DECNet: Dense embedding contrast for unsupervised semantic segmentation.

Unsupervised semantic segmentation is important for understanding that each pixel belongs to known categories without annotation. Recent studies have demonstrated promising outcomes by employing a vision transformer backbone pre-trained on an image-level dataset in a self-supervised manner. However, those methods always depend on complex architectures or meticulously designed inputs. Naturally, we are attempting to explore the investment with a straightforward approach. To prevent over-complication, we introduce a simple Dense Embedding Contrast network (DECNet) for unsupervised semantic segmentation in this paper. Specifically, we propose a Nearest Neighbor Similarity strategy (NNS) to establish well-defined positive and negative pairs for dense contrastive learning. Meanwhile, we optimize a contrastive objective named Ortho-InfoNCE to alleviate the false negative problem inherent in contrastive learning for further enhancing dense representations. Finally, extensive experiments conducted on COCO-Stuff and Cityscapes datasets demonstrate that our approach outperforms state-of-the-art methods.

Advances in Machine Learning for Wearable Sensors.

Recent years have witnessed tremendous advances in machine learning techniques for wearable sensors and bioelectronics, which play an essential role in real-time sensing data analysis to provide clinical-grade information for personalized healthcare. To this end, supervised learning and unsupervised learning algorithms have emerged as powerful tools, allowing for the detection of complex patterns and relationships in large, high-dimensional data sets. In this Review, we aim to delineate the latest advancements in machine learning for wearable sensors, focusing on key developments in algorithmic techniques, applications, and the challenges intrinsic to this evolving landscape. Additionally, we highlight the potential of machine-learning approaches to enhance the accuracy, reliability, and interpretability of wearable sensor data and discuss the opportunities and limitations of this emerging field. Ultimately, our work aims to provide a roadmap for future research endeavors in this exciting and rapidly evolving area.

Stratification of Group A Streptococcal Pharyngitis Children Using Unsupervised Learning.

Background and objectives Group A Streptococcus (GAS) is the most frequent cause of bacterial pharyngitis, and it is advised to selectively use rapid antigen detection testing (RADT). Currently, the decision to perform this test is based on pediatricians' observations, but the criteria are not well-defined. Therefore, we utilized unsupervised learning to categorize patients based on the clinical manifestations of GAS pharyngitis. Our goal was to pinpoint the clinical symptoms that should prompt further examination and treatment in patients diagnosed with pharyngitis. Methods We analyzed categorical data from 305 RADT-positive patients aged three to 15 years using the K-modes clustering method. Each explanatory variable's relationship with cluster variables was statistically examined. Finally, we tested the differences between clusters for continuous variables statistically. Results The K-modes method categorized the cases into two clusters. Cluster 1 included older children with lymph node tenderness, while Cluster 2 consisted of younger children with cough and rhinorrhea. Conclusion Differentiating streptococcal pharyngitis from common cold or upper respiratory tract infection based on clinical symptoms alone is challenging, particularly in young patients. Future research should focus on identifying indicators that can aid in suspecting streptococcal infection in young patients.

3D Unsupervised deep learning method for magnetic resonance imaging-to-computed tomography synthesis in prostate radiotherapy.

Background and purpose: Magnetic resonance imaging (MRI)-to-computed tomography (CT) synthesis is essential in MRI-only radiotherapy workflows, particularly through deep learning techniques known for their accuracy. However, current supervised methods are limited to specific center's learnings and depend on registration precision. The aim of this study was to evaluate the accuracy of unsupervised and supervised approaches in the context of prostate MRI-to-CT generation for radiotherapy dose calculation. Methods: CT/MRI image pairs from 99 prostate cancer patients across three different centers were used. A comparison between supervised and unsupervised conditional Generative Adversarial Networks (cGAN) was conducted. Unsupervised training incorporates a style transfer method with. Content and Style Representation for Enhanced Perceptual synthesis (CREPs) loss. For dose evaluation, the photon prescription dose was 60 Gy delivered in volumetric modulated arc therapy (VMAT). Imaging endpoint for sCT evaluation was Mean Absolute Error (MAE). Dosimetric endpoints included absolute dose differences and gamma analysis between CT and sCT dose calculations. Results: The unsupervised paired network exhibited the highest accuracy for the body with a MAE at 33.6 HU, the highest MAE was 45.5 HU obtained with unsupervised unpaired learning. All architectures provided clinically acceptable results for dose calculation with gamma pass rates above 94 % (1 % 1 mm 10 %). Conclusions: This study shows that multicenter data can produce accurate sCTs via unsupervised learning, eliminating CT-MRI registration. The sCTs not only matched HU values but also enabled precise dose calculations, suggesting their potential for wider use in MRI-only radiotherapy workflows.

Towards a rigorous analysis of mutual information in contrastive learning.

Contrastive learning has emerged as a cornerstone in unsupervised representation learning. Its primary paradigm involves an instance discrimination task utilizing InfoNCE loss where the loss has been proven to be a form of mutual information. Consequently, it has become a common practice to analyze contrastive learning using mutual information as a measure. Yet, this analysis approach presents difficulties due to the necessity of estimating mutual information for real-world applications. This creates a gap between the elegance of its mathematical foundation and the complexity of its estimation, thereby hampering the ability to derive solid and meaningful insights from mutual information analysis. In this study, we introduce three novel methods and a few related theorems, aimed at enhancing the rigor of mutual information analysis. Despite their simplicity, these methods can carry substantial utility. Leveraging these approaches, we reassess three instances of contrastive learning analysis, illustrating the capacity of the proposed methods to facilitate deeper comprehension or to rectify pre-existing misconceptions. The main results can be summarized as follows: (1) While small batch sizes influence the range of training loss, they do not inherently limit learned representation's information content or affect downstream performance adversely; (2) Mutual information, with careful selection of positive pairings and post-training estimation, proves to be a superior measure for evaluating practical networks; and (3) Distinguishing between task-relevant and irrelevant information presents challenges, yet irrelevant information sources do not necessarily compromise the generalization of downstream tasks.

Noise-imitation learning: unpaired speckle noise reduction for optical coherence tomography.

OBJECTIVE: Optical coherence tomography (OCT) is widely used in clinical practice for its non-invasive, high-resolution imaging capabilities. However, speckle noise inherent to its low coherence principle can degrade image quality and compromise diagnostic accuracy. While deep learning methods have shown promise in reducing speckle noise, obtaining well-registered image pairs remains challenging, leading to the development of unpaired methods. Despite their potential, existing unpaired methods suffer from redundancy in network structures or interaction mechanisms. Therefore, a more streamlined method for unpaired OCT denoising is essential. APPROACH: In this work, we propose a novel unpaired method for OCT image denoising, referred to as noise-imitation learning (NIL). NIL comprises three primary modules: the noise extraction module, which extracts noise features by denoising noisy images; the noise imitation module, which synthesizes noisy images and generates fake clean images; and the adversarial learning module, which differentiates between real and fake clean images through adversarial training. The complexity of NIL is significantly lower than that of previous unpaired methods, utilizing only one generator and one discriminator for training. MAIN RESULTS: By efficiently fusing unpaired images and employing adversarial training, NIL can extract more speckle noise information to enhance denoising performance. Building on NIL, we propose an OCT image denoising pipeline, NIL-NAFNet. This pipeline achieved PSNR, SSIM, and RMSE values of 31.27 dB, 0.865, and 7.00, respectively, on the PKU37 dataset. Extensive experiments suggest that our method outperforms state-of-the-art unpaired methods both qualitatively and quantitatively. SIGNIFICANCE: These findings indicate that the proposed NIL is a simple yet effective method for unpaired OCT speckle noise reduction. The OCT denoising pipeline based on NIL demonstrates exceptional performance and efficiency. By addressing speckle noise without requiring well-registered image pairs, this method can enhance image quality and diagnostic accuracy in clinical practice.

Unsupervised speckle noise reduction technique for clinical ultrasound imaging.

PURPOSE: Deep learning-based image enhancement has significant potential in the field of ultrasound image processing, as it can accurately model complicated nonlinear artifacts and noise, such as ultrasonic speckle patterns. However, training deep learning networks to acquire reference images that are clean and free of noise presents significant challenges. This study introduces an unsupervised deep learning framework, termed speckle-to-speckle (S2S), designed for speckle and noise suppression. This framework can complete its training without the need for clean (speckle-free) reference images. METHODS: The proposed network leverages statistical reasoning for the mutual training of two in vivo images, each with distinct speckle patterns and noise. It then infers speckle- and noise-free images without needing clean reference images. This approach significantly reduces the time, cost, and effort experts need to invest in annotating reference images manually. RESULTS: The experimental results demonstrated that the proposed approach outperformed existing techniques in terms of the signal-to-noise ratio, contrast-to-noise ratio, structural similarity index, edge preservation index, and processing time (up to 86 times faster). It also performed excellently on images obtained from ultrasound scanners other than the ones used in this work. CONCLUSION: S2S demonstrates the potential of employing an unsupervised learning-based technique in medical imaging applications, where acquiring a ground truth reference is challenging.

Online Adaptive Kalman Filtering for Real-Time Anomaly Detection in Wireless Sensor Networks.

Wireless sensor networks (WSNs) are essential for a wide range of applications, including environmental monitoring and smart city developments, thanks to their ability to collect and transmit diverse physical and environmental data. The nature of WSNs, coupled with the variability and noise sensitivity of cost-effective sensors, presents significant challenges in achieving accurate data analysis and anomaly detection. To address these issues, this paper presents a new framework, called Online Adaptive Kalman Filtering (OAKF), specifically designed for real-time anomaly detection within WSNs. This framework stands out by dynamically adjusting the filtering parameters and anomaly detection threshold in response to live data, ensuring accurate and reliable anomaly identification amidst sensor noise and environmental changes. By highlighting computational efficiency and scalability, the OAKF framework is optimized for use in resource-constrained sensor nodes. Validation on different WSN dataset sizes confirmed its effectiveness, showing 95.4% accuracy in reducing false positives and negatives as well as achieving a processing time of 0.008 s per sample.

Cross-view motion consistent self-supervised video inter-intra contrastive for action representation understanding.

Self-supervised contrastive learning draws on power representational models to acquire generic semantic features from unlabeled data, and the key to training such models lies in how accurately to track motion features. Previous video contrastive learning methods have extensively used spatially or temporally augmentation as similar instances, resulting in models that are more likely to learn static backgrounds than motion features. To alleviate the background shortcuts, in this paper, we propose a cross-view motion consistent (CVMC) self-supervised video inter-intra contrastive model to focus on the learning of local details and long-term temporal relationships. Specifically, we first extract the dynamic features of consecutive video snippets and then align these features based on multi-view motion consistency. Meanwhile, we compare the optimized dynamic features for instance comparison of different videos and local spatial fine-grained with temporal order in the same video, respectively. Ultimately, the joint optimization of spatio-temporal alignment and motion discrimination effectively fills the challenges of the missing components of instance recognition, spatial compactness, and temporal perception in self-supervised learning. Experimental results show that our proposed self-supervised model can effectively learn visual representation information and achieve highly competitive performance compared to other state-of-the-art methods in both action recognition and video retrieval tasks.

Deep coherence learning: An unsupervised deep beamformer for high quality single plane wave imaging in medical ultrasound.

Plane wave imaging (PWI) in medical ultrasound is becoming an important reconstruction method with high frame rates and new clinical applications. Recently, single PWI based on deep learning (DL) has been studied to overcome lowered frame rates of traditional PWI with multiple PW transmissions. However, due to the lack of appropriate ground truth images, DL-based PWI still remains challenging for performance improvements. To address this issue, in this paper, we propose a new unsupervised learning approach, i.e., deep coherence learning (DCL)-based DL beamformer (DL-DCL), for high-quality single PWI. In DL-DCL, the DL network is trained to predict highly correlated signals with a unique loss function from a set of PW data, and the trained DL model encourages high-quality PWI from low-quality single PW data. In addition, the DL-DCL framework based on complex baseband signals enables a universal beamformer. To assess the performance of DL-DCL, simulation, phantom and in vivo studies were conducted with public datasets, and it was compared with traditional beamformers (i.e., DAS with 75-PWs and DMAS with 1-PW) and other DL-based methods (i.e., supervised learning approach with 1-PW and generative adversarial network (GAN) with 1-PW). From the experiments, the proposed DL-DCL showed comparable results with DMAS with 1-PW and DAS with 75-PWs in spatial resolution, and it outperformed all comparison methods in contrast resolution. These results demonstrated that the proposed unsupervised learning approach can address the inherent limitations of traditional PWIs based on DL, and it also showed great potential in clinical settings with minimal artifacts.

Stratification of Heart Sounds Morphology Through Unsupervised Learning.

The use of heart sounds for the assessment of the hemodynamic condition of the heart in telemonitoring applications is object of wide research at date. Many different approaches have been tried out for the analysis of the first (S1) and second (S2) heart sounds, but their morphological interpretation is still to be explored: in fact, the sound morphology is not unique and this impact the separability of the heart sounds components with methods based on envelopes or model optimization. In this study, we propose a method to stratify S1 and S2 according to their morphology to explore their diversity and increase their morphological interpretability. The method we propose is based on unsupervised learning, which we obtain using the cascade of four Self-Organizing Maps (SOMs) of decreasing dimensions. When tested on a publicly available heart sounds dataset, the proposed clustering approach proved to be robust and consistent, with over 80% of the heartbeats of the same patient being clustered together. The identified heart sounds templates highlight differences in the time and energy domains which may open to new directions of analysis in the future.

Robust Unsupervised Super-Resolution of Infant MRI via Dual-Modal Deep Image Prior.

Magnetic resonance imaging (MRI) is commonly used for studying infant brain development. However, due to the lengthy image acquisition time and limited subject compliance, high-quality infant MRI can be challenging. Without imposing additional burden on image acquisition, image super-resolution (SR) can be used to enhance image quality post-acquisition. Most SR techniques are supervised and trained on multiple aligned low-resolution (LR) and high-resolution (HR) image pairs, which in practice are not usually available. Unlike supervised approaches, Deep Image Prior (DIP) can be employed for unsupervised single-image SR, utilizing solely the input LR image for de novo optimization to produce an HR image. However, determining when to stop early in DIP training is non-trivial and presents a challenge to fully automating the SR process. To address this issue, we constrain the low-frequency k-space of the SR image to be similar to that of the LR image. We further improve performance by designing a dual-modal framework that leverages shared anatomical information between T1-weighted and T2-weighted images. We evaluated our model, dual-modal DIP (dmDIP), on infant MRI data acquired from birth to one year of age, demonstrating that enhanced image quality can be obtained with substantially reduced sensitivity to early stopping.

A new possibilistic-based clustering method for probability density functions and its application to detecting abnormal elements.

Cluster analysis can also detect abnormalities besides building a basis for identifying elements into clusters. Detecting abnormalities is a highly developed feature in the field of unsupervised learning. However, existing studies have mainly focused on discrete data, not probability density functions. This paper enables a possibilistic approach to solving the clustering for probability density functions dealing with abnormal elements. First, the data are extracted using the density function. Then, they are passed through the proposed algorithm to produce a possibilistic partition. Finally, a decision rule is established to recognize which function is abnormal. We compare the proposed algorithm with baseline algorithms in clustering PDFs, such as k-means, FCF, and Self-Updated Clustering. The results of three numerical examples applied to the image are typical for this new method. Furthermore, The proposed algorithm reaches accuracy at 100% over simulated benchmark data and outperforms baseline methods. Additionally, two last examples apply to image data reaching G-mean up from 96 to 100% (Sensitivity: 92-100% and Specificity: 100%). The proposed method can be researched and used to understand the internal structures of big data in the digital age through the probability density functions.

Disentangling Cobionts and Contamination in Long-Read Genomic Data using Sequence Composition.

The recent acceleration in genome sequencing targeting previously unexplored parts of the tree of life presents computational challenges. Samples collected from the wild often contain sequences from several organisms, including the target, its cobionts, and contaminants. Effective methods are therefore needed to separate sequences. Though advances in sequencing technology make this task easier, it remains difficult to taxonomically assign sequences from eukaryotic taxa that are not well-represented in databases. Therefore, reference-based methods alone are insufficient. Here, I examine how we can take advantage of differences in sequence composition between organisms to identify symbionts, parasites and contaminants in samples, with minimal reliance on reference data. To this end, I explore data from the Darwin Tree of Life project, including hundreds of high-quality HiFi read sets from insects. Visualising two-dimensional representations of read tetranucleotide composition learned by a Variational Autoencoder can reveal distinct components of a sample. Annotating the embeddings with additional information, such as coding density, estimated coverage, or taxonomic labels allows rapid assessment of the contents of a dataset. The approach scales to millions of sequences, making it possible to explore unassembled read sets, even for large genomes. Combined with interactive visualisation tools, it allows a large fraction of cobionts reported by reference-based screening to be identified. Crucially, it also facilitates retrieving genomes for which suitable reference data are absent.

Clustering analysis across different speeds reveals two distinct running techniques with no differences in running economy.

Establishing the links between running technique and economy remains elusive due to high inter-individual variability. Clustering runners by technique may enable tailored training recommendations, yet it is unclear if different techniques are equally economical and whether clusters are speed-dependent. This study aimed to identify clusters of runners based on technique and to compare cluster kinematics and running economy. Additionally, we examined the agreement of clustering partitions of the same runners at different speeds. Trunk and lower-body kinematics were captured from 84 trained runners at different speeds on a treadmill. We used Principal Component Analysis for dimensionality reduction and agglomerative hierarchical clustering to identify groups of runners with a similar technique, and we evaluated cluster agreement across speeds. Clustering runners at different speeds independently produced different partitions, suggesting single speed clustering can fail to capture the full speed profile of a runner. The two clusters identified using data from the whole range of speeds showed differences in pelvis tilt and duty factor. In agreement with self-optimisation theories, there were no differences in running economy, and no differences in participants' characteristics between clusters. Considering inter-individual technique variability may enhance the efficacy of training designs as opposed to 'one size fits all' approaches.

Multi-objective Bayesian optimization with enhanced features for adaptively improved glioblastoma partitioning and survival prediction.

Glioblastoma, an aggressive brain tumor prevalent in adults, exhibits heterogeneity in its microstructures and vascular patterns. The delineation of its subregions could facilitate the development of region-targeted therapies. However, current unsupervised learning techniques for this task face challenges in reliability due to fluctuations of clustering algorithms, particularly when processing data from diverse patient cohorts. Furthermore, stable clustering results do not guarantee clinical meaningfulness. To establish the clinical relevance of these subregions, we will perform survival predictions using radiomic features extracted from them. Following this, achieving a balance between outcome stability and clinical relevance presents a significant challenge, further exacerbated by the extensive time required for hyper-parameter tuning. In this study, we introduce a multi-objective Bayesian optimization (MOBO) framework, which leverages a Feature-enhanced Auto-Encoder (FAE) and customized losses to assess both the reproducibility of clustering algorithms and the clinical relevance of their outcomes. Specifically, we embed the entirety of these processes within the MOBO framework, modeling both using distinct Gaussian Processes (GPs). The proposed MOBO framework can automatically balance the trade-off between the two criteria by employing bespoke stability and clinical significance losses. Our approach efficiently optimizes all hyper-parameters, including the FAE architecture and clustering parameters, within a few steps. This not only accelerates the process but also consistently yields robust MRI subregion delineations and provides survival predictions with strong statistical validation.

AutoCorNN: An Unsupervised Physics-Aware Deep Learning Model for Geometric Distortion Correction of Brain MRI Images Towards MR-Only Stereotactic Radiosurgery.

Geometric distortions in brain MRI images arising from susceptibility artifacts at air-tissue interfaces pose a significant challenge for high-precision radiation therapy modalities like stereotactic radiosurgery, necessitating sub-millimeter accuracy. To achieve this goal, we developed AutoCorNN, an unsupervised physics-aware deep-learning model for correcting geometric distortions. Two publicly available datasets, the MPI-Leipzig Mind-Brain-Body with 318 subjects, and the Vestibular Schwannoma-SEG dataset, encompassing 242 patients were utilized. AutoCorNN integrates two 2D convolutional encoder-decoder neural networks with the forward physical model of MRI signal generation to predict undistorted MR and field map images from distorted MR input. The network is trained in an unsupervised manner by minimizing the mean absolute error between the measured and estimated k-space data, without requiring ground truth images during training or deployment. The model was evaluated on vestibular schwannoma cases. AutoCorNN achieved a peak signal-to-noise ratio (PSNR) of 41.35 ± 0.02 dB, a root mean square error (RMSE) of 0.02 ± 0.003, and a structural similarity index (SSIM) of 0.99 ± 0.02 outperforming uncorrected and B0-mapping correction methods. Geometric distortions of about 1.6 mm were observed at the air-tissue interfaces at the air canal and nasal cavity borders. Geometrically, distortion correction increased the target volume from 3.12 ± 0.52 cc to 3.84 ± 0.54 cc. Dosimetrically, AutoCorNN improved target coverage (0.96 ± 0.01 to 0.97 ± 0.02), conformity index (0.92 ± 0.03 to 0.94 ± 0.03), and reduced dose gradients outside the target. AutoCorNN achieves accurate geometric distortion correction comparable to conventional iterative methods while offering substantial computational acceleration, enabling precise target delineation and conformal dose delivery for improved radiation therapy outcomes.

Do Cognitive Subtypes Exist in People at Clinical High Risk for Psychosis? Results From the EU-GEI Study.

BACKGROUND AND HYPOTHESIS: Cognition has been associated with socio-occupational functioning in individuals at Clinical High Risk for Psychosis (CHR-P). The present study hypothesized that clustering CHR-P participants based on cognitive data could reveal clinically meaningful subtypes. STUDY DESIGN: A cohort of 291 CHR-P subjects was recruited through the multicentre EU-GEI high-risk study. We explored whether an underlying cluster structure was present in the cognition data. Clustering of cognition data was performed using k-means clustering and density-based spatial clustering of applications with noise. Cognitive subtypes were validated by comparing differences in functioning, psychosis symptoms, transition outcome, and grey matter volume between clusters. Network analysis was used to further examine relationships between cognition scores and clinical symptoms. STUDY RESULTS: No underlying cluster structure was found in the cognitive data. K-means clustering produced "spared" and "impaired" cognition clusters similar to those reported in previous studies. However, these clusters were not associated with differences in functioning, symptomatology, outcome, or grey matter volume. Network analysis identified cognition and symptoms/functioning measures that formed separate subnetworks of associations. CONCLUSIONS: Stratifying patients according to cognitive performance has the potential to inform clinical care. However, we did not find evidence of cognitive clusters in this CHR-P sample. We suggest that care needs to be taken in inferring the existence of distinct cognitive subtypes from unsupervised learning studies. Future research in CHR-P samples could explore the existence of cognitive subtypes across a wider range of cognitive domains.

An unsupervised learning-based guidewire shape registration for vascular intervention surgery robot.

In a vascular interventional surgery robot(VISR), a high transparency master-slave system can aid physicians in the more precise manipulation of guidewires for navigation and operation within blood vessels. However, deformations arising from the movement of the guidewire can affect the accuracy of the registration, thus reducing the transparency of the master-slave system. In this study, the degree of the guidewire's deformation is analyzed based on the Kirchhoff model. An unsupervised learning-based guidewire shape registration method(UL-GSR) is proposed to estimate geometric transformations by learning displacement field functions. It can effectively achieve precise registration of flexible bodies. This method not only demonstrates high registration accuracy but also performs robustly under different complexity degrees of guidewire shapes. The experiments have demonstrated that the UL-GSR method significantly improves the accuracy of shape point set registration between the master and slave sides, thus enhancing the transparency and operational reliability of the VISR system.