搜索结果：voxel

For the past several decades, it has been popular to reconstruct Fourier imaging data using model-based approaches that can easily incorporate physical constraints and advanced regularization/machine learning priors. The most common modeling approach is to represent the continuous image as a linear combination of shifted "voxel" basis functions. Although well-studied and widely-deployed, this voxel-based model is associated with longstanding limitations, including high computational costs, slow convergence, and a propensity for artifacts. In this work, we reexamine this model from a fresh perspective, identifying new issues that may have been previously overlooked (including undesirable approximation, wrap-around, and nullspace characteristics). Our insights motivate us to propose a new model that is more resilient to the limitations (old and new) of the previous approach. Specifically, the new model is based on a Fourier-domain basis expansion rather than the standard image-domain voxel-based approach. Illustrative results, which are presented in the context of non-Cartesian MRI reconstruction, demonstrate that the new model enables improved image quality (reduced artifacts) and/or reduced computational complexity (faster computations and improved convergence).

Ultrasonic row-column addressed (RCA) arrays enable volumetric imaging with reduced channel count while maintaining a large aperture. With full matrix capturing (FMC) data and total focusing method (TFM) reconstruction, RCA probes can deliver image quality comparable to that of fully addressed 2-D arrays. However, the large number of transmit-receive events required for FMC-TFM severely limit the achievable volume rate, and even with reduced transmissions in plane wave imaging (PWI), conventional delay-and-sum (DAS) beamformer remains computationally inefficient for large 3-D voxel grids. To overcome these problems, an efficient wavenumber-domain algorithm for 3-D PWI with RCA arrays is proposed in this paper. The forward model for RCA PWI acquisition is derived, and the inverse Stolt mapping is established. By treating the transmit angle as an explicit dimension, the measurement-domain spectrum is completed. Two orthogonal image volumes are then reconstructed and coherently combined using a conjugate cross-correlation to suppress sidelobes and improve signal-to-noise ratio (SNR). Moreover, the calculation in wavenumber-domain dramatically reduces the computational cost. Simulation and experimental results demonstrate that the proposed method performs better in terms of imaging quality and efficiency. Specifically, it achieves an acceleration factor of approximately 231-fold over the DAS method for a volume of 256 × 256 × 512 voxels, indicating strong potential for real-time volumetric ultrasound imaging in practical applications.

Task-based precision mapping has become a promising technique in functional MRI (fMRI) to robustly characterize and map an individual's unique activity patterns. These experiments consist of acquiring extensive imaging data in one participant, ultimately improving the sensitivity and specificity of individual-specific functional localization. Despite its advantages, studies have primarily focused on understanding individual-specific cortical activation, preventing a holistic view of a systems-level functional response, and to date, best approaches for the statistical analysis of controlled task-based, densely sampled, whole-brain data have not yet been fully established. Therefore, in this study, we collected whole-brain (i.e. covering cortex, cerebellum, and brainstem) multi-echo densely sampled data of the auditory system, a system with major subcortical components, and evaluated activation sensitivity as well as activation stability across data subsets of commonly-used whole-brain and region-specific inference testing approaches. The whole-brain approaches involved standard voxel-level and cluster-level inference schemes with varying statistical thresholds and a non-parametric permutation inference approach. The region-specific approaches involved an exploratory top % t-statistics methods and non-parametric permutation inference approaches. We found that a whole-brain voxel-level approach with a false discovery rate (FDR) correction (p<0.05) presented highest sensitivity across regions and subjects as well as most consistent detection of expected auditory regions, even with lower scan duration. In addition, we found that a region-specific top % t-statistic approach may be a useful exploratory functional localization tool and a complementary method to standard inference testing approaches.

Assessing muscle activity is essential for diagnosis and treatment of movement disorders such as dystonia and spasticity. While task-based muscle functional magnetic resonance imaging (m-fMRI) enables non-invasive imaging of muscle activation, conventional methods rely on comparisons between rest and activity, which are unsuitable for patients with sustained muscle contractions. This pilot study introduces a resting-state muscle fMRI (rs-m-fMRI) approach based on regional homogeneity (ReHo) to evaluate muscle activity from spontaneous BOLD fluctuations during sustained isometric contraction without block-design contrasts. Eight healthy male participants performed separate isometric plantar and dorsal foot flexion tasks during 3&#xa0;T MRI scanning. rs-m-fMRI data were analyzed using ReHo to assess local synchronization of BOLD signal. Calf muscle activation was quantified as the percentage of suprathreshold z-transformed ReHo voxels within each segmented muscle and activation thresholds were derived via ROC analysis. ROC analysis demonstrated moderate discrimination between expected active and inactive muscle regions (AUC&#xa0;=&#xa0;0.63), with sensitivity of 0.57 and specificity of 0.61 at the selected threshold. Consistent condition-related differences were observed between active and inactive muscles during both conditions, with a higher percentage of suprathreshold zReHo voxels in voluntarily contracted muscles. This pilot study demonstrates the feasibility of detecting contraction-related ReHo differences using rs-m-fMRI from a single continuous acquisition. The activation threshold was internally calibrated using expected agonist and antagonist muscle groups and therefore does not represent an externally validated classifier. Further studies incorporating independent physiological validation, reproducibility assessment and larger patient cohorts are required before clinical translation.

Rhythmic motor paradigms are widely used to study sensorimotor timing, yet magnetic resonance imaging (MRI) research has largely focused on central processes, with limited insight into peripheral neuromuscular mechanisms. Motor unit MRI (MUMRI), a motion-sensitive technique in which muscle contraction induces intravoxel water redistribution and transient signal attenuation, enables in&#xa0;vivo visualization of muscle activity. In this study, we developed and validated a combined behavioral-MUMRI paradigm to characterize muscle recruitment during rhythmic foot tapping. Healthy participants performed an auditory-paced tapping task inside an MRI scanner while timing was recorded via an MRI-compatible force transducer and muscle activity was measured using single-slice MUMRI. A variable-latency cueing design systematically sampled the temporal relationship between auditory cues, motor execution, and image acquisition, allowing identification of the optimal latency window for detecting contraction-related signal changes. Fixed-latency acquisitions were then used to assess reproducibility. Behavioral results showed stable performance across conditions, with low variability in tapping accuracy (mean coefficient of variation [CoV]&#x2009;&#x2248;0.078). Transient, localized signal reductions consistent with muscle contraction were observed in anterior lower leg muscles during dorsiflexion. Voxel-wise analyses demonstrated high within-condition reproducibility and latency-dependent spatial patterns, with the greatest average consistency when tapping aligned with scanner rhythm (r&#x2009;&#x2248;0.68). These findings establish a robust framework for integrating rhythmic motor tasks with MUMRI, highlighting the importance of precise temporal alignment for reliable measurement of muscle activity. This approach provides a reproducible method for linking motor behavior to peripheral neuromuscular dynamics and offers potential for advancing both basic and clinical MRI research.

To evaluate the generation of artifacts from intraradicular pins and cementation materials in cone-beam computed tomography images, using different exposure protocols. Sixty extracted single-rooted premolars human were divided into six groups such as G1: Fiberglass pin with Allcem cement, G2: Metal core fiberglass pin core with Allcem cement, G3: metal core fiberglass pin with Allcem Core cement, G4: Fiberglass pin with Allcem Core cement, G5: Fiberglass pin with Panavia cement, and G6: metal core fiberglass pin with Panavia cement. CBCT images were acquired on the Orthophos XG&#xae; 3D (Sirona Dental Systems, Bensheim, Germany) following the protocols with the exposure parameters; 1) high definition (HD) mode (70 kV and 6 mA, voxel size 100 &#xb5;m, scan time 14s, 500 basis images), and 2) non-HD mode (70 kV and 10 mA, voxel size 160 &#xb5;m, scan time 5s, 200 basis images). Two-way analysis of variance (ANOVA) and Tukey HSD post-hoc tests, Fisher's exact test and the McNemar test were performed to detect statistically significant factors. For both HD and non-HD acquisition modes, the percentage of severe artifacts was higher in groups G3 and G6 (P<0.05). In the quantitative analysis, no significant differences were found among the parameters studied after accounting for all variables (P>0.05). The type of fiberglass pin and cement used did not affect the production of the artifacts. When evaluating the scanning modes, the non-HD mode, which had lower mAs and fewer image bases, showed less artifact severity on the assessed CBCT scans.

Linking circuit level activity to large scale functional organization requires imaging methods combining high spatial resolution, broad coverage, and single trial sensitivity. We present volumetric functional ultrasound imaging (3D-fUS) in behaving macaques, enabling imaging of ~1 cm&#xb3; cortical volumes at high spatiotemporal resolution (100 &#xd7; 150 &#xd7; 150 &#x3bc;m&#xb3; voxels, 1.67 Hz). Visually evoked responses were reliably detected at the level of single trials and single voxels, substantially reducing experimental time. To enable model-based analyses analogous to functional magnetic resonance imaging (fMRI), we estimated a fUS hemodynamic response function (fUS-HRF) that was consistent across subjects, cortical areas, and visual stimuli and was well approximated by a gamma function. Compared with fMRI HRFs, the fUS-HRF exhibited faster dynamics, enabling shorter and more closely spaced stimulus presentations. Together, these results establish 3D-fUS as a fast, volumetric, and circuit relevant imaging modality for efficient investigation of distributed cortical dynamics in primates.

Magnetic resonance-guided linear accelerators (MR-Linacs) have recently been introduced for radiotherapy of brain metastases (BMs), including hypofractionated stereotactic radiotherapy (HSRT). However, optimal strategies for planning HSRT within MR-guided adaptive workflows remain to be established. This study aimed to evaluate the influence of field number and minimum monitor unit per segment (MU/segment) on plan quality, and robustness of dose calculations in MR-guided HSRT for patients with solitary medium-sized BMs. This retrospective study included 20 patients who underwent HSRT for solitary medium-sized BMs, receiving a prescription dose of 30 Gy in three fractions. Four intensity-modulated radiotherapy (IMRT) plans were systematically generated for each patient by varying the number of fields (9 vs. 15) and minimum MU/segment (15 vs. 5): 9FL-IMRT (9 fields, 15 MU/segment), 9FS-IMRT (9 fields, 5 MU/segment), 15FL-IMRT (15 fields, 15 MU/segment), and 15FS-IMRT (15 fields, 5 MU/segment). Plan quality, treatment efficiency, and delivery accuracy were assessed based on dose distributions optimized using structure-based bulk electron density assignment, with patient-specific quality assurance (QA) evaluated via global gamma passing rate (GPR) using 3%/2 mm, 2%/2 mm, and 2%/1 mm criteria. To assess robustness against density-related spatial uncertainties, each plan was reoptimized after simulating skull misalignment with random translational offsets within &#xb1;2 mm. These reoptimized plans were recalculated on original computed tomography (CT) datasets using voxel-based electron density as the reference standard. Robustness was quantified by comparing dose distributions between bulk density-based and voxel-based recalculations through GPR analysis using 3%/2 mm, 2%/2 mm, 2%/1 mm, and 1%/1 mm criteria. All four IMRT configurations provided clinically acceptable plans with similar target coverage and no significant differences in conformity, gradient, or homogeneity indices (all P>0.05). Although some dose parameters for normal brain tissue reached statistical significance (overall P<0.001), absolute differences were minor (&#x2264;1 cm3) and clinically irrelevant. The 9FL-IMRT configuration yielded the lowest MU (2,037&#xb1;174) and shortest beam-on time (BOT; 6.69&#xb1;0.92 min), while the 15FL-IMRT configuration required only a minimal increase in BOT (7.59&#xb1;1.18 min) and maintained fewer segments (33&#xb1;7). Delivery accuracy was consistently high across all techniques, with mean GPR values &#x2265;95% at 3%/2 mm, showing no significant inter-technique differences (P=0.67). However, under conditions of simulated skull displacement, notable robustness differences emerged under stricter criteria, with 15FL-IMRT consistently showing the highest GPR at 2%/1 mm (96.02%&#xb1;2.93%, overall P=0.004) and 1%/1 mm (91.87%&#xb1;3.55%, overall P<0.001), significantly outperforming 9FL-IMRT and 9FS-IMRT, indicating enhanced tolerance to density-related spatial uncertainties. The 15FL-IMRT configuration provided an optimal balance among plan quality, deliverability, and robustness of dose calculation, supporting its adoption as the preferred planning approach for MR-guided HSRT in patients with solitary medium-sized BMs.

Episodic ataxia type 2 (EA2) is characterized by episodes of vertigo and ataxia due to mutations in CACNA1A that encodes the &#x3b1;1A subunit of the P/Q-type voltage-gated calcium channel. This study aimed to identify neural correlates of cognitive dysfunction in EA2 by investigating brain atrophy in these patients and determining the relationships between regional brain volumes and intellectual dysfunction. We recruited 12 patients with EA2 (including 6 males; age=30.5&#xb1;14.3 years, mean&#xb1;standard deviation) in 3 university hospitals of South Korea from 2019 to 2023. Regional brain volumes were quantified using voxel-based morphometry and the brainstem-structures feature of FreeSurfer. The results were compared with those for healthy controls. The relationships between regional gray-matter volumes (GMVs) and cognitive function were assessed using voxel-wise multiple regression analyses within the general linear model framework. Brain volumetry revealed a significant decrease in cerebellar volumes, particularly in the vermis (lobules IV, V, and VIII), bilateral flocculi (lobule X), and brainstem. The Full-Scale Intelligence Quotient was positively correlated with the GMVs in the left parahippocampal gyrus, right caudate nucleus, and right cerebellar crus II; the Verbal Comprehension Index was correlated with the GMVs in the bilateral cerebellar crura I and II; and the Processing Speed Index was correlated with the GMV in the right parahippocampal gyrus. Volumetric analyses revealed brain atrophy in patients with EA2 that was correlated with the clinical features observed in this disorder. These findings may further expand the imaging spectrum of disorders associated with CACNA1A mutations, although the identified correlations need to be interpreted with caution.

Purpose To develop a novel, fully automated marker-based method for registering PET images from a small, non-stationary, retrofitted preclinical PET detector to simultaneously acquired MRI images. Methods We manufactured a nose cone tract with geometric markers from a material visible when imaged with a zero-echo-time MRI sequence. A one-time universal calibration determined the relation between the PET image space and the nose cone tract markers. An experiment-specific calibration was needed to determine the location of the nose cone tract markers in laboratory space, and, hence, in MRI image space. The robustness of the method was evaluated through systematic experiments. Results Experiments demonstrated successful registration of the two imaging modalities with a spatial registration error below the PET voxel size. Conclusion The proposed registration approach is a practical method for registering simultaneously acquired PET and MR images with sub-voxel precision when the PET device is non-stationary and the usable space inside the detector is too small for dual-modality fiducial markers.

Prostate cancer (PCa) patients frequently experience biochemical recurrence (BCR) following definitive primary treatment. Although fluorine-18-labeled prostate-specific membrane antigen positron emission tomography/computed tomography (1 8F-PSMA PET/CT) is an imaging modality that is highly sensitive for BCR, false-positive findings owing to benign nonspecific uptake complicate diagnosis. Existing artificial intelligence (AI) tools have attempted to address this challenge but are often limited by their reliance on ground-truth labels derived from expert visual interpretation; thus, these tools reproduce expert opinion rather than confirming disease status. The purpose of this study was to develop a computer-aided diagnosis (CAD) system for classifying benign and malignant findings on 1 8F-PSMA PET/CT imaging in patients with PCa and BCR. Post-therapy imaging follow-up was used as an objective reference standard for malignancy. A dataset of 69 patients with PCa and BCR who underwent 1 8F-PSMA PET/CT imaging was used to develop a CAD. The system was evaluated under two classification schemes based on different ground truths: Task 1 used post-therapy imaging follow-up as the objective reference standard for malignant findings (in a subset of 45 patients), whereas task 2 relied on expert visual interpretation at baseline. In total, after data augmentation and filtering, 334 findings were analyzed for task 1, and 467 findings were analyzed for task 2. Suspicious findings were manually segmented using LifeX software (version 25.06.1). One-dimensional intensity profiles were extracted along the x-, y-, and z-axes at the maximum intensity voxel of each finding, from which profile-based and Gaussian-fit features were derived. The intensity profiles were used directly as inputs to a multilayer perceptron (MLP) classifier and were also used to extract profile- and Gaussian-fit-based features to train a random forest (RF) classifier using feature-importance analysis. The final model incorporated a stacking ensemble architecture combining the MLP and RF base models; logistic regression was the meta-classifier. The model was trained and evaluated using stratified 10-fold cross-validation. In each fold, 90% of the findings were assigned to the training set for model development, including feature selection, and the remaining 10% were held out as an independent test set for performance evaluation. Within each training set, five-fold internal cross-validation was used as the validation procedure for feature selection and stacking. For task 1, the stacking ensemble model achieved an accuracy of 92.5% (SD, 3.5%), sensitivity of 92.3% (SD, 4.6%), specificity of 93.0% (SD, 9.2%), and an area under the receiver operating characteristic curve (AUROC) of 0.97. For task 2, performance was similar: accuracy, 91.7% (SD, 5.2%); sensitivity, 93.8% (SD, 4.5%); specificity, 85.8% (SD, 8.6%); and AUROC, 0.97. Feature-importance analysis revealed that raw intensity magnitude and spatial gradients along the x-axis were the most discriminative features for classification. The proposed CAD system achieved highly accurate classification of 1 8F-PSMA PET/CT findings, leveraging post-therapy imaging follow-up as a more objective reference standard for identifying malignancy than expert-based visual interpretation alone. The system achieved high and consistent performance across both the follow-up-based and expert-based labeling tasks. The integration of this tool into clinical workflow could improve diagnostic confidence and support the personalized management of PCa patients with BCR.

Liver cancer remains a major cause of cancer mortality, and precise CT-based liver tumor segmentation is critical for early diagnosis and personalized treatment. In practice, fully 3D&#x2011;supervised training is limited by prohibitive annotation costs and memory and compute demands. Achieving 3D segmentation from purely 2D supervision is attractive yet challenging due to through-plane context loss, inter&#x2011;slice inconsistency, weak boundaries, and low contrast and noise. We propose LiT&#x2011;WSAG, a 2D&#x2011;trained framework that reconstructs voxel&#x2011;wise 3D masks. Built upon SAM&#x2011;Med2D, it incorporates a wavelet&#x2011;guided dual&#x2011;scale channel adapter (WDSC&#x2011;Adapter) to decouple low- and high-frequency content and enhance multi-scale representations, a lightweight L2.5D neighborhood fusion to recover through&#x2011;plane context at near&#x2011;2D cost, and adversarial training with a ramped GAN weight ('mean-loss' scheduling) to enforce shape and boundary consistency. Extensive experiments demonstrate significant gains over the baseline across multiple datasets. On the LiTS dataset, LiT&#x2011;WSAG achieves slice&#x2011;wise and volumetric DSC of 89.74% and 69.27% (+&#x2009;8.85 and&#x2009;+&#x2009;33.64 percentage points over SAM-Med2D); on the WAW&#x2011;TACE dataset, the corresponding scores are 96.08% and 82.03%. Furthermore, on the Synapse multi-organ dataset, it attains a state-of-the-art average DSC of 88.23%. These results indicate that LiT&#x2011;WSAG delivers robust 3D liver tumor segmentation from 2D supervision, balancing boundary accuracy, noise robustness, and computational efficiency.

To construct and validate a 3D MRI-based cartilage model of the elbow joint by comparing it with laser-scanned models and to assess how cartilage morphology influences intra-articular stress distribution through finite element analysis (FEA). Seven formalin-fixed cadaveric elbows were scanned using high-resolution MRI (voxel size: 0.375 &#xd7; 0.375 &#xd7; 0.400 mm), with sequences selected based on the contrast-to-noise ratio. Cartilage models of the distal humerus, radial head, and proximal ulna were reconstructed and validated against laser-scanned reference models. Validation metrics included cartilage thickness differences, surface-to-surface distances, and Dice similarity coefficients. Subsequently, FEA was performed using models with and without cartilage, under physiological loading, to evaluate the intra-articular stress distribution. MRI-based and laser-scanned models showed similar cartilage thickness distributions, with differences within or close to the imaging resolution. The median cartilage thicknesses measured using the MRI-based and laser-scanned models were 1.15 mm (IQR 0.97 to 1.54) and 1.25 mm (IQR 1.14 to 1.29) in the distal humerus, 1.15 mm (IQR 0.97 to 1.18) and 1.03 mm (IQR 1.01 to 1.14) in the radial head, and 1.06 mm (IQR 0.93 to 1.21) and 1.05 mm (IQR 1.00 to 1.10) in the proximal ulna, respectively. The surface-to-surface root mean square distances were < 0.40 mm, and the Dice coefficients exceeded 0.96. FEA revealed physiologically plausible stress concentration patterns at anatomical contact areas only in models incorporating the cartilage. The 3D MRI-based cartilage models have high geometrical accuracy and biomechanical validity. These models offer a noninvasive tool for assessing cartilage morphology and joint mechanics, and may provide a foundation for future studies investigating the pathophysiology of elbow osteoarthritis.

Current hepatocellular carcinoma (HCC) surveillance guidelines rely on manually defined LI-RADS (Liver Imaging Reporting and Data System) features rather than imaging data analysis. This study evaluates the feasibility of machine-learning (ML)-based image analysis frameworks to identify and localize hepatic parenchyma at elevated risk for HCC. In this retrospective study, cirrhotic patients with HCC diagnosis undergoing MRI between 2008 and 2023 were included. The analysis included negative screening MRI preceding a positive screening MRI confirming a LR-5 lesion within 18 months. Volume-of-interest (VOI) annotations of 'non-malignant' and 'malignant' liver tissue on the screening MRI were manually or automatically placed, mapped from future HCC lesion on positive screening MRI. Radiomics were extracted from these VOIs using PyRadiomics. Logistic regression (LR), random forest (RF), and eXtreme Gradient Boosting (XGB) models were trained and validated across four manual/automatic annotation combinations. Exploratory voxel-level heatmaps were generated to visualize high-risk HCC areas. Model performance was summarized using median values and 95% non-parametric confidence intervals. 121 patients (65&#x2009;&#xb1;&#x2009;9.56; 99 men) were included. The best model performances for LR, RF, and XGB were achieved when training and validating on manual annotations: (AUC [95% CI]: 0.75 [0.72-0.78]; 0.80 [0.78-0.81]; 0.79 [0.77-0.81], respectively). Predictions from LR-3 lesions outperformed regions without visible precursor territory (RF AUC [95% CI]: 0.86 [0.84-0.88] vs. 0.66 [0.63-0.70]). In 36 out of 121 patients (30%), heatmaps showed visually increased probability signals in the region where HCC subsequently became visible on follow-up imaging. The results demonstrate the feasibility of ML-based MRI analysis for identifying liver regions at increased risk for HCC development prior to radiologic diagnosis, particularly in the presence of precursor lesions.

Excessive cerebral iron deposition has been implicated in cognitive dysfunction across several neurological disorders. We evaluated severity-dependent patterns of cerebral iron accumulation in obstructive sleep apnea (OSA) using quantitative susceptibility mapping (QSM) and assessed their potential role in mediating cognitive impairment. The cohort comprised 139 OSA patients, stratified by severity (68 mild-moderate [OSA-M: apnea-hypopnea index (AHI) 5-30 events/hour] and 71 severe [OSA-S: AHI&#x2009;>&#x2009;30 events/hour]), and 48 healthy controls. All underwent polysomnography, Montreal Cognitive Assessment and 3&#xa0;T MRI with multi-echo gradient echo sequences for QSM analysis. Whole-brain voxel-wise comparisons characterized iron deposition patterns. Correlation analysis and mediation models evaluated associations between OSA severity, regional iron content, and cognition. With increasing OSA severity, iron content increased in the bilateral precentral gyri, bilateral medial superior frontal gyri, right putamen and middle cingulate gyrus. Notably, higher QSM values in the left precentral gyrus and right putamen were negatively correlated with cognitive impairment, particularly visuospatial function (p&#x2009;<&#x2009;.05). Mediation analysis demonstrated iron deposition in the right putamen partially mediated associations between AHI, N3 sleep stage proportion, percentage of total sleep time with oxygen saturation&#x2009;<&#x2009;90%, and cognitive function, particularly visuospatial and executive abilities. Cerebral iron overload may contribute to cognitive dysfunction across OSA severity levels. The partial mediating effect of putaminal iron supports chronic intermittent hypoxia-induced iron dysregulation as a potential neuropathological mechanism. These findings, predominantly from a male cohort, identify cerebral iron deposition as a potential therapeutic target for mitigating cognitive decline in OSA.

Accurate preoperative assessment of lymphovascular invasion (LVI) in patients with rectal cancer (RC) is important for guiding postoperative management. This study aimed to develop and validate a deep learning model based on magnetic resonance imaging (MRI)-derived microvascular network simulation parameters for the preoperative assessment of LVI in RC patients. A total of 453 patients with pathologically confirmed rectal adenocarcinoma from two medical centers were retrospectively enrolled. All patients underwent multi-b-value diffusion-weighted imaging (DWI) before surgery. First, a steady-state Navier-Stokes hemodynamic model of the tumor microvascular network was constructed based on the multi-b-value DWI images. Subsequently, voxel-wise least squares fitting was performed to match the DWI signals with the dictionary, enabling the inversion and generation of spatial parametric maps for mean flow velocity (V-m), velocity standard deviation (V-s), and angiogenic branching index (ANB). These parametric maps were then input into a Vision Transformer (ViT) network to extract deep features from each modality. A cross-attention fusion module was designed to capture spatial interactions among the parametric maps and construct a multiparametric fusion model. The model's performance was comprehensively evaluated using the area under the curve (AUC), calibration curves, and decision curve analysis (DCA). The multiparametric fusion model achieved favorable performance, with AUCs of 0.901 [95% confidence interval (CI): 0.808-0.993] and 0.863 (95% CI: 0.800-0.926) in the internal and external validation cohorts, respectively. DCA demonstrated that within the threshold range of 0.2-0.8, the fusion model provided substantially greater clinical net benefit than the individual parameter models. Gradient-weighted Class Activation Mapping (Grad-CAM) visualization revealed that the model's attention was primarily focused on the invasive front of the tumor and regions with high peritumoral vascular density, providing supportive visual evidence and suggesting potential biological relevance. The deep learning model based on MRI-simulated microvascular network parameters provides a promising and noninvasive approach for the preoperative assessment of LVI status in RC patients. The model demonstrated encouraging performance in both internal and external validation cohorts. However, further prospective and multicenter validation is required before clinical application.