Knowledge distillation (KD) transfers the ``dark knowledge'' from a complex teacher model to a compact student model. However, heterogeneous architecture distillation, such as Vision Transformer (ViT) to ResNet18, faces challenges due to differences in spatial feature representations. Traditional KD methods are mostly designed for homogeneous architectures and hence struggle to effectively address the disparity. Although heterogeneous KD approaches have been developed recently to solve these issues, they often incur high computational costs and complex designs, or overly rely on logit alignment, which limits their ability to leverage the complementary features. To overcome these limitations, we propose Heterogeneous Complementary Distillation (HCD), a simple yet effective framework that integrates complementary teacher and student features to align representations in shared logits. These logits are decomposed and constrained to facilitate diverse knowledge transfer to the student. Specifically, HCD processes the student’s intermediate features through convolutional projector and adaptive pooling, concatenates them with teacher's feature from the penultimate layer and then maps them via the Complementary Feature Mapper (CFM) module, comprising fully connected layer, to produce shared logits. We further introduce Sub-logit Decoupled Distillation (SDD) that partitions the shared logits into n sub-logits, which are fused with teacher's logits to rectify classification. To ensure sub-logit diversity and reduce redundant knowledge transfer, we propose an Orthogonality Loss (OL). By preserving student-specific strengths and leveraging teacher knowledge, HCD enhances robustness and generalization in students. Extensive experiments on the CIFAR-100, fine-grained (e.g., CUB200, Aircraft) and ImageNet-1K datasets demonstrate that HCD outperforms state-of-the-art KD methods, establishing it as an effective solution for heterogeneous KD.

Remote photoplethysmography (rPPG) provides a convenient solution for contactless physiological measurement, and is a promising technique for daily health monitoring and clinical application. However, traditional supervised methods rely heavily on labeled data, incurring substantial annotation costs. Current unsupervised contrastive learning methods encounter challenges under challenging conditions, including variations in illumination and motion of the head or limbs. In this paper, we propose a novel unsupervised end-to-end framework, called CoDPN for physiological measurement. CoDPN employs a dual-path architecture consisting of the Short-term Information Extraction Path (SIEP) and Long-term Information Extraction Path (LIEP), which capture short-term contextual relevance and long-term periodic dependence of rPPG signals, respectively. Next, we propose a collaborative learning strategy to integrate the latent features from both the SIEP and LIEP, facilitating the exchange of complementary information. Furthermore, our unsupervised learning strategy leverages rPPG features in both the frequency and time domains, guiding the CoDPN to extract rPPG signals consistent with physiological information while reducing dependence on labeled datasets. We conduct extensive experiments on three benchmark datasets (UBFC-rPPG, PURE, and UBFC-phys) to implement physiological measurement, including heart rate (HR), heart rate variability (HRV) and respiratory frequency (RF). The experimental results demonstrate that our CoDPN outperforms other state-of-the-art methods under complex conditions.

Accurate identification and prediction of diabetes complications contribute to improved patient health. However, existing prediction models predominantly employ single-task learning (STL) paradigms, failing to fully leverage the intrinsic correlations among different complications that arise from shared underlying pathophysiological mechanisms, thereby limiting predictive accuracy. To address this, we propose a disease-grouping hierarchical mixed expert model (DGHME). This model integrates clinical-pathological grouping knowledge into a multi-task learning (MTL) architecture, constructing a hierarchical network comprising a bottom-layer self-attention shared expert, group-internal shared experts, task-private experts, and a global shared expert. Through an adaptive gating mechanism and uncertainty-based loss weighting strategy, it achieves refined learning of both disease-common and disease-specific features. Finally, comparative experiments against existing multi-task baseline models on a real-world dataset demonstrate the superiority of the proposed DGHME. Our ablation studies further indicate that DGHME aids in accurately identifying high-risk patients and enables more effective complication prediction.

State-of-the-art logit distillation methods exhibit versatility, simplicity, and efficiency. Despite the advances, existing studies have yet to delve thoroughly into fine-grained relationships within logit knowledge. In this paper, we propose Local Dense Relational Logit Distillation (LDRLD), a novel method that captures inter-class relationships through recursively decoupling and recombining logit information, thereby providing more detailed and clearer insights for student learning. To further optimize the performance, we introduce an Adaptive Decay Weight (ADW) strategy, which can dynamically adjust the weights for critical category pairs using Inverse Rank Weighting (IRW) and Exponential Rank Decay (ERD). Specifically, IRW assigns weights inversely proportional to the rank differences between pairs, while ERD adaptively controls weight decay based on total ranking scores of category pairs. Furthermore, after the recursive decoupling, we distill the remaining non-target knowledge to ensure knowledge completeness and enhance performance. Ultimately, our method improves the student's performance by transferring fine-grained knowledge and emphasizing the most critical relationships. Extensive experiments on datasets such as CIFAR-100, ImageNet-1K, and Tiny-ImageNet demonstrate that our method compares favorably with state-of-the-art logit-based distillation approaches. The code will be made publicly available.

In bone cancer imaging, positron emission tomography (PET) is ideal for the diagnosis and staging of bone cancers due to its high sensitivity to malignant tumors. The diagnosis of bone cancer requires tumor analysis and localization, where accurate and automated wholebody bone segmentation (WBBS) is often needed. Current WBBS for PET imaging is based on paired Computed Tomography (CT) images. However, mismatches between CT and PET images often occur due to patient motion, which leads to erroneous bone segmentation and thus, to inaccurate tumor analysis. Furthermore, there are some instances where CT images are unavailable for WBBS. In this work, we propose a novel multimodal fusion network (MMF-Net) for WBBS of PET images, without the need for CT images. Specifically, the tracer activity ($\lambda$-MLAA), attenuation map ($\mu$-MLAA), and synthetic attenuation map ($\mu$-DL) images are introduced into the training data. We first design a multi-encoder structure employed to fully learn modalityspecific encoding representations of the three PET modality images through independent encoding branches. Then, we propose a multimodal fusion module in the decoder to further integrate the complementary information across the three modalities. Additionally, we introduce revised convolution units, SE (Squeeze-and-Excitation) Normalization and deep supervision to improve segmentation performance. Extensive comparisons and ablation experiments, using 130 whole-body PET image datasets, show promising results. We conclude that the proposed method can achieve WBBS with moderate to high accuracy using PET information only, which potentially can be used to overcome the current limitations of CT-based approaches, while minimizing exposure to ionizing radiation.

The differences between cross-modality medical images are significant, so several studies are working on unsupervised domain adaptation (UDA) segmentation, which aims to adapt a segmentation model trained on a labeled source domain to an unlabeled target domain. The conventional UDA segmentation strategy aims to integrate image generation and segmentation. However, conventional image generation modules only consider information from a single domain (source or target), resulting in visual inconsistencies. The image generation module may also lack anatomical constraints, leading to incorrect pseudo-label generation. To address these issues, we propose an improved symmetric generative adversarial network (ISGAN). Unlike conventional approaches that perform domain adaptation only in the source or target domain, ISGAN adopts a symmetric architecture using two-path domain adaptation to reduce the visual difference. In addition, ISGAN adopts a bidirectional training strategy to optimize the image generation and segmentation modules. The bidirectional training strategy introduces the anatomical constraints into the image generation module, thereby reducing the generation of incorrect pseudo labels. Finally, we validate ISGAN on two cross-modality datasets (the MMWHS cardiac dataset and Abdomen dataset). ISGAN delivers promising segmentation and generalization performance compared with state-of-the-art UDA methods.

Unsupervised domain adaptation (UDA) methods have shown great potential in cross-modality medical image segmentation tasks, where target domain labels are unavailable. However, the domain shift among different image modalities remains challenging, because the conventional UDA methods are based on convolutional neural networks (CNNs), which tend to focus on the texture of images and cannot establish the global semantic relevance of features due to the locality of CNNs. This paper proposes a novel end-to-end Swin Transformer-based generative adversarial network (ST-GAN) for cross-modality cardiac segmentation. In the generator of ST-GAN, we utilize the local receptive fields of CNNs to capture spatial information and introduce the Swin Transformer to extract global semantic information, which enables the generator to better extract the domain-invariant features in UDA tasks. In addition, we design a multi-scale feature fuser to sufficiently fuse the features acquired at different stages and improve the robustness of the UDA network. We extensively evaluated our method with two cross-modality cardiac segmentation tasks on the MS-CMR 2019 dataset and the M&Ms dataset. The results of two different tasks show the validity of ST-GAN compared with the state-of-the-art cross-modality cardiac image segmentation methods.

Abstract Despite their great advancement, current pedestrian detection methods focus on single static images, which fail to employ richer information available from the video sequences. Compared with still images, videos can offer temporal information of objects in the time dimension, thus providing the potential to obtain more robust detection performance. Here, a novel pedestrian detection method based on visible part detection and temporal calibration is proposed. Specifically, a part‐aware module to predict the visible body part of each pedestrian instance, which enables us to obtain precise motion information of partially occluded pedestrians in a video sequence, is first developed. Then, the temporal coherence for each pedestrian instance based on the predicted motion information is constructed. After that, an adaptive temporal calibration method is introduced to effectively calibrate the final detection result. This method on two video pedestrian detection benchmarks, that is, Caltech‐New and MOT17Det, is evaluated. Experimental results show that this method performs favourably against existing pedestrian detection approaches.

A novel temporal convolutional network (TCN) model is utilized to reconstruct the central aortic blood pressure (aBP) waveform from the radial blood pressure waveform. The method does not need manual feature extraction as traditional transfer function approaches. The data acquired by the SphygmoCor CVMS device in 1,032 participants as a measured database and a public database of 4,374 virtual healthy subjects were used to compare the accuracy and computational cost of the TCN model with the published convolutional neural network and bi-directional long short-term memory (CNN-BiLSTM) model. The TCN model was compared with CNN-BiLSTM in the root mean square error (RMSE). The TCN model generally outperformed the existing CNN-BiLSTM model in terms of accuracy and computational cost. For the measured and public databases, the RMSE of the waveform using the TCN model was 0.55 ± 0.40 mmHg and 0.84 ± 0.29 mmHg, respectively. The training time of the TCN model was 9.63 min and 25.51 min for the entire training set; the average test time was around 1.79 ms and 8.58 ms per test pulse signal from the measured and public databases, respectively. The TCN model is accurate and fast for processing long input signals, and provides a novel method for measuring the aBP waveform. This method may contribute to the early monitoring and prevention of cardiovascular disease.

Automatic coronary artery segmentation is of great value in diagnosing coronary disease. In this paper, we propose an automatic coronary artery segmentation method for coronary computerized tomography angiography (CCTA) images based on a deep convolutional neural network. The proposed method consists of three steps. First, to improve the efficiency and effectiveness of the segmentation, a 2D DenseNet classification network is utilized to screen out the non-coronary-artery slices. Second, we propose a coronary artery segmentation network based on the 3D-UNet, which is capable of extracting, fusing and rectifying features efficiently for accurate coronary artery segmentation. Specifically, in the encoding process of the 3D-UNet network, we adapt the dense block into the 3D-UNet so that it can extract rich and representative features for coronary artery segmentation; In the decoding process, 3D residual blocks with feature rectification capability are applied to improve the segmentation quality further. Third, we introduce a Gaussian weighting method to obtain the final segmentation results. This operation can highlight the more reliable segmentation results at the center of the 3D data blocks while weakening the less reliable segmentations at the block boundary when merging the segmentation results of spatially overlapping data blocks. Experiments demonstrate that our proposed method achieves a Dice Similarity Coefficient (DSC) value of 0.826 on a CCTA dataset constructed by us. The code of the proposed method is available at https://github.com/alongsong/3D_CAS.

Pedestrian detection is a crucial task in intelligent transportation systems, which can be applied in autonomous vehicles and traffic scene video surveillance systems. The past few years have witnessed much progress on the research of pedestrian detection methods, especially through the successful use of the deep learning based techniques. However, occlusion and large scale variation remain the challenging issues for pedestrian detection. In this work, we propose a Part-Aware Multi-Scale Fully Convolutional Network (PAMS-FCN) to tackle these difficulties. Specifically, we present a part-aware Region-of-Interest (RoI) pooling module to mine body parts with different responses, and select the part with the strongest response via voting. As such, a partially visible pedestrian instance can receive a high detection confidence score, making it less likely to become a missing detection. This module operates in parallel with an instance RoI pooling module to combine local parts and global context information. To handle vast scale variation, we construct a fully convolutional network in which multi-scale feature maps are generated efficiently, and small-scale and large-scale pedestrians are detected separately. By integrating these structures, the proposed detector achieves the state-of-the-art performance on the Caltech, KITTI, INRIA and ETH pedestrian detection datasets.

The objective of this study was to investigate the measurement instrument-dependent variability in the morphology of the ballistocardiogram (BCG) waveform in human subjects and computational methods to mitigate the variability. The BCG was measured in 22 young healthy subjects using a high-performance force plate and a customized commercial weighing scale under upright standing posture. The timing and amplitude features associated with the major I, J, K waves in the BCG waveforms were extracted and quantitatively analyzed. The results indicated that 1) the I, J, K waves associated with the weighing scale BCG exhibited delay in the timings within the cardiac cycle relative to the ECG R wave as well as attenuation in the absolute amplitudes than the respective force plate counterparts, whereas 2) the time intervals between the I, J, K waves were comparable. Then, two alternative computational methods were conceived in an attempt to mitigate the discrepancy between force plate versus weighing-scale BCG: a transfer function and an amplitude-phase correction. The results suggested that both methods effectively mitigated the discrepancy in the timings and amplitudes associated with the I, J, K waves between the force plate and weighing-scale BCG. Hence, signal processing may serve as a viable solution to the mitigation of the instrument-induced morphological variability in the BCG, thereby facilitating the standardized analysis and interpretation of the timing and amplitude features in the BCG across wide-ranging measurement platforms.

Analysis of the morphology and function of the right ventricle (RV) can be used for the prediction and diagnosis of cardiovascular disease. Accurate description of the structure and function of heart can be provided by analyzing cardiac magnetic resonance imaging (MRI) images. Noise interference and intensity inhomogeneity of MRI images can be addressed by using a local intensity clustering (LIC) model. However, the segmentation of the RV in MRI images still remains a challenge mainly due to its ill-defined borders. To address such a challenge, an algorithm for segmenting the RV based on a local motion intensity clustering (LMIC) model is proposed in this paper. The LMIC model combines the LIC model with the motion intensity information, due to cardiac motion and blood flow. The motion intensity is calculated by using the Lucas Kanade optical flow method and utilized in the LMIC model as an energy parameter. Because the motion intensity of the RV region is stronger than other areas, the RV can be accurately segmented by this approach. Experimental results demonstrate that the LMIC model is able to address the challenge of the ill-defined RV borders in cardiac MRI images and improved RV segmentation accuracy over existing methods.

We propose an online multiple object tracking algorithm that exploits optical flow and convolutional features to handle noisy detections as well as frequent occlusion. To achieve robust tracking, we develop a data association method that deals with tracking scenarios of increasing difficulty. For easy scenarios, we use motion affinity to associate detections with objects. For ambiguous situations, we propose to use an appearance model based on convolutional features and correlation filters to complement template matching methods. For difficult cases where objects are under heavy occlusion, we carry out occlusion analysis, which exploits the relationship between targets and occluders to predict potential object locations. To deal with noisy detections, false positives are detected and removed on both raw detection and tracklet levels, while missing and inaccurate detections are recovered or corrected via short-term tracking. Experimental results on two benchmark datasets demonstrate that the proposed online algorithm performs favorably against the state-of-the-art methods.

Aortic pulse wave reflects cardiovascular status, but, unlike the peripheral pulse wave, is difficult to be measured reliably using noninvasive techniques. Thus, the estimation of aortic pulse wave from peripheral ones is of great significance. This study proposed an adaptive transfer function (ATF) method to estimate the aortic pulse wave from the brachial pulse wave. Aortic and brachial pulse waves were derived from 26 patients who underwent cardiac catheterization. Generalized transfer functions (GTF) were derived based on the autoregressive exogenous model. Then, the GTF was adapted by its peak resonance frequency. And the optional peak resonance frequency for an individual was determined by regression formulas using brachial systolic blood pressure. The method was validated using the leave-one-out cross validation method. Compared with previous studies, the ATF method showed better performance in estimating the aortic pulse wave and predicting the feature parameters. The prediction error of the aortic systolic blood pressure and pulse pressure were 0.2 ± 3.1 and -0.9 ± 3.1 mmHg, respectively. The percentage errors of augmentation index, percentage notch amplitude, and ejection duration were -2.1 ± 32.7%, 12.4 ± 9.2%, and -2.4 ± 3.3%, respectively.

Despite significant progress in pedestrian detection has been made in recent years, detecting pedestrians in crowded scenes remains a challenging problem. In this paper, we propose to use visual contexts based on scale and occlusion cues from detections at proximity to better detect pedestrians for surveillance applications. Specifically, we first apply detectors based on full body and parts to generate initial detections. Scale prior at each image location is estimated using the cues provided by neighboring detections, and the confidence score of each detection is refined according to its consistency with the estimated scale prior. Local occlusion analysis is exploited in refining detection confidence scores which facilitates the final detection cluster based Non-Maximum Suppression. Experimental results on benchmark data sets show that the proposed algorithm performs favorably against the state-of-the-art methods.

Multiple-object tracking is an important task in automated video surveillance. In this paper, we present a multiple-human-tracking approach that takes the single-frame human detection results as input and associates them to form trajectories while improving the original detection results by making use of reliable temporal information in a closed-loop manner. It works by first forming tracklets, from which reliable temporal information is extracted, and then refining the detection responses inside the tracklets, which also improves the accuracy of tracklets' quantities. After this, local conservative tracklet association is performed and reliable temporal information is propagated across tracklets so that more detection responses can be refined. The global tracklet association is done last to resolve association ambiguities. Experimental results show that the proposed approach improves both the association and detection results. Comparison with several state-of-the-art approaches demonstrates the effectiveness of the proposed approach.

In order to assess the compliance of ingestible wireless device (IWD) within safety guidelines, the SAR, and near fields of IWD in two realistic human body models, whose dielectric values are increased from the original by +/-10% and +/- 20% are studied using the finite-difference time-domain method. The radiation characteristics of the IWD in the human body models with changed and unchanged dielectric values are compared. Simulations are carried out at 13 scenarios where the IWD is placed at center positions of abdomens in the two models at the operation frequency of 430, 800, and 1200 MHz, respectively. Results show that variation of radiation intensity near the surface of abdomen is around 2.5, 2.6, and 3.5 dB within 20% variation of dielectric values corresponding to the frequency of 430, 800, and 1200 MHz, respectively. Electric fields in the anterior of the human body models are higher than those in the posterior for all scenarios. SAR values increase with the increase of conductivities of human body tissues, and usually decrease with the increase of relative permittivities of human body tissues. The effect of the dielectric values of human body on SAR is orientation-, human-body-, and frequency-dependent. A variation up to 20% in conductivities and relative permittivities alone or simultaneously always causes a SAR variation less than 10%, 20%, and 30% at the frequency of 430, 800, and 1200 MHz, respectively. As far as the compliance of safety was concerned, the IWD was safe to be used at the input power less than 12.6, 9.3, and 8.4 mW, according to the IEEE safety standards at the frequency of 430, 800, and 1200 MHz, respectively.

This paper compares the radial artery pulses of 105 young graduate students. The radial artery pulses after performing progressive ergometer for five minutes are different from those at rest. All the pulses become floating and fast. The contours of pulses have three kinds of variability. The incisures of 39 subjects become especially low; sometimes the incisures are lower than the onset of pulse waveform. The tidal waves and dicrotic waves of 32 subjects become higher. The pulses of 34 subjects become smooth. Their incisures and dicrotic waves become lower. These changes can instruct the exercise and training of the young students and athletes

The dynamical analysis of pulse variability gives new insight into researches of cardiovascular system's dynamics. Firstly, long-term pulse variability analysis for the researches on cardiovascular system was proposed. Secondly, approximate entropy was applied to analyze three groups of long-term pulse waveform variabilities and we found that the pulses' approximate entropies of patients with cardiovascular disease preferred to smaller value and less irregularity. What more, the pulse variability of the patient with pacemaker newly implanted was also studied. Finally, the pulse variability's clinical value for cardiovascular system was concluded.

The pulse waveform plays an important role in pulse diagnosis, which is the key technique in traditional Chinese medicine. However, its baseline wander introduced in the acquisition process will result in misdiagnosis. Therefore a wavelet based cascade adaptive filter to remove this wander is presented. This cascade adaptive filter works in two stages. The first stage is a discrete Meyer wavelet filter and the second stage is the cubic spline estimation. Compared with some traditional methods, such as cubic spline estimation and linear-phase FIR least-squares error minimization digital filter, the proposed approach has better performance for removing the baseline wander of the pulse waveform.