• Developed FPQuant, a multi-task deep learning framework for fingerprint phenotyping. • Achieved 97.18% accuracy in six fingerprint pattern classifications. • Demonstrated 98.63% precision in singularity detection with optimized localization. • Identified unreported geographic variations in fingerprint traits. Fingerprint morphology, while evolutionary conserved yet individually distinct, emerges as a pivotal biometric identifier in anthropological research and forensic investigation. Current methodologies for precise identification and quantification of complex morphological features—particularly ridge counting and mean ridge-furrow pairs ridge breadth—remain constrained by labor-intensive and monolithic pattern recognition systems. This study presents FPQuant (Fingerprint Phenomics Quantification), a multi-task deep learning framework integrating the most comprehensive fingerprint pattern classification, singularity detection, and quantification of 12 morphometric phenotypes to date. Leveraging NSPT database of 28,867 expert-curated fingerprints, FPQuant achieved state-of-the-art performance with 97.18% (6-class), 98.62% (5-class), and 98.67% (4-class) pattern classification accuracy; 98.63% precision in topological singularity detection through optimized discrete keypoint localization; and expert-level precision in critical quantitative measurements including ridge counting. Cross-database validation demonstrated extraordinary generalizability with 96.20% of 5-class accuracy on NIST-4 and 97.75% of singularity precision on FVC2002 DB1. Notably, FPQuant’s integrated phenotypic capability revealed uncharacterized geographic variation in six morphometric traits, establishing novel fingerprint morphometric biomarkers for anthropological research. This study creates a scalable technical paradigm that bridging fingerprint phenomics with large-scale population study, while providing potential new research avenues across anthropology, forensics and biometric authentication.

The uterine niche is a localized defect formed at the site of a cesarean scar and is clinically associated with chronic pelvic pain and abnormal bleeding. Accurate segmentation is essential for reliable diagnosis, yet remains challenging due to intrinsic signal artifacts, severe speckle noise, and sparse medical annotations. To address these challenges, we propose a robust ultrasound video segmentation framework that integrates three complementary components. First, an entropy-based progressive complexity curriculum regulates information complexity during training, stabilizing optimization and improving robustness under noisy and data-scarce conditions. Second, a graph Fourier temporal enhancement module injects informative adjacent-frame features and multi-scale temporal context into feature representations, clarifying blurred boundaries and strengthening the delineation of fine anatomical structures. Third, a FlickerSteady loss explicitly constrains inter-frame prediction variations, suppressing temporal flickering and ensuring smooth and continuous segmentation over time. Extensive experiments on clinical ultrasound datasets demonstrate that the proposed framework significantly outperforms existing methods in both spatial accuracy and temporal stability. The source code will be released at https://github.com/Joungjimin/US-Video-MedSAM2.

Deep learning models have been widely investigated for computing and analyzing brain images across various downstream tasks such as disease diagnosis and age regression. Most existing models are tailored for specific tasks and diseases, posing a challenge in developing a foundation model for diverse tasks. This paper proposes a Dense Transformer Foundation Model with Mixture of Experts (DenseFormer-MoE), which integrates dense convolutional network, Vision Transformer and Mixture of Experts (MoE) to progressively learn and consolidate local and global features from T1-weighted magnetic resonance images (sMRI) for multiple tasks including diagnosing multiple brain diseases and predicting brain age. First, a foundation model is built by combining the vision Transformer with Densenet, which are pre-trained with Masked Autoencoder and self-supervised learning to enhance the generalization of feature representations. Then, to mitigate optimization conflicts in multi-task learning, MoE is designed to dynamically select the most appropriate experts for each task. Finally, our method is evaluated on multiple renowned brain imaging datasets including UK Biobank (UKB), Alzheimer's Disease Neuroimaging Initiative (ADNI), and Parkinson's Progression Markers Initiative (PPMI). Experimental results and comparison demonstrate that our method achieves promising performances for prediction of brain age and diagnosis of brain diseases.

Accurate segmentation of brain tumors plays an important role for clinical diagnosis and treatment. Multimodal magnetic resonance imaging (MRI) can provide rich and complementary information for accurate brain tumor segmentation. However, some modalities may be absent in clinical practice. It is still challenging to integrate the incomplete multimodal MRI data for accurate segmentation of brain tumors. In this paper, we propose a brain tumor segmentation method based on multimodal transformer network with incomplete multimodal MRI data. The network is based on U-Net architecture consisting of modality specific encoders, multimodal transformer and multimodal shared-weight decoder. First, a convolutional encoder is built to extract the specific features of each modality. Then, a multimodal transformer is proposed to model the correlations of multimodal features and learn the features of missing modalities. Finally, a multimodal shared-weight decoder is proposed to progressively aggregate the multimodal and multi-level features with spatial and channel self-attention modules for brain tumor segmentation. A missing-full complementary learning strategy is used to explore the latent correlation between the missing and full modalities for feature compensation. For evaluation, our method is tested on the multimodal MRI data from BraTS 2018, BraTS 2019 and BraTS 2020 datasets. The extensive results demonstrate that our method outperforms the state-of-the-art methods for brain tumor segmentation on most subsets of missing modalities.

Deep neural networks have been successfully investigated in the computational analysis of structural magnetic resonance imaging (sMRI) data for the diagnosis of dementia, such as Alzheimer's disease (AD). The disease-related changes in sMRI may be different in local brain regions, which have variant structures but with some correlations. In addition, aging increases the risk of dementia. However, it is still challenging to capture the local variations and long-range correlations of different brain regions and make use of the age information for disease diagnosis. To address these problems, we propose a hybrid network with multi-scale attention convolution and aging transformer for AD diagnosis. First, to capture the local variations, a multi-scale attention convolution is proposed to learn the feature maps with multi-scale kernels, which are adaptively aggregated by an attention module. Then, to model the long-range correlations of brain regions, a pyramid non-local block is used on the high-level features to learn more powerful features. Finally, we propose an aging transformer subnetwork to embed the age information into image features and capture the dependencies between subjects at different ages. The proposed method can learn not only the subject-specific rich features but also the inter-subject age correlations in an end-to-end framework. Our method is evaluated with T1-weighted sMRI scans from a large cohort of subjects on the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. Experimental results demonstrate that our method has achieved promising performance for AD-related diagnosis.

Deep learning has been widely investigated in brain image computational analysis for diagnosing brain diseases such as Alzheimer's disease (AD). Most of the existing methods built end-to-end models to learn discriminative features by group-wise analysis. However, these methods cannot detect pathological changes in each subject, which is essential for the individualized interpretation of disease variances and precision medicine. In this article, we propose a brain status transferring generative adversarial network (BrainStatTrans-GAN) to generate corresponding healthy images of patients, which are further used to decode individualized brain atrophy. The BrainStatTrans-GAN consists of generator, discriminator, and status discriminator. First, a normative GAN is built to generate healthy brain images from normal controls. However, it cannot generate healthy images from diseased ones due to the lack of paired healthy and diseased images. To address this problem, a status discriminator with adversarial learning is designed in the training process to produce healthy brain images for patients. Then, the residual between the generated and input images can be computed to quantify pathological brain changes. Finally, a residual-based multi-level fusion network (RMFN) is built for more accurate disease diagnosis. Compared to the existing methods, our method can model individualized brain atrophy for facilitating disease diagnosis and interpretation. Experimental results on T1-weighted magnetic resonance imaging (MRI) data of 1,739 subjects from three datasets demonstrate the effectiveness of our method.

Brain age is considered as an important biomarker for detecting aging-related diseases such as Alzheimer's Disease (AD). Magnetic resonance imaging (MRI) have been widely investigated with deep neural networks for brain age estimation. However, most existing methods cannot make full use of multimodal MRIs due to the difference in data structure. In this paper, we propose a graph transformer geometric learning framework to model the multimodal brain network constructed by structural MRI (sMRI) and diffusion tensor imaging (DTI) for brain age estimation. First, we build a two-stream convolutional autoencoder to learn the latent representations for each imaging modality. The brain template with prior knowledge is utilized to calculate the features from the regions of interest (ROIs). Then, a multi-level construction of the brain network is proposed to establish the hybrid ROI connections in space, feature and modality. Next, a graph transformer network is proposed to model the cross-modal interaction and fusion by geometric learning for brain age estimation. Finally, the difference between the estimated age and the chronological age is used as an important biomarker for AD diagnosis. Our method is evaluated with the sMRI and DTI data from UK Biobank and Alzheimer's Disease Neuroimaging Initiative database. Experimental results demonstrate that our method has achieved promising performances for brain age estimation and AD diagnosis.

With the advance of medical imaging technologies, multimodal images such as magnetic resonance images (MRI) and positron emission tomography (PET) can capture subtle structural and functional changes of brain, facilitating the diagnosis of brain diseases such as Alzheimer's disease (AD). In practice, multimodal images may be incomplete since PET is often missing due to high financial costs or availability. Most of the existing methods simply excluded subjects with missing data, which unfortunately reduced the sample size. In addition, how to extract and combine multimodal features is still challenging. To address these problems, we propose a deep learning framework to integrate a task-induced pyramid and attention generative adversarial network (TPA-GAN) with a pathwise transfer dense convolution network (PT-DCN) for imputation and classification of multimodal brain images. First, we propose a TPA-GAN to integrate pyramid convolution and attention module as well as disease classification task into GAN for generating the missing PET data with their MRI. Then, with the imputed multimodal images, we build a dense convolution network with pathwise transfer blocks to gradually learn and combine multimodal features for final disease classification. Experiments are performed on ADNI-1/2 datasets to evaluate our method, achieving superior performance in image imputation and brain disease diagnosis compared to state-of-the-art methods.

Fingerprint recognition has been widely investigated and achieved great success for personal recognition. Most of existing fingerprint recognition algorithms can work well on adults but cannot be directly used for children, especially for infants. Compared with adult fingerprints, the size of infant fingerprints is smaller with lower resolution under the same acquisition conditions. In addition, infant fingerprint images suffer from various degradations from the physiological effects and bad collection conditions. Some studies focused on using high-quality and high-resolution sensors to capture infant fingerprints for reliable recognition, which will increase the costs. In this paper, we propose a deep learning based method to perform the super-resolution and enhancement of infant fingerprints by an end-to-end way for more reliable recognition, which is compatible with the existing recognition system. In this method, a dense pyramid convolution neural network is built for joint deep learning of fingerprint super-resolution and enhancement, with a minutia attention block added for more accurate reconstruction of local details. The network is trained with adult fingerprints for image transformation and tested on infant fingerprint dataset. Experimental results show that the proposed method achieves promising improvements for infant fingerprint recognition.

Latent fingerprints are one of the most important evidences used to identify criminals in the law enforcement and forensic agencies. Automated recognition of latent fingerprints is still challenging due to their poor image quality caused by unclear ridge structure and various overlapping patterns. Segmentation and enhancement are important to identify valid fingerprint regions, reduce the noise and improve the clarity of ridge structure for more accurate fingerprint recognition. In this paper, we propose a deep convolutional neural network architecture with the nested UNets for automatic segmentation and enhancement of latent fingerprints. First, to prepare training data, we synthetically generate the latent fingerprints and their segmentation and enhancement ground truth data for training. Then, a deep architecture of nested UNets is proposed to transform low-quality latent image into the segmentation mask and high-quality fingerprint through the pixels-to-pixels and end-to-end training. Finally, the test latent fingerprint is segmented and enhanced with the deep nested UNets to improve the image quality in one shot. The enhancement network is optimized by combining the local and global losses, which not only helps reconstruct the global structure, but also enhance the local ridge details of latent fingerprints. The proposed network can make use of multi-level feature maps in a pyramid way of nested UNets for segmentation and enhancement. Experimental results and comparison on NIST SD27 and IIITD-MOLF latent fingerprint databases demonstrate the promising performance of the proposed method.

Hippocampus is one of the first involved regions in Alzheimer's disease (AD) and mild cognitive impairment (MCI), a prodromal stage of AD. Hippocampal atrophy is a validated, easily accessible, and widely used biomarker for AD diagnosis. Most of existing methods compute the shape and volume features for hippocampus analysis using structural magnetic resonance images (MRI). However, the regions adjacent to hippocampus may be relevant to AD, and the visual features of the hippocampal region are important for disease diagnosis. In this paper, we have proposed a new hippocampus analysis method to combine the global and local features of hippocampus by three-dimensional densely connected convolutional networks and shape analysis for AD diagnosis. The proposed method can make use of the local visual and global shape features to enhance the classification. Tissue segmentation and nonlinear registration are not required in the proposed method. Our method is evaluated with the T1-weighted structural MRIs from 811 subjects including 192 AD, 396 MCI (231 stable MCI and 165 progressive MCI), and 223 normal control in Alzheimer's disease neuroimaging initiative database. Experimental results show the proposed method achieves a classification accuracy of 92.29% and area under the ROC curve of 96.95% for AD diagnosis. Results comparison demonstrates the proposed method performs better than other methods.

Latent fingerprints are captured from the fingerprint impressions left unintentionally at the surfaces of the crime scene. They are often used as an important evidence to identify criminals in law enforcement agencies. Different from the widely used plain and rolled fingerprints, the latent fingerprints are usually of poor quality consisting of complex background with a lot of nonfingerprint patterns and various noises. Latent fingerprint segmentation is an important image processing step to separate fingerprint foreground from background for more accurate and efficient feature extraction and matching. Traditional methods are usually based on the local features such as gray scale variance and gradients, which are sensitive to noise and cannot work well for latent images. This paper proposes a latent fingerprint segmentation method based on combination of ridge density and orientation consistency, which are global and local features of fingerprints, respectively. First, a texture image is obtained by decomposition of latent image with a total variation model. Second, we propose to detect the ridge segments from the texture image, and then compute the density of ridge segments and ridge orientation consistency to characterize the global and local fingerprint patterns. Finally, fingerprint segmentation is performed by combining the ridge density and orientation consistency for latent images. The proposed method has been evaluated on NIST SD27 latent fingerprint database. Experimental results and comparison demonstrate the promising performance of the proposed method.

Latent fingerprint identification plays an important role for identifying and convicting criminals in law enforcement agencies. Latent fingerprint images are usually of poor quality with unclear ridge structure and various overlapping patterns. Although significant advances have been achieved on developing automated fingerprint identification system, it is still challenging to achieve reliable feature extraction and identification for latent fingerprints due to the poor image quality. Prior to feature extraction, fingerprint enhancement is necessary to suppress various noises, and improve the clarity of ridge structures in latent fingerprints. Motivated by the recent success of sparse representation in image denoising, this paper proposes a latent fingerprint enhancement algorithm by combining the total variation model and multiscale patch-based sparse representation. First, the total variation model is applied to decompose the latent fingerprint into cartoon and texture components. The cartoon component with most of the nonfingerprint patterns is removed as the structured noise, whereas the texture component consisting of the weak latent fingerprint is enhanced in the next stage. Second, we propose a multiscale patch-based sparse representation method for the enhancement of the texture component. Dictionaries are constructed with a set of Gabor elementary functions to capture the characteristics of fingerprint ridge structure, and multiscale patch-based sparse representation is iteratively applied to reconstruct high-quality fingerprint image. The proposed algorithm cannot only remove the overlapping structured noises, but also restore and enhance the corrupted ridge structures. In addition, we present an automatic method to segment the foreground of latent image with the sparse coefficients and orientation coherence. Experimental results and comparisons on NIST SD27 latent fingerprint database are presented to show the effectiveness of the proposed algorithm and its superiority over existing algorithms.

Clustering is a useful technique to provide the organization of multimedia database. Using single prototype to represent each cluster may not adequately model the different types of clusters and hence limits the clustering performance on the complex data structure. This paper proposes a clustering algorithm based on multi-prototype representation of cluster. The square-error clustering is used to produce a number of prototypes to locate the regions of high density. The prototypes are organized into a given number of clusters in agglomerative method based on a proposed separation measure. New prototypes are iteratively added to improve the poor cluster boundaries. As a result, the proposed algorithm can discover the clusters of complex structure. Experimental results demonstrate the effectiveness of the proposed clustering algorithm.

This paper presents a front-end filtering algorithm for fingerprint identification, which uses orientation field and dominant ridge distance as retrieval features. We propose a new distance measure that better quantifies the similarity evaluation between two orientation fields than the conventional Euclidean and Manhattan distance measures. Furthermore, fingerprints in the data base are clustered to facilitate a fast retrieval process that avoids exhaustive comparisons of an input fingerprint with all fingerprints in the data base. This makes the proposed approach applicable to large databases. Experimental results on the National Institute of Standards and Technology data base-4 show consistent better retrieval performance of the proposed approach compared to other continuous and exclusive fingerprint classification methods as well as minutia-based indexing schemes