Dual Attention and Patient Similarity Network for drug recommendation.

MOTIVATION: Artificially making clinical decisions for patients with multi-morbidity has long been considered a thorny problem due to the complexity of the disease. Drug recommendations can assist doctors in automatically providing effective and safe drug combinations conducive to treatment and reducing adverse reactions. However, the existing drug recommendation works ignored two critical information. (i) Different types of medical information and their interrelationships in the patient's visit history can be used to construct a comprehensive patient representation. (ii) Patients with similar disease characteristics and their corresponding medication information can be used as a reference for predicting drug combinations. RESULTS: To address these limitations, we propose DAPSNet, which encodes multi-type medical codes into patient representations through code- and visit-level attention mechanisms, while integrating drug information corresponding to similar patient states to improve the performance of drug recommendation. Specifically, our DAPSNet is enlightened by the decision-making process of human doctors. Given a patient, DAPSNet first learns the importance of patient history records between diagnosis, procedure and drug in different visits, then retrieves the drug information corresponding to similar patient disease states for assisting drug combination prediction. Moreover, in the training stage, we introduce a novel information constraint loss function based on the information bottleneck principle to constrain the learned representation and enhance the robustness of DAPSNet. We evaluate the proposed DAPSNet on the public MIMIC-III dataset, our model achieves relative improvements of 1.33%, 1.20% and 2.03% in Jaccard, F1 and PR-AUC scores, respectively, compared to state-of-the-art methods. AVAILABILITY AND IMPLEMENTATION: The source code is available at the github repository: https://github.com/andylun96/DAPSNet.

Experimental demonstration of wavefront reconstruction and correction techniques for variable targets based on distorted grating and deep learning.

This research presents a practical approach for wavefront reconstruction and correction adaptable to variable targets, with the aim of constructing a high-precision, general extended target adaptive optical system. Firstly, we delve into the detailed design of a crucial component, the distorted grating, simplifying the optical system implementation while circumventing potential issues in traditional phase difference-based collection methods. Subsequently, normalized fine features (NFFs) and structure focus features (SFFs) which both are independent of the imaging target but corresponded precisely to the wavefront aberration are proposed. The two features provide a more accurate and robust characterization of the wavefront aberrations. Then, a Noise-to-Denoised Generative Adversarial Network (N2D-GAN) is employed for denoising real images. And a lightweight network, Attention Mechanism-based Efficient Network (AM-EffNet), is applied to achieve efficient and high-precision mapping between features and wavefronts. A prototype of object-independent adaptive optics system is demonstrated by experimental setup, and the effectiveness of this method in wavefront reconstruction for different imaging targets has been verified. This research holds significant relevance for engineering applications of adaptive optics, providing robust support for addressing challenges within practical systems.

Multi-line-of-sight Hartmann-Shack wavefront sensing based on image segmentation and K-means sorting.

Multi-line-of-sight wavefront sensing, crucial for next-generation astronomy and laser applications, often increases system complexity by adding sensors. This research introduces, to the best of our knowledge, a novel method for multi-line-of-sight Hartmann-Shack wavefront sensing by using a single sensor, addressing challenges in centroid estimation and classification under atmospheric turbulence. This method contrasts with existing techniques that rely on multiple sensors, thereby reducing system complexity. Innovations include combining edge detection and peak extraction for precise centroid calculation, improved k-means clustering for robust centroid classification, and a centroid filling algorithm for subapertures with light loss. The method's effectiveness was confirmed through simulations for a five-line-of-sight system and experimental setup for two-line and three-line-of-sight systems, demonstrating its potential in real atmospheric aberration correction conditions. Experimental findings indicate that, when implemented in a closed-loop configuration, the method significantly reduces wavefront residuals from 1 λ to 0.1 λ under authentic atmospheric turbulence conditions. Correspondingly, the quality of the far-field spot is enhanced by a factor of 2 to 4. These outcomes collectively highlight the method's robust capability in enhancing optical system performance in environments characterized by genuine atmospheric turbulence.

Performance of the neural network-based prediction model in closed-loop adaptive optics.

Adaptive optics (AO) technology is an effective means to compensate for atmospheric turbulence, but the inherent delay error of an AO system will cause the compensation phase of the deformable mirror (DM) to lag behind the actual distortion, which limits the correction performance of the AO technology. Therefore, the feed-forward prediction of atmospheric turbulence has important research value and application significance to offset the inherent time delay and improve the correction bandwidth of the AO system. However, most prediction algorithms are limited to an open-loop system, and the deployment and the application in the actual AO system are rarely reported, so its correction performance improvement has not been verified in practice. We report, to our knowledge, the first successful test of a deep learning-based spatiotemporal prediction model in an actual 3 km laser atmospheric transport AO system and compare it with the traditional closed-loop control methods, demonstrating that the AO system with the prediction model has higher correction performance.

Respiration-Responsive Colorful Room-Temperature Phosphorescent Materials and Assembly-Induced Phosphorescence Enhancement Strategies.

It is still very challenging to obtain colorful and long-afterglow room-temperature phosphorescent (RTP) materials from pure organic polymers. Herein, it is found that chitosan (CS), a natural polymer, not only has its own RTP, but also reacts with different phosphorescent molecules to obtain a multicolor, long-afterglow RTP material. CS can emit RTP with a lifetime of 48 ms. In addition, CS is rich in amino groups, and grafting different phosphorescent molecules onto CS by an amidation reaction can modulate it to emit different colors of phosphorescence and obtain a series of colorful CS derivatives. The obtained polymer films also have ultra-long RTP due to the good film-forming ability. In addition, one of the CS derivatives selected with α-cyclodextrin is used to construct RTP materials with lifetimes of up to seconds. The host-guest interactions are used to suppress nonradiative relaxation and build crystalline domains, thus synergistically enhancing the RTP. Interestingly, the RTP properties of the CS derivative films are extremely sensitive to water and heat stimuli, because water broke the hydrogen bonds between adjacent CS molecules and thus altered the rigid environment in the material. Finally, they can be used as a stimuli-responsive ink and for monitoring environmental humidity.

Object-independent wavefront sensing method based on an unsupervised learning model for overcoming aberrations in optical systems.

This Letter introduces the idea of unsupervised learning into object-independent wavefront sensing for the first time, to the best of our knowledge, which can achieve fast phase recovery of arbitrary objects without labels. First, a fine feature extraction method which only depends on the wavefront aberrations is proposed. Then, a lightweight neural network and an optical feature system are combined to form an unsupervised learning model, and the neural network is promoted to be well trained by reversely outputting fine features. Simulation results prove that the proposed method can effectively overcome the aberrations (static or variable) existing in the optical system and achieve wavefront sensing of different objects with high precision and efficiency.

Crystal Growth of Defect-Free Single-Crystalline Fibers in Super-High Supersaturation Solution.

By controlling the supersaturation and choosing high-quality seeds, we successfully suppress the prismatic growth of the tetragonal potassium dihydrogen phosphate (KDP) SCFs, realize the rapid growth along the [001] direction, and obtain SCFs less than 10 µm in width with lengths of centimeters. The experimental results show that there exist critical supersaturation points, 22.40% and 41.41% at 25 °C, for initiating the growth of KDP SCF on its pyramidal and prismatic faces, respectively, which are quite different from those of the bulk crystals. We use the mechanism of 2D nucleation on smooth faces to explain the peculiar phenomena, assuming that there is no 2D or 3D defect on the surfaces of the seed fiber crystal. The assumption is supported by AFM observation of the surface micromorphology of the SCFs. Our solution growth technique developed can be used to grow ultrafine SCFs unable to be achieved by existing melt growth techniques.

Highly Stable Spatio-Temporal Prediction Network of Wavefront Sensor Slopes in Adaptive Optics.

Adaptive Optics (AO) technology is an effective means to compensate for wavefront distortion, but its inherent delay error will cause the compensation wavefront on the deformable mirror (DM) to lag behind the changes in the distorted wavefront. Especially when the change in the wavefront is higher than the Shack-Hartmann wavefront sensor (SHWS) sampling frequency, the multi-frame delay will seriously limit its correction performance. In this paper, a highly stable AO prediction network based on deep learning is proposed, which only uses 10 frames of prior wavefront information to obtain high-stability and high-precision open-loop predicted slopes for the next six frames. The simulation results under various distortion intensities show that the prediction accuracy of six frames decreases by no more than 15%, and the experimental results also verify that the open-loop correction accuracy of our proposed method under the sampling frequency of 500 Hz is better than that of the traditional non-predicted method under 1000 Hz.

Dual-Role of Polyelectrolyte-Tethered Benzimidazolium Cation in Promoting CO2 /Pure Water Co-Electrolysis to Ethylene.

Electrochemical CO2 reduction reaction (CO2 RR), as a promising route to realize negative carbon emissions, is known to be strongly affected by electrolyte cations (i.e., cation effect). In contrast to the widely-studied alkali cations in liquid electrolytes, the effect of organic cations grafted on alkaline polyelectrolytes (APE) remains unexplored, although APE has already become an essential component of CO2 electrolyzers. Herein, by studying the organic cation effect on CO2 RR, we find that benzimidazolium cation (Beim+ ) significantly outperforms other commonly-used nitrogenous cations (R4 N+ ) in promoting C2+ (mainly C2 H4 ) production over copper electrode. Cyclic voltammetry and in situ spectroscopy studies reveal that the Beim+ can synergistically boost the CO2 to *CO conversion and reduce the proton supply at the electrocatalytic interface, thus facilitating the *CO dimerization toward C2+ formation. By utilizing the homemade APE ionomer, we further realize efficient C2 H4 production at an industrial-scale current density of 331 mA cm-2 from CO2 /pure water co-electrolysis, thanks to the dual-role of Beim+ in synergistic catalysis and ionic conduction. This study provides a new avenue to boost CO2 RR through the structural design of polyelectrolytes.

Distortion correction for particle image velocimetry using multiple-input deep convolutional neural network and Hartmann-Shack sensing.

Aberrations degrade the accuracy of quantitative, imaging-based measurements, like particle image velocimetry (PIV). Adaptive optical elements can in principle correct the wavefront distortions, but are limited by their technical specifications. Here we propose an actuator-free correction based on a multiple-input deep convolutional neural network which uses an additional input from a wavefront sensor to correct time-varying distortions. It is applied for imaging flow velocimetry to conduct measurements through a fluctuating air-water phase boundary. Dataset for neural network is generated by an experimental setup with a deformable mirror. Correction performance of trained model is estimated in terms of image quality, which is improved significantly, and flow measurement results, where the errors induced by the distortion from fluctuating phase boundary can be corrected by 82 %. The technique has the potential to replace classical closed-loop adaptive optical systems where the performance of the actuators is not sufficient.

Anti-Myocardial Infarction Effects of Radix Aconiti Lateralis Preparata Extracts and Their Influence on Small Molecules in the Heart Using Matrix-Assisted Laser Desorption/Ionization-Mass Spectrometry Imaging.

Radix Aconiti Lateralis Preparata (fuzi) is the processed product of Aconitum carmichaelii Debeaux tuber, and has great potential anti-myocardial infarction effects, including improving myocardial damage and energy metabolism in rats. However, the effects of Radix Aconiti Lateralis Preparata extracts in a rat model of myocardial infarction have not yet been fully illustrated. Herein, Radix Aconiti Lateral Preparata was used to prepare Radix Aconiti Lateralis Preparata extract (RAE), fuzi polysaccharides (FPS), and fuzi total alkaloid (FTA). Then, we aimed to compare the effects of RAE, FPS, and FTA in MI rats and further explore their influence on small molecules in the heart. We reported that Radix Aconiti Lateralis Preparata extract (RAE) and fuzi total alkaloid (FTA) significantly improved left ventricular function and structure, and reduced myocardial damage and infarct size in rats with myocardial infarction by the left anterior descending artery ligation. In contrast, fuzi polysaccharides (FPS) was less effective than RAE and FTA, indicating that alkaloids might play a major role in the treatment of myocardial infarction. Moreover, via matrix-assisted laser desorption/ionization-mass spectrometry imaging (MALDI-MSI), we further showed that RAE and FTA containing alkaloids as the main common components regulated myocardial energy metabolism-related molecules and phospholipids levels and distribution patterns against myocardial infarction. In particular, it was FTA, not RAE, that could also regulate potassium ions and glutamine to play a cardioprotective role in myocardial infarction, which revealed that an appropriate dose of alkaloids generated more obvious cardiotonic effects. These findings together suggested that Radix Aconiti Lateralis Preparata extracts containing an appropriate dose of alkaloids as its main pharmacological active components exerted protective effects against myocardial infarction by improving myocardial energy metabolism abnormalities and changing phospholipids levels and distribution patterns to stabilize the cardiomyocyte membrane structure. Thus, RAE and FTA extracted from Radix Aconiti Lateralis Preparata are potential candidates for the treatment of myocardial infarction.

Environmental coherence and excitation effects in non-Markovian dynamics.

We study the non-Markovian dynamics of a qubit system coupled respectively to coherent state, squeezing vacuum state, and mixed state environments through dephasing interaction. Special attention is paid to the problem of environmental coherence and excitation on the effect of non-Markovianity of system dynamics. Some nontrivial and unexpected results are found. The number of environmental excitations serves to enhance the non-Markovianity of system dynamics, but the enhancement slows down with the increasing of the variance of excitation number. However, environmental coherence can play dual effects, which enhances in some cases and suppresses in other cases the non-Markovianity of system dynamics.

Multiple serous cavity effusion screening based on smear images using vision transformer.

Serous cavity effusion is a prevalent pathological condition encountered in clinical settings. Fluid samples obtained from these effusions are vital for diagnostic and therapeutic purposes. Traditionally, cytological examination of smears is a common method for diagnosing serous cavity effusion, renowned for its convenience. However, this technique presents limitations that can compromise its efficiency and diagnostic accuracy. This study aims to overcome these challenges and introduce an improved method for the precise detection of malignant cells in serous cavity effusions. We have developed a transformer-based classification framework, specifically employing the vision transformer (ViT) model, to fulfill this objective. Our research involved collecting smear images and corresponding cytological reports from 161 patients who underwent serous cavity drainage. We meticulously annotated 4836 patches from these images, identifying regions with and without malignant cells, thus creating a unique dataset for smear image classification. The findings of our study reveal that deep learning models, particularly the ViT model, exhibit remarkable accuracy in classifying patches as malignant or non-malignant. The ViT model achieved an impressive area under the receiver operating characteristic curve (AUROC) of 0.99, surpassing the performance of the convolutional neural network (CNN) model, which recorded an AUROC of 0.86. Additionally, we validated our models using an external cohort of 127 patients. The ViT model sustained its high-level screening performance, achieving an AUROC of 0.98 at the patient level, compared to the CNN model's AUROC of 0.84. The visualization of our ViT models confirmed their capability to precisely identify regions containing malignant cells in multiple serous cavity effusion smear images. In summary, our study demonstrates the potential of deep learning models, particularly the ViT model, in automating the screening process for serous cavity effusions. These models offer significant assistance to cytologists in enhancing diagnostic accuracy and efficiency. The ViT model stands out for its advanced self-attention mechanism, making it exceptionally suitable for tasks that necessitate detailed analysis of small, sparsely distributed targets like cellular clusters in serous cavity effusions.

Spiking Neural Network for Ultralow-Latency and High-Accurate Object Detection.

Spiking Neural Networks (SNNs) have attracted significant attention for their energy-efficient and brain-inspired event-driven properties. Recent advancements, notably Spiking-YOLO, have enabled SNNs to undertake advanced object detection tasks. Nevertheless, these methods often suffer from increased latency and diminished detection accuracy, rendering them less suitable for latency-sensitive mobile platforms. Additionally, the conversion of artificial neural networks (ANNs) to SNNs frequently compromises the integrity of the ANNs' structure, resulting in poor feature representation and heightened conversion errors. To address the issues of high latency and low detection accuracy, we introduce two solutions: timestep compression and spike-time-dependent integrated (STDI) coding. Timestep compression effectively reduces the number of timesteps required in the ANN-to-SNN conversion by condensing information. The STDI coding employs a time-varying threshold to augment information capacity. Furthermore, we have developed an SNN-based spatial pyramid pooling (SPP) structure, optimized to preserve the network's structural efficacy during conversion. Utilizing these approaches, we present the ultralow latency and highly accurate object detection model, SUHD. SUHD exhibits exceptional performance on challenging datasets like PASCAL VOC and MS COCO, achieving a remarkable reduction of approximately 750 times in timesteps and a 30% enhancement in mean average precision (mAP) compared to Spiking-YOLO on MS COCO. To the best of our knowledge, SUHD is currently the deepest spike-based object detection model, achieving ultralow timesteps for lossless conversion.

Biodegradable MXene Quantum Dots with High Near-Infrared Photothermal Performance for Cancer Treatment.

Photothermal therapy (PTT) offers significant potential in cancer treatment due to its short, simple, and less harmful nature. However, obtaining a photothermal agent (PTA) with good photothermal performance and biocompatibility remains a challenge. MXenes, which are PTAs, have shown promising results in cancer treatment. This study presents the preparation of Ti3C2 MXene quantum dots (MXene QDs) using a simple hydrothermal and ultrasonic method and their use as a PTA for cancer treatment. Compared to conventional MXene QDs synthesized using only the hydrothermal method, the ultrasonic process increased the degree of oxidation on the surface of the MXene QDs. This resulted in the presence of more hydrophilic groups such as hydroxyl groups on the MXene QD surfaces, leading to excellent dispersion in the aqueous system and biocompatibility of the prepared MXene QDs without the need for surface modification. The MXene QDs showed great photothermal performance with a photothermal conversion efficiency of 62.5%, resulting in the highest photothermal conversion efficiency among similar materials reported thus far. Both in vitro and in vivo experiments have proved the potent tumor inhibitory effect of the MXene QD-mediated PTT, with minimal harm to mice. Therefore, these MXene QDs hold a significant promise for clinical applications.

Prophylactic Supplementation with Lactobacillus Reuteri or Its Metabolite GABA Protects Against Acute Ischemic Cardiac Injury.

The gut microbiome has emerged as a potential target for the treatment of cardiovascular disease. Ischemia/reperfusion (I/R) after myocardial infarction is a serious complication and whether certain gut bacteria can serve as a treatment option remains unclear. Lactobacillus reuteri (L. reuteri) is a well-studied probiotic that can colonize mammals including humans with known cholesterol-lowering properties and anti-inflammatory effects. Here, the prophylactic cardioprotective effects of L. reuteri or its metabolite γ-aminobutyric acid (GABA) against acute ischemic cardiac injury caused by I/R surgery are demonstrated. The prophylactic gavage of L. reuteri or GABA confers cardioprotection mainly by suppressing cardiac inflammation upon I/R. Mechanistically, GABA gavage results in a decreased number of proinflammatory macrophages in I/R hearts and GABA gavage no longer confers any cardioprotection in I/R hearts upon the clearance of macrophages. In vitro studies with LPS-stimulated bone marrow-derived macrophages (BMDM) further reveal that GABA inhibits the polarization of macrophages toward the proinflammatory M1 phenotype by inhibiting lysosomal leakage and NLRP3 inflammasome activation. Together, this study demonstrates that the prophylactic oral administration of L. reuteri or its metabolite GABA attenuates macrophage-mediated cardiac inflammation and therefore alleviates cardiac dysfunction after I/R, thus providing a new prophylactic strategy to mitigate acute ischemic cardiac injury.

MG-Trans: Multi-Scale Graph Transformer With Information Bottleneck for Whole Slide Image Classification.

Multiple instance learning (MIL)-based methods have become the mainstream for processing the megapixel-sized whole slide image (WSI) with pyramid structure in the field of digital pathology. The current MIL-based methods usually crop a large number of patches from WSI at the highest magnification, resulting in a lot of redundancy in the input and feature space. Moreover, the spatial relations between patches can not be sufficiently modeled, which may weaken the model's discriminative ability on fine-grained features. To solve the above limitations, we propose a Multi-scale Graph Transformer (MG-Trans) with information bottleneck for whole slide image classification. MG-Trans is composed of three modules: patch anchoring module (PAM), dynamic structure information learning module (SILM), and multi-scale information bottleneck module (MIBM). Specifically, PAM utilizes the class attention map generated from the multi-head self-attention of vision Transformer to identify and sample the informative patches. SILM explicitly introduces the local tissue structure information into the Transformer block to sufficiently model the spatial relations between patches. MIBM effectively fuses the multi-scale patch features by utilizing the principle of information bottleneck to generate a robust and compact bag-level representation. Besides, we also propose a semantic consistency loss to stabilize the training of the whole model. Extensive studies on three subtyping datasets and seven gene mutation detection datasets demonstrate the superiority of MG-Trans.

A semi-supervised multi-task learning framework for cancer classification with weak annotation in whole-slide images.

Cancer region detection (CRD) and subtyping are two fundamental tasks in digital pathology image analysis. The development of data-driven models for CRD and subtyping on whole-slide images (WSIs) would mitigate the burden of pathologists and improve their accuracy in diagnosis. However, the existing models are facing two major limitations. Firstly, they typically require large-scale datasets with precise annotations, which contradicts with the original intention of reducing labor effort. Secondly, for the subtyping task, the non-cancerous regions are treated as the same as cancerous regions within a WSI, which confuses a subtyping model in its training process. To tackle the latter limitation, the previous research proposed to perform CRD first for ruling out the non-cancerous region, then train a subtyping model based on the remaining cancerous patches. However, separately training ignores the interaction of these two tasks, also leads to propagating the error of the CRD task to the subtyping task. To address these issues and concurrently improve the performance on both CRD and subtyping tasks, we propose a semi-supervised multi-task learning (MTL) framework for cancer classification. Our framework consists of a backbone feature extractor, two task-specific classifiers, and a weight control mechanism. The backbone feature extractor is shared by two task-specific classifiers, such that the interaction of CRD and subtyping tasks can be captured. The weight control mechanism preserves the sequential relationship of these two tasks and guarantees the error back-propagation from the subtyping task to the CRD task under the MTL framework. We train the overall framework in a semi-supervised setting, where datasets only involve small quantities of annotations produced by our minimal point-based (min-point) annotation strategy. Extensive experiments on four large datasets with different cancer types demonstrate the effectiveness of the proposed framework in both accuracy and generalization.

A Structure-Aware Hierarchical Graph-Based Multiple Instance Learning Framework for pT Staging in Histopathological Image.

Pathological primary tumor (pT) stage focuses on the infiltration degree of the primary tumor to surrounding tissues, which relates to the prognosis and treatment choices. The pT staging relies on the field-of-views from multiple magnifications in the gigapixel images, which makes pixel-level annotation difficult. Therefore, this task is usually formulated as a weakly supervised whole slide image (WSI) classification task with the slide-level label. Existing weakly-supervised classification methods mainly follow the multiple instance learning paradigm, which takes the patches from single magnification as the instances and extracts their morphological features independently. However, they cannot progressively represent the contextual information from multiple magnifications, which is critical for pT staging. Therefore, we propose a structure-aware hierarchical graph-based multi-instance learning framework (SGMF) inspired by the diagnostic process of pathologists. Specifically, a novel graph-based instance organization method is proposed, namely structure-aware hierarchical graph (SAHG), to represent the WSI. Based on that, we design a novel hierarchical attention-based graph representation (HAGR) network to capture the critical patterns for pT staging by learning cross-scale spatial features. Finally, the top nodes of SAHG are aggregated by a global attention layer for bag-level representation. Extensive studies on three large-scale multi-center pT staging datasets with two different cancer types demonstrate the effectiveness of SGMF, which outperforms state-of-the-art up to 5.6% in the F1 score.

Childhood Leukemia Classification via Information Bottleneck Enhanced Hierarchical Multi-Instance Learning.

Leukemia classification relies on a detailed cytomorphological examination of Bone Marrow (BM) smear. However, applying existing deep-learning methods to it is facing two significant limitations. Firstly, these methods require large-scale datasets with expert annotations at the cell level for good results and typically suffer from poor generalization. Secondly, they simply treat the BM cytomorphological examination as a multi-class cell classification task, thus failing to exploit the correlation among leukemia subtypes over different hierarchies. Therefore, BM cytomorphological estimation as a time-consuming and repetitive process still needs to be done manually by experienced cytologists. Recently, Multi-Instance Learning (MIL) has achieved much progress in data-efficient medical image processing, which only requires patient-level labels (which can be extracted from the clinical reports). In this paper, we propose a hierarchical MIL framework and equip it with Information Bottleneck (IB) to tackle the above limitations. First, to handle the patient-level label, our hierarchical MIL framework uses attention-based learning to identify cells with high diagnostic values for leukemia classification in different hierarchies. Then, following the information bottleneck principle, we propose a hierarchical IB to constrain and refine the representations of different hierarchies for better accuracy and generalization. By applying our framework to a large-scale childhood acute leukemia dataset with corresponding BM smear images and clinical reports, we show that it can identify diagnostic-related cells without the need for cell-level annotations and outperforms other comparison methods. Furthermore, the evaluation conducted on an independent test cohort demonstrates the high generalizability of our framework.