate LET-aware evaluation of proton therapy plans. This study developed a cascade deep learning framework for simultaneous three-dimensional prediction of dose and LETd distributions for conventional dose-based optimization (DoseOpt) and LETd-incorporated optimization (LETOpt) in prostate intensity-modulated proton therapy.<i>Approach.</i>A total of 150 patients with prostate cancer were included, each with DoseOpt and LETOpt plans. Planning CT images and anatomical structure masks were used as inputs. The cascade network transferred an intermediate latent representation from the dose prediction pathway to the LETd prediction pathway. Stepwise, multitask, and anatomy-only LETd (only-LET) networks served as comparators. Model performance was evaluated using five-fold cross-validation.<i>Main results.</i>All dose-predicting models showed high spatial agreement, with cascade Dice similarity coefficients of approximately 0.93, 0.96, and 0.95 for the 50%, 75%, and 95% prescription isodose volumes. No significant differences in structure-specific dose errors remained after false discovery rate correction. The cascade model achieved the lowest global LETd mean absolute error (MAE) for DoseOpt and LETOpt, 0.108 ± 0.025 and 0.111 ± 0.025 keV<i>μ</i>m<sup>-1</sup>, respectively, significantly lower than those of the multitask, stepwise, and only-LET models. Structure-specific LETd performance was endpoint dependent, with favorable cascade performance for several clinical target volume and organ-at-risk endpoints. LETOpt primarily redistributed LETd while largely preserving dose, and the cascade model reproduced the principal plan-specific changes, including reductions in rectal and rectal-wall LETd. Mean inference time was 0.505 ± 0.020 s per patient.<i>Significance.</i>The proposed framework enables rapid simultaneous prediction of dose and LETd for predefined planning strategies from patient anatomy without requiring a precomputed dose distribution, supporting LET-aware plan comparison and providing a basis for future LETd-guided biological optimization.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":" ","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-11-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149288292","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Hui Huang, Gangxin Hou, Haonan Li, Sen Yang, Zhonghua Kuang, Ning Ren, Yongfeng Yang, Zheng Liu
{"title":"Beam-width-corrected DOI resolution measurement of dual-ended readout PET detectors.","authors":"Hui Huang, Gangxin Hou, Haonan Li, Sen Yang, Zhonghua Kuang, Ning Ren, Yongfeng Yang, Zheng Liu","doi":"10.1088/1361-6560/aebd34","DOIUrl":"10.1088/1361-6560/aebd34","url":null,"abstract":"<p><p><i>Objective.</i>Dual-ended readout PET detectors provide depth-of-interaction (DOI) information for reducing parallax errors in high-resolution PET imaging. This study compared LYSO detectors with different inter-crystal reflectors and investigated the effect of irradiation beam width on DOI resolution measurements under electronic collimation.<i>Approach.</i>Two 28 × 28 LYSO arrays with 0.75 × 0.75 × 10 mm<sup>3</sup>crystals and BaSO<sub>4</sub>or E60 reflectors were characterized in terms of flood-map quality, energy resolution, and DOI resolution. Part of E60 detector (8 × 8 crystals) was further measured using three irradiation geometries with beam widths of 0.80 mm, 0.62 mm, and 0.46 mm. A beam-width correction was applied to the measured DOI resolution, and the effects of energy selection and empirical flood-map-based inter-crystal scattering (ICS) filtering were evaluated.<i>Main results.</i>The BaSO<sub>4</sub>detector showed better flood-map quality and energy resolution, whereas the E60 detector showed better DOI resolution. In the 400-650 keV energy window, the measured average DOI resolutions of entire 28 × 28 LYSO pixels with irradiation beam width of 1.00 mm were 1.92 and 1.75 mm for the BaSO<sub>4</sub>and E60 detectors, respectively. For the selected 8 × 8 crystal region of the E60 detector, the corrected DOI resolution by subtracting the irradiation beam width improved from 1.19 mm for all events to 1.01 mm after energy selection, and further to 0.86 mm after additional empirical flood-map-based ICS filtering.<i>Significance.</i>Irradiation beam width introduces a geometry-dependent contribution to DOI resolution measurements. Beam-width correction enables more consistent DOI characterization under different irradiation geometries.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":" ","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-11-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149324045","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Anatomical correlates of delivered TIS fields across deep and cortical targets.","authors":"Tianxiang Guo, Suiping Zhou","doi":"10.1088/1361-6560/aebc0b","DOIUrl":"10.1088/1361-6560/aebc0b","url":null,"abstract":"<p><p>The objective of this study was to quantify inter-individual variability in target-region Temporal interference stimulation (TIS) envelope amplitude (EA) and to examine its associations with anatomical features across deep and cortical targets.We built individualized finite element head models forhealthy HCP participants aged 22-35 years and evaluated four montages across five prespecified target ROI-montage pairs: two hippocampal pairs, left and right thalamic pairs under a single symmetric montage, and one motor-cortex pair. For each target ROI-montage pair, we quantified mean TIS EA within target region (EA, V m) and the target-to-brain envelope-amplitude ratio (TBR) as measures of target-field magnitude and spatial selectivity, respectively. Two prespecified global volume ratios were assessed with Pearson correlations, and 20 anatomical features were evaluated using two-stage LASSO-OLS models with family-grouped repeated hold-out validation. Bootstrap selection frequencies were calculated for the EAmodels.Mean EAranged from 0.252-0.512 V mwith substantial inter-individual variability (CV = 28.1%-40.0%). Two correlated global brain-fraction measures were positively associated with EAacross the five target ROI-montage pairs, but multivariable models explained only a modest proportion of variance (-; median family-grouped test-). EAand TBR models selected different feature sets, with no predictor reachingin either thalamic TBR model; the left and right thalamic targets under the single symmetric T1 montage correlated strongly ().This population-scale computational dosimetry analysis quantifies how fixed TIS currents and montages translate into target-field magnitudes across individual head models. The results distinguish the anatomical correlates of delivered amplitude from those of spatial selectivity and provide quantitative information for anatomy-informed TIS planning.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":" ","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-11-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149299805","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Study on the capture and thrombolytic efficacy of magnetic particle clusters under a deeply focused rotating magnetic field.","authors":"Shoulong Dong, Zongnan Liu, Yuan Lei, Sizhe Xiang, Lisheng Zhao, Dengfeng He, Chenguo Yao","doi":"10.1088/1361-6560/aeb643","DOIUrl":"10.1088/1361-6560/aeb643","url":null,"abstract":"<p><p><i>Objective</i>. Magnetic particles can serve as carriers for tissue plasminogen activator (tPA) and can be targeted and controllably manipulated under a rotating magnetic field (RMF), providing a potential approach for localized thrombolysis in venous thrombosis. However, current magnetic-control strategies remain limited at larger working distances and under flow conditions because of magnetic-field attenuation, insufficient focusing capability, and blood-flow-induced washout. Therefore, maintaining effective capture and stable motion of magnetic particle clusters (MPCs) remains an important challenge.<i>Approach.</i>To address this issue, we developed a deeply focused rotating magnetic field (DF-RMF) system driven by two synchronously rotating permanent magnets. The system generates a magnetic-field region with an axial dual-peak distribution and centripetal action, thereby promoting MPC accumulation within the target region while maintaining rotational motion. Numerical simulations and<i>in vitro</i>experiments were conducted to evaluate the magnetic-field characteristics, MPC rotational behavior, capture performance under flow conditions, and thrombolytic efficacy. Experiments were performed in phosphate-buffered saline (PBS) and a glycerol-water solution (GWS) with higher blood-relevant viscosity.<i>Main results.</i>Compared with a single-permanent-magnet rotating magnetic field, the DF-RMF provided better particle capture performance at larger working distances. In PBS, particle capture could still be maintained to some extent at a working distance of 5 cm, while particle retention was still observed at a flow velocity of 14 cm s<sup>-1</sup>and a working distance of 2 cm. Experiments in GWS showed that higher blood-relevant viscosity reduced the long-distance capture capability of the DF-RMF. Finite-element simulations showed that MPC rotation generated localized recirculating flow, velocity gradients, and periodic pressure variations in the surrounding fluid, potentially facilitating tPA transport near the thrombus interface. In the fully occluded thrombus model, the DF-RMF-assisted group achieved complete recanalization, with a thrombus clearance rate approximately five times that of the tPA-only group.<i>Significance.</i>The DF-RMF improved the long-distance capture and anti-washout retention of magnetic particles under flow conditions and enhanced the<i>in vitro</i>thrombolytic effect of tPA, providing a new approach for magnetically assisted thrombolysis.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":" ","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-11-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149224731","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Bohan Yang, Hang Lian, Gang Liu, Yang Zhong, Riao Dao, Yujia Qian, Ke Shi, Anke Tang, Yong Luo, Qi Kong, Jingnan Liu
{"title":"Unsupervised deep learning model for fast energy layer pre-selection of delivery-efficient proton arc therapy plan optimization of nasopharyngeal carcinoma.","authors":"Bohan Yang, Hang Lian, Gang Liu, Yang Zhong, Riao Dao, Yujia Qian, Ke Shi, Anke Tang, Yong Luo, Qi Kong, Jingnan Liu","doi":"10.1088/1361-6560/aec10b","DOIUrl":"https://doi.org/10.1088/1361-6560/aec10b","url":null,"abstract":"<p><strong>Objective: </strong>Proton arc therapy (PAT) is an emerging and promising modality in radiotherapy, offering improved dose distribution and treatment robustness over conventional intensity-modulated proton therapy. Yet, identifying the optimal energy layer (EL) sequence remains challenging due to the high computational demand and prolonged treatment delivery time. This study proposes unsupervised deep learning for fast EL pre-selection, targeting EL switching (ELS) time and plan quality.

Approach. We introduce a spot-count representation that combines structure-specific spot counts with spatial dose-influence features, organized by gantry angle and EL. A U-Net-style model, SPArc_dldp, uses this representation and water-equivalent thickness features to score candidate ELs, with training objectives based on spot counts and ELS time. We evaluated SPArc_dldp against SPArc_pso in 35 paired nasopharyngeal carcinoma cases, including nine-scenario robust optimization and evaluation.

Main results. SPArc_dldp achieved a significantly higher nominal clinical target volume (CTV) conformity index (CI: 0.613±0.034 versus 0.557±0.040; p<0.001). It showed better homogeneity (HI: 0.395±0.048 versus 0.414±0.060; p=0.135). EL-selection computation took 2.218±0.022 ms for DL and 14.18±2.71 s for PSO. DL had significantly shorter ELS time (28.59±4.01 versus 72.19±5.56 s; p<0.001), but higher optic-nerve and parotid mean doses. Following robust optimization, DL had a higher CTV root-mean-square deviation dose-volume histogram area (RVH-AUC: 10.71±1.00 versus 8.69±0.73 per-fraction Gy(RBE)·%; p<0.001), indicating greater dose variation.

Significance. The proposed spot-count representation supports rapid EL pre-selection for PAT planning. In this exploratory comparison, SPArc_dldp achieved higher nominal CI and shorter ELS time, alongside higher mean doses to some OARs and greater CTV dose variation.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":" ","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-11-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149395024","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Cramér-Rao lower bound for spectral propagation-based x-ray phase-contrast imaging.","authors":"Giavanna Jadick, Patrick La Rivière","doi":"10.1088/1361-6560/aebd38","DOIUrl":"10.1088/1361-6560/aebd38","url":null,"abstract":"<p><p><i>Objective.</i>Spectral x-ray phase-contrast imaging (XPCI) is a promising emerging method for quantitative phase retrieval. With an energy-discriminating detector, the independent measurements needed for quantitative phase retrieval or material decomposition can be acquired in a single shot. The existing Cramér-Rao lower bound (CRLB) model for attenuation-based spectral imaging, however, cannot capture the spatially distributed phase information associated with propagation-based (PB) XPCI. In this work, we develop a generalized CRLB framework for spectral PB-XPCI.<i>Approach.</i>The PB-XPCI forward model was formulated for nonideal, polychromatic photon-counting detector measurements. Two-material toy models were constructed with simple edge geometry to parallel the attenuation-based single-ray CRLB, with Poisson Fisher information computed over both energy-resolved and spatially distributed detector measurements. The conventional single-ray and proposed single-edge CRLBs were compared for ideal dual-monoenergy acquisitions. The single-edge model was then used to optimize detector energy thresholds for a nonideal 30 kVp polychromatic system, and CRLB predictions were validated against simulations.<i>Main results.</i>The single-edge model yielded substantially lower CRLBs relative to the fluence-matched single-ray model, while free-space propagation provided additional energy-dependent improvements. For the idealized monoenergetic single-edge model, the CRLB for each basis material was minimized when both measurements used the same energy, indicating that spatial measurements supplied sufficient independent information for material decomposition. For the polychromatic system, the CRLB exhibited material-dependent optimal energy thresholds, which shifted with the number of available energy bins. The single-edge and 2D-geometry CRLB models identified similar optimal energy thresholds within 0.1-1.9 keV, while simulated maximum-likelihood thickness estimates closely approached the CRLB with mean variance ratios of 1.006-1.020.<i>Significance.</i>The proposed CRLB model retains the interpretability of the familiar attenuation-based model while including PB phase information, providing a useful theoretical foundation for evaluation and optimization of emerging spectral PB-XPCI systems as well as proposed phase retrieval and material decomposition algorithms.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":" ","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-11-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149324046","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"OpenTOPAS-SPECT: an extension for full-physics clinical and preclinical SPECT simulation.","authors":"Victor V Onecha, Alejandro Bertolet","doi":"10.1088/1361-6560/aebd36","DOIUrl":"10.1088/1361-6560/aebd36","url":null,"abstract":"<p><p><i>Objective.</i>Single-photon emission computed tomography (SPECT) simulations must resolve collimator transport, detector energy deposition, and low-efficiency projection formation within one model. General transport codes leave users to build the imaging components, whereas dedicated SPECT codes offer efficient camera models with less freedom to represent new systems. We developed OpenTOPAS-SPECT to bring projection-domain SPECT simulation into the declarative parameter-file workflow of OpenTOPAS.<i>Approach.</i>We added parametric parallel-hole, converging, pinhole, multi-pinhole, and slit-slat collimators; sodium-iodide and cadmium-zinc-telluride detector configurations; energy-resolved and forced-detection scorers; native decay-chain sources; and an evidence-graded library of 14 camera presets. We tested preset construction, multi-view consistency, and variance reduction, and benchmarked projection-domain quantities against measured, independent-code, and analytic references.<i>Main results.</i>At 140.5 keV and 10 cm from the collimator face, the Siemens Symbia low-energy high-resolution (LEHR) configuration gave 7.49 × 10<sup>-5</sup>counts per decay, or 0.82 of the measured 9.1 × 10<sup>-5</sup>. GE collimator sensitivities were 1.08-1.15× the SIMIND range midpoints, while the 364.5 keV low-to-high-energy sensitivity ratio was 59.5 versus 67.7. Planar point-spread widths were 4%-9% below analytic geometric resolution from 5 to 30 cm. In this Symbia LEHR point-source geometry, forced detection reached matched precision 6.9× faster than full transport and reproduced its absolute sensitivity to 1% without calibration. Across 20 configurations (five collimators, 113 keV-1 MeV), forced detection stayed within 4% of full transport while septa carried up to a fifth of the signal, but fell 7% to 25% below it where they carried 25% or more. The release-gating suite ran all 23 example decks and built all 14 presets.<i>Significance.</i>OpenTOPAS-SPECT provides reusable SPECT components within the OpenTOPAS parameter system, with bounded projection-domain benchmarks for the first stable public release. The evidence supports the tested configurations and observables; reconstructed-image accuracy and broader system coverage remain untested.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":" ","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-11-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149323969","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Image reconstruction based on FWHM-energy joint weighting for magnetic particle imaging.","authors":"Chunhao Yin, Shijie Sun, Tong Zhao, Yuntao Bing, Lijun Xu, Jing Zhong","doi":"10.1088/1361-6560/aebd62","DOIUrl":"10.1088/1361-6560/aebd62","url":null,"abstract":"<p><p><i>Objective.</i>Magnetic particle imaging (MPI) enables high-sensitivity quantitative imaging by detecting the nonlinear magnetization response of superparamagnetic nanoparticles. field-free-point (FFP)-gridded MPI has been developed to reduce system complexity, and its multi-harmonic extension exploits harmonic-dependent information to balance spatial resolution and SNR. However, an appropriate weighting strategy is required to utilize complementary information across harmonics. We propose an image reconstruction method based on full width at half maximum (FWHM)-energy joint weighting, which combines PSF-derived FWHM with either L2-norm harmonic energy or noise energy to achieve a favorable trade-off between spatial resolution enhancement and reconstruction stability.<i>Approach.</i>In the FFP-gridded multi-harmonic fusion framework, the 3rd, 5th, 7th, and 9th harmonic components are jointly used for image reconstruction. Since different harmonics exhibit distinct characteristics in signal intensity, noise level, and spatial resolution, optimizing their weights is essential for improving reconstruction quality. Traditional weighting strategies based on amplitude, signal energy, and noise characteristics were implemented as comparison methods. On this basis, two FWHM-energy joint weighting strategies were developed: L2-norm energy-FWHM weighting for balancing resolution improvement and artifact suppression, and noise-energy-FWHM weighting for enhancing the spatial resolution of noise-suppressed reconstruction.<i>Main Results.</i>The proposed methods were validated through numerical simulations and phantom experiments under different noise conditions and phantoms. The results demonstrate that the FWHM-energy joint weighting strategies improved reconstruction quality compared with traditional weighting methods. By incorporating spatial resolution information into harmonic weighting, the proposed methods achieved a more favorable balance among spatial resolution, artifact suppression, and reconstruction stability.<i>Significance.</i>The proposed method introduces FWHM-energy joint weighting strategies for multi-harmonic MPI reconstruction, enabling spatial resolution enhancement while preserving reconstruction stability. Future work will focus on extending this framework to adaptive weighting schemes and more complex scenarios, thereby improving the applicability of multi-harmonic MPI reconstruction.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":" ","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-11-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149324047","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Xin Song, QuanQi Liu, YiMin Wang, Ke Wang, Siyang Xu, JiaYi Ma
{"title":"EFM-GAN: enhanced feature modeling for multimodal MRI synthesis under missing-modality settings.","authors":"Xin Song, QuanQi Liu, YiMin Wang, Ke Wang, Siyang Xu, JiaYi Ma","doi":"10.1088/1361-6560/aebd5f","DOIUrl":"https://doi.org/10.1088/1361-6560/aebd5f","url":null,"abstract":"<p><p><i>Objective</i>. Multimodal magnetic resonance imaging (MRI) synthesis has been widely investigated for recovering unavailable imaging modalities under incomplete acquisition scenarios. However, existing approaches often rely on implicit feature fusion strategies, where modality-specific appearance information and cross-modal anatomical representations are entangled in a shared latent space. Such entanglement may lead to unstable feature interaction and limit the preservation of modality-invariant structural information. Moreover, although multimodal MRI sequences are generally spatially registered, modality-specific acquisition mechanisms introduce heterogeneous feature distributions, leading to feature-level semantic inconsistency during cross-modal representation learning. Lesion-related details are also frequently underrepresented, particularly in anatomically complex regions that are difficult to reconstruct.<i>Approach</i>. To address these limitations, an enhanced feature modeling generative adversarial network (EFM-GAN) is proposed with an explicit parallel feature modeling framework. In EFM-GAN, parallel feature branches explicitly model modality-specific appearance and cross-modal shared representations. A bidirectional progressive alignment module is further introduced to progressively reduce feature-level semantic discrepancies between modality-specific representations before feature fusion, while preserving modality-dependent appearance information. Furthermore, a reconstruction difficulty-aware loss is designed to adaptively reweight reconstruction constraints, assigning greater emphasis to regions that are harder to reconstruct.<i>Main results</i>. Extensive experiments on the BRaTS2020 and BRaTS2021 datasets demonstrate that EFM-GAN achieves higher peak signal-to-noise ratio and structural similarity index than the evaluated baseline methods across multiple target modalities, with improved structural consistency and synthesis fidelity.<i>Significance</i>. The results demonstrate the effectiveness of explicitly modeling modality-specific and shared representations, progressively aligning heterogeneous features, and emphasizing reconstruction-difficult regions for missing-modality MRI synthesis.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":"71 21","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-11-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149387523","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Closed-form, engine-portable uncertainty prediction of deformable image registration for dose accumulation in adaptive radiotherapy.","authors":"Riccardo Ruggieri, Nicola Bianchi, Stefania Naccarato, Ruggero Ruggieri","doi":"10.1088/1361-6560/aec10c","DOIUrl":"https://doi.org/10.1088/1361-6560/aec10c","url":null,"abstract":"<p><strong>Objective: </strong>Voxel-level uncertainty estimates of the deformation vector field (DVF) produced by commercial deformable image registration (DIR) engines are flagged as an open issue by AAPM TG-132 but seldom available, preventing principled propagation into accumulated dose-volume histograms (DVHs). We propose and validate an engine-portable per-voxel σ predictor for the DVF (μ) of a deterministic DIR, calibrated by one scalar f against landmark Target Registration Error (TRE), with no access to the engine's internals.

Approach. The framework combines (i) an anisotropic Gaussian correlation kernel, longer along the cranio-caudal axis, generalising the isotropic kernel of Smolders et al. (2023); and (ii) a heuristic-residual σ_h=f[wR̃+(1-w)d̃] (default w=0.7), where R̃ combines a self-similarity descriptor residual with an image-gradient suppression factor and d̃ is |Jacobian determinant-1|, both normalised to [0,1]; σ_h is closed-form, computed from the image pair and the DVF without optimisation or training. For each combination of dataset and DIR engine (a cell), f is calibrated against TRE via grid search minimising the Landmark Expected Calibration Error (LECE).

Main results. Across 4 publicly annotated computed tomography datasets (Lung-, Abdominal- and Liver-DIR-QA and DIR-LAB Castillo; 44,778 landmarks) and 4 DIR engines (ANTs SyNRA, ConvexAdam, uniGradICON, Elastix), with f fitted on other patients (5-fold cross-validation), the calibrated σ_h achieves, in the per-direction convention of Smolders et al., LECE within or below their 0.41-1.34mm range in all 16 cells (cases with final root-mean-square TRE ≤10mm retained). A constant σ calibrated likewise also meets this range, but the variation of σ_h raises the held-out log-likelihood in 13 cells; σ_h ranks landmarks by error only weakly (Spearman ρ=-0.11 to +0.34).

Significance. A one-scalar, post-hoc calibration yields voxel-level DVF uncertainty for any deterministic DIR that exports its DVF, partly addressing the AAPM TG-132 gap; propagation into accumulated DVH bands is illustrated on one lung stereotactic body radiotherapy case.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":" ","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-11-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149394948","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}