Plant disease recognition remains challenging due to complex imaging conditions, inter-class visual similarity, and the need to deploy accurate models on resource-constrained devices. Although convolutional neural networks (CNNs) are effective at local pattern extraction, their locality bias limits global context modeling. Transformer-based models alleviate this issue, yet many existing designs do not explicitly exploit the complementarity between spatial and channel interactions. To address this limitation, we propose a Dual-Channel Feature Enhancement Network (DC-FEN) as the teacher model, which integrates spatial attention and channel attention to capture lesion-sensitive spatial cues and discriminative channel semantics. To enable lightweight deployment, we further introduce a Category-Guided Knowledge Distillation (CGKD) framework that transfers both category-level semantic knowledge and feature-level relational knowledge from the teacher to a MobileNetV3 student. Experimental results show that the proposed distillation framework consistently improves the performance of the lightweight MobileNetV3 student over its original baseline.

Coronary atherosclerosis underlies life-threatening conditions such as myocardial infarction and stroke, yet its cellular dynamics remain incompletely understood. Here, through single-cell RNA sequencing of 27,941 cells from 56 human coronary segments, we constructed a disease-stage-resolved cellular atlas, revealing pathological remodelling of endothelial cells (ECs) into a progenitor-like state (EC5SLCO4A1+) with low expression of canonical EC dysfunction signatures. EC5SLCO4A1+ abundance increased with atherosclerotic stage, and its emergence is driven by PRDM15 through direct transcriptional activation. Analysis of the EC5SLCO4A1+ interaction network revealed extensive crosstalk with immune cell types, the interaction between which contributed to atherosclerotic progression. Endothelial overexpression of Prdm15 in vivo exacerbated atherosclerosis, while its suppression ameliorated the disease phenotype, with diminished EC5SLCO4A1+-like cells and immune infiltration. Our findings underscore the central role of EC subtype remodelling in the progression of human coronary atherosclerosis and reveal tractable targets for therapeutic intervention.

Multimodal 3D object detection leverages complementary geometric and semantic cues, but integrating depth-completed pseudo-LiDAR with original LiDAR remains challenging due to positional uncertainty and boundary noise. To address these discrepancies, we propose VIM-Net, a progressive framework that models cross-modal consistency across voxel, BEV, and RoI levels. Specifically, the Centroid-guided Retrieval Voxel Fusion (CRVF) module performs voxel-level alignment via centroid-based neighborhood retrieval and position-aware aggregation. The Cross-modal Deformable Attention (CDA) module applies bidirectional deformable attention in Bird’s-Eye View (BEV) space to model contextual correspondence. The Bidirectional Position-aware Cross-modal Fusion (Bi-PCF) module further refines instance-level features by incorporating 3D positional encodings. Experiments on KITTI, nuScenes, and Waymo show that VIM-Net achieves competitive performance compared with state-of-the-art 3D detectors. Notably, it achieves 81.47% Car AP3D on the KITTI Hard test set and provides consistent gains in sparse, long-range, and occluded scenarios.

Plant disease symptoms combine local texture changes with patterns distributed across a leaf, while practical recognition models must remain compact. We introduce DC-FEN, a MobileNetV3-based design that models spatial-token relations and channel interactions in parallel and injects them through gated residual fusion. We also examine output-distribution, direct-feature, and token-relation transfer under same-backbone and heterogeneous teachers. PlantVillage and Plant Pathology 2021 (FGVC8) are evaluated with duplicate-audited, group-aware 70/15/15 splits, an explicit unresolved-leaf sensitivity check, validation-only selection, five training seeds, class-sensitive metrics, and paired seed-wise descriptive summaries. On PlantVillage, the no-additional-attention student, DC-FEN teacher, and DC-FEN joint student obtain macro F1 scores of 96.46±0.91%, 96.90±0.40%, and 96.55±0.25%. On FGVC8, the corresponding scores are 87.29±0.63%, 87.14±0.52%, and 87.20±0.26%. At the prespecified FGVC8 threshold of 0.5, DCAB changed sample-wise F1 by −0.02±0.55 percentage points relative to the unmodified backbone; validation-selected global and label-specific thresholds changed this contrast to +0.28±0.55 and +0.55±0.29 points, while threshold-free macro mAP remained essentially unchanged. A duplicate-audited PlantDoc pressure test reduced frozen-checkpoint accuracy to 30.34±1.10% and 29.57±1.10%, showing that external generalization remains unestablished. A ResNet50 teacher gives logit-only students 97.42±0.51% macro F1 on PlantVillage and 89.82±0.43% sample-wise F1 on FGVC8. After separately weighting the direct and relation terms, the corresponding joint students obtain 97.37±0.56% and 89.94±0.27%, recovering the degradation seen with unit internal weights while remaining close to logit-only transfer. Thus, the study evaluates the benefits and limits of explicit spatial–channel interaction and shows that adding intermediate transfer constraints does not guarantee a stronger student.

Accurate recognition of plant diseases plays a critical role in maintaining agricultural productivity and food security. This study proposes LANet, an innovative network architecture aimed at improving the precision of identifying and categorizing plant diseases from images. LANet consists of two primary components. Firstly, Layer Attention applies self-attention across feature layers at different scales to capture the weights between cross-scale features, addressing the data loss caused by downsampling and enhancing plant species recognition. Secondly, we introduce a transfer learning method, where a plant species classification network is initially trained, its parameters are then frozen, and several convolutional layers are added to train the disease classification network. This allows the model to leverage the learned classification information for more accurate disease recognition. Our model demonstrates an accuracy of 85.62%, outperforming nine other models significantly. Comparative analyses using the public PlantVillage dataset highlight the superior performance of our method in disease recognition.

A star-shaped triazine derivative (T-TPZ) was synthesized, which was multifunctional for the detection of Fe3+ ions and optical power limiting applications. The optical characteristics of T-TPZ were thoroughly analyzed using UV-vis absorption spectra, fluorescence emission spectra, and density functional theory (DFT) calculations. Notably, within the UV-vis range, T-TPZ exhibited several electronic transitions, including n-pi*, 1 pi-pi*, and intramolecular charge transfer (ICT) transitions. Interactions between the amide group of T-TPZ and Fe3+ ions initiated an ICT process resulting in fluorescence quenching, with a detection limit of 2.73 x 10-7 M. Remarkably, the fluorescence intensity decreased by 90% within 30 seconds, highlighting the high selectivity and sensitivity of T-TPZ. Job's plot analysis revealed a coordination stoichiometry of 1 : 3 between T-TPZ and Fe3+ ions. Variations in the T-TPZ concentration within PMMA films were also examined to further explore optical power limiting properties, demonstrating new insights and directions for the development of multifunctional materials.

Accurate detection of objects from LiDAR point clouds is crucial for autonomous driving and environment modeling. However, uncertainties in ground truth labels due to occlusions, sparsity, and truncation can hinder model training and performance. This paper introduces two strategies to address these issues: 1) Soft Regression Loss (SoRL) and 2) Discrete Quantization Sampling (DQS). SoRL utilizes Gaussian distributions for object predictions, measuring uncertainty based on the probability of ground truth labels within these distributions. This method effectively accounts for deviations in object location and orientation. Meanwhile, DQS introduces uncertainty scores for dynamic sample selection, aiming to refine the quality of positive samples for regression. Based on the proposed modules, we design a lightweight multi-stage object detection framework. Notably, these modules can enhance existing 3D object detection methods without affecting significantly inference speeds. Experiments over benchmark datasets show the effectiveness of our method, especially for cars in sparse point clouds.

Recently, 3D object detection from LiDAR point clouds has advanced rapidly. Although the second stage can improve the detection performance significantly, prior works concern little about the essential differences among different stages for the performance enhancement. To address this, this paper proposes a Hierarchical Refinement Network (HRNet) with two novel strategies. Firstly, we build the detection head on multi-scale voxel features to optimize the regression branch progressively with an effective Scale-aware Attentive Propagation (SAP) module. Then, we propose a Dynamic Sample Selection (DSS) module for the recalculation of the IoU during each stage to obtain more balanced positive and negative sample selections. Experiments over benchmark datasets show the effectiveness of our HRNet, particularly for car detection in the sparse point clouds.

Recently, 3-D object detection from LiDAR point clouds has advanced rapidly, but the detection performance still has shortcomings in complex scenarios. In this article, we introduce a novel two-stage 3-D object detector that uses only raw point clouds as input. Specifically, we propose two novel strategies that can effectively enhance the detection performance: the feature interpolation aggregation (FIA) and graph attention pooling (GAP). The former is used to initialize the graph centroid features across different resolutions, which is essential for constructing local graphs. The latter constructs hierarchical local graphs on multiscale voxels to extract region of interest (RoI) features by considering the contributions of neighboring voxels. This enables the aggregation of more effective and robust features for detection. The proposed innovations enhance the detector's feature extraction capability and improve its robustness in complex scenarios. Experimental results on the KITTI and Waymo datasets demonstrate that VoxelGraph-RCNN achieves competitive performance compared with the state-of-the-art detectors. Specifically, the proposed model surpasses the baseline Voxel R-CNN by 0.43%, 1.04%, and 0.87% in detecting objects with easy, moderate, and hard difficulty levels on the KITTI datasets, respectively.

Background:The napkin-ring sign (NRS) was accepted as unstable plaques at coronary computed tomography angiography (CCTA). However, the incidence is relatively low. We sought to assess whether the newly defined diamond-attenuation-sign [DAS, defined as a qualitative plaque feature in a mixed plaque (MP) on CCTA cross-section images by the presence of two features: a visual calcification (in the shape of a diamond) accompanied by an annular-shape lower attenuation plaque tissue surrounding the lumen like a ring], could be accurately identified as unstable atherosclerotic plaques.Methods:Eight heart transplant recipients (8 male; mean age, 48.5±11.6 years; range, 37-65 years) underwent CCTA exams prior to heart transplant surgery. Segment-based CCTA sections were independently evaluated for various plaque patterns including non-calcified plaque (NCP) with NRS (NCP-NRS), NCP without NRS (NCP-non-NRS), MP with DAS (MP-DAS), MP without DAS sign (MP-non-DAS), and calcified plaque (CP).Results:NCP-NRS plaques in 6.4% (23/358), NCP-non-NRS plaques in 24.0% (86/358), MP-DAS plaques in 18.2% (65/358), MP-non-DAS plaques in 20.1% (72/358), and calcified-plaques in 7.0% (25/358) of all cases. The specificity and positive predictive values of the MP-DAS and NCP-NRS signs to identify unstable plaque features were excellent (97.1% vs. 98.6%, 90.8% vs. 87.0%, respectively). DAS plaques were more frequently seen on CCTA exams than that of NRS (39.3% vs. 13.3%, respectively, P=0.001). The diagnostic performance of MP-DAS to identify unstable coronary lesions was superior compared to NCP-NRS [area under the receiver operating characteristic curve (ROC), 0.756; 95% CI: 0.717-0.791 vs. 0.558; 95% CI: 0.514-0.600, respectively, P<0.001].Conclusions:Both the DAS and NRS had a high specificity and positive predictive value for the presence of unstable lesions. DAS was a better identification of unstable atherosclerotic plaques in the assessment of plaque-calcification-pattern (PCP).

To explore whether radiomics-based machine learning (ML) models could outperform conventional diagnostic methods at identifying vulnerable lesions on coronary computed tomographic angiography (CCTA). In this retrospective study, 36 heart transplant recipients with coronary heart disease (CAD) and end-stage heart failure were included. Pathological cross-section samples of 350 plaques were collected and coregistered to patients’ preoperative CCTA images. A total of 1184 radiomic features were extracted from CCTA images. Through feature selection and stratified fivefold cross-validation, we derived eight radiomics-based ML models for lesion vulnerability prediction. An independent set of 196 plaques from another 8 CAD patients who underwent heart transplants was collected to validate radiomics-based ML models’ diagnostic accuracy against conventional CCTA feature-based diagnosis (presence of at least 2 high-risk plaque features). The performance of the prediction models was assessed by the area under the receiver operating characteristic curve (AUC) with 95% confidence intervals (CI). The training group used to develop radiomics-based ML models contained 200/350 (57.1%) vulnerable plaques and the external validation group was composed of 67.3% (132/196) vulnerable plaques. The radiomics-based ML model based on eight radiomic features showed excellent cross-validation diagnostic accuracy (AUC: 0.900 ± 0.033). In the validation group, diagnosis based on conventional CCTA features demonstrated moderate performance (AUC: 0.656 [95% CI: 0.593 –0.718]), while the radiomics-based ML model showed higher diagnostic ability (0.782 [95% CI: 0.710 –0.846]). Radiomics-based ML models showed better diagnostic ability than the conventional CCTA features at assessing coronary plaque vulnerability. • CCTA has great potential in the diagnosis of vulnerable coronary artery lesions. • Radiomics model built through CCTA could discriminate coronary vulnerable lesions in good diagnostic ability. • Radiomics model could improve the ability of vulnerability diagnosis against traditional CCTA method, sensitivity especially.

OBJECTIVE:We sought to distinguish lipid plaques using a CT quantitative pixel density histogram, based on the pathological diagnosis of lipid cores as the gold standard. MATERIALS AND METHODS:Eight patients awaiting heart transplantation due to end-stage coronary heart disease underwent coronary CT angiography (CCTA) spectroscopy prior to heart transplantation; coronary artery pathological analysis was performed for all patients. Lipid-core plaques were defined pathologically as manifesting a lipid core diameter > 200 μm, a circumference > 60 degrees, and a cap thickness < 450 μm. The percentage distributions of CT pixel attenuation ≤ 20, 30, 40, and 50 HU were calculated using quantitative histogram analysis. RESULTS:A total of 271 transverse sections were co-registered between CCTA and pathological analysis. Overall, 26 lipid cores and 16 fibrous plaques were identified by pathological analysis. There was no significant difference in median CT attenuation between the lipid and fibrous plaques (51 HU [interquartile range, 46-63] vs. 57 HU [interquartile range, 50-64], p = 0.659). The median percentage of CT pixel attenuation ≤ 30 HU accounted for 11% (5-17) of lipid-core plaques and 0% (0-2) of fibrous plaques (p < 0.001). The sensitivity and specificity of the method for diagnosing lipid plaques by the average CT pixel attenuation ≤ 30 HU were 80.8% and 87.5%, respectively. The area under the receiver operator characteristics curve was 0.898 (95% confidence interval: 0.765-0.970; 3.0% was the best cut-off value). The diagnostic performance was significantly higher than those of the average pixel CT attenuation percentages ≤ 20, 40, and 50 HU and the mean CT attenuation (p < 0.05). CONCLUSION:In in vivo conditions, with the pathological lipid core as the gold standard, quantification of the percentage of average CT pixel attenuation ≤ 30 HU in the histogram can be useful for accurate identification of lipid plaques.

The application of time-resolved fluorescence spectroscopy (TRFS) to in vivo tissue diagnosis requires a method for fast acquisition of fluorescence decay profiles in multiple spectral bands. This study focusses on development of a clinically compatible fiber-optic based multispectral TRFS (ms-TRFS) system together with validation of its accuracy and precision for fluorescence lifetime measurements. It also presents the expansion of this technique into an imaging spectroscopy method. A tandem array of dichroic beamsplitters and filters was used to record TRFS decay profiles at four distinct spectral bands where biological tissue typically presents fluorescence emission maxima, namely, 390, 452, 542, and 629 nm. Each emission channel was temporally separated by using transmission delays through 200 μm diameter multimode optical fibers of 1, 10, 19, and 28 m lengths. A Laguerre-expansion deconvolution algorithm was used to compensate for modal dispersion inherent to large diameter optical fibers and the finite bandwidth of detectors and digitizers. The system was found to be highly efficient and fast requiring a few nano-Joule of laser pulse energy and <1 ms per point measurement, respectively, for the detection of tissue autofluorescent components. Organic and biological chromophores with lifetimes that spanned a 0.8-7 ns range were used for system validation, and the measured lifetimes from the organic fluorophores deviated by less than 10% from values reported in the literature. Multi-spectral lifetime images of organic dye solutions contained in glass capillary tubes were recorded by raster scanning the single fiber probe in a 2D plane to validate the system as an imaging tool. The lifetime measurement variability was measured indicating that the system provides reproducible results with a standard deviation smaller than 50 ps. The ms-TRFS is a compact apparatus that makes possible the fast, accurate, and precise multispectral time-resolved fluorescence lifetime measurements of low quantum efficiency sub-nanosecond fluorophores.

Cardiovascular disease has long been the number one cause of death and morbidity for adults in developed countries (Roger et al. 2010). Coronary artery disease is the most common type of cardiovascular disease and is caused by the development of atherosclerotic plaques in the coronary arteries of the heart; plaque is composed of lipids, cholesterol, calcium, and other types of cells found circulating in the blood stream (Ross 1993). Heart attack and other acute coronary syndromes are caused by plaque rupture, and although there are oftentimes no precluding physical symptoms to this phenomenon, plaques with a higher risk of rupture are morphologically different than their more stable counterparts and are deemed high-risk or “vulnerable plaques” (Libby 2002; Virmani et al. 2000). ere is currently no method clinically available for assessing risk of plaque rupture; thus, many research efforts are being expended to develop imaging techniques for this purpose. is chapter presents a review of the time-resolved fluorescence techniques applied to characterization and diagnosis of atherosclerosis.

This work reports a multimodal system for label-free tissue diagnosis combining fluorescence lifetime imaging (FLIm), ultrasound backscatter microscopy (UBM), and photoacoustic imaging (PAI). This system provides complementary biochemical, structural and functional features allowing for enhanced in vivo detection of oral carcinoma. Results from a hamster oral carcinoma model (normal, precancer and carcinoma) are presented demonstrating the ability of FLIm to delineate biochemical composition at the tissue surface, UBM and related radiofrequency parameters to identify disruptions in the tissue microarchitecture and PAI to map optical absorption associated with specific tissue morphology and physiology.

ObjectivesUltrasound backscatter microscopy (UBM), or ultrasound biomicroscopy, is a noninvasive, label‐free, and ionizing radiation–free technique allowing high‐resolution 3‐dimensional structural imaging. The goal of this study was to evaluate UBM for resolving anatomic features associated with squamous cell carcinoma of the oral cavity.MethodsThe study was conducted in a hamster buccal pouch model. A carcinogen was topically applied to cheeks of 14 golden Syrian hamsters. Six additional hamsters served as healthy controls. A high‐frequency (41 MHz, 6‐mm focal depth, lateral and axial resolutions of 65 and 37 μm, respectively) UBM system was used for scanning the oral cavity after 14 weeks of carcinogen application. Histologic analyses were conducted on scanned regions.ResultsThe histologic structure of buccal tissue and microvasculature networks could be visualized from the UBM images. Epithelial and mucosal hypertrophy and neoplastic changes were identified in animals subjected to the carcinogen. In animals with invasive squamous cell carcinoma, lesion development and destruction of the structural integrity of tissue layers were noted.ConclusionsIn this pilot study, UBM generated sufficient contrast for morphologic features associated with oral carcinoma compared to healthy tissue. This modality may present a practical technique for detection of oral neoplasms that is potentially translatable to humans.

We report a novel method for estimating fluorescence impulse response function (fIRF) from noise-corrupted time-domain fluorescence measurements of biological tissue. This method is based on the use of high-order Laguerre basis functions and a constrained least-squares approach that addresses the problem of overfitting due to increased model complexity. The new method was extensively evaluated on fluorescence data from simulation, fluorescent standard dyes, ex vivo tissue samples of atherosclerotic plaques and in vivo oral carcinoma. Current results demonstrate that this method allows for rapid and accurate deconvolution of multiple channel fluorescence decays without adaptively adjusting the Laguerre scale parameter. The appropriate choice of the scale parameter is essential for accurate estimation of the fIRF. The method described here is anticipated to play an important role in the development of computational techniques for real-time analysis of time-resolved fluorescence data from biological tissues and to support the advancement of fluorescence lifetime instrumentation for biomedical diagnostics by providing a means for on-line robust analysis of fluorescence decay.

Vaults are naturally-occurring ribonucleoprotein particles found in nearly all eukaryotic cells. They were named for their morphological resemblance to the vaulted ceilings of gothic cathedrals. These ubiquitous nanoparticles are quite abundant with 10(4)-10(6) copies found in the cytoplasm depending on cell type. The structural shell of the particle can self-assemble from 78 copies of a single protein, the major vault protein. This finding has allowed vaults to be bioengineered, resulting in a variety of new functions and capabilities directed toward overcoming many limitations posed by current gene and drug delivery systems. In this study, we demonstrate that recombinant vaults, with the addition of a cell penetration peptide, TAT, can be rapidly delivered to cells in vitro with significantly elevated binding and uptake efficiency. This TAT-vault nanoparticle could be a valuable tool for improving the retention and penetration of therapeutic drugs at tumor sites.

We report the development and validation of an intravascular rotary catheter for bimodal interrogation of arterial pathologies. This is based on a point-spectroscopy scanning time-resolved fluorescence spectroscopy technique enabling reconstruction of fluorescence lifetime images (FLIm) and providing information on arterial intima composition and intravascular ultrasound (IVUS) providing information on arterial wall morphology. The catheter design allows for independent rotation of the ultrasonic and optical channels within an 8 Fr outer diameter catheter sheath and integrates a low volume flushing channel for blood removal in the optical pathways. In the current configuration, the two channels consist of (a) a standard 3 Fr IVUS catheter with single element transducer (40 MHz) and (b) a side-viewing fiber optic (400 μm core). Experiments conducted in tissue phantoms showed the ability of the catheter to operate in an intraluminal setting and to generate coregistered FLIm and IVUS in one pull-back scan. Current results demonstrate the feasibility of the catheter for simultaneous bimodal interrogation of arterial lumen and for generation of robust fluorescence lifetime data under IVUS guidance. These results facilitate further development of a FLIm-IVUS technique for intravascular diagnosis of atherosclerotic cardiovascular diseases including vulnerable plaques.

Tissue diagnostic features generated by a bimodal technique integrating scanning time-resolved fluorescence spectroscopy (TRFS) and ultrasonic backscatter microscopy (UBM) are investigated in an in vivo hamster oral carcinoma model. Tissue fluorescence is excited by a pulsed nitrogen laser and spectrally and temporally resolved using a set of filters/dichroic mirrors and a fast digitizer, respectively. A 41-MHz focused transducer (37-μm axial, 65-μm lateral resolution) is used for UBM scanning. Representative lesions of the different stages of carcinogenesis show that fluorescence characteristics complement ultrasonic features, and both correlate with histological findings. These results demonstrate that TRFS-UBM provide a wealth of co-registered, complementary data concerning tissue composition and structure as it relates to disease status. The direct co-registration of the TRFS data (sensitive to surface molecular changes) with the UBM data (sensitive to cross-sectional structural changes and depth of tumor invasion) is expected to play an important role in pre-operative diagnosis and intra-operative determination of tumor margins.