Unsupervised deep learning framework for data-driven gating in positron emission tomography

BACKGROUND

Physiological motion, such as respiratory motion, has become a limiting factor in the spatial resolution of positron emission tomography (PET) imaging as the resolution of PET detectors continue to improve. Motion-induced misregistration between PET and CT images can also cause attenuation correction artifacts. Respiratory gating can be used to freeze the motion and to reduce motion induced artifacts.

PURPOSE

In this study, we propose a robust data-driven approach using an unsupervised deep clustering network that employs an autoencoder (AE) to extract latent features for respiratory gating.

METHODS

We first divide list-mode PET data into short-time frames. The short-time frame images are reconstructed without attenuation, scatter, or randoms correction to avoid attenuation mismatch artifacts and to reduce image reconstruction time. The deep AE is then trained using reconstructed short-time frame images to extract latent features for respiratory gating. No additional data are required for the AE training. K-means clustering is subsequently used to perform respiratory gating based on the latent features extracted by the deep AE. The effectiveness of our proposed Deep Clustering method was evaluated using physical phantom and real patient datasets. The performance was compared against phase gating based on an external signal (External) and image based principal component analysis (PCA) with K-means clustering (Image PCA).

RESULTS

The proposed method produced gated images with higher contrast and sharper myocardium boundaries than those obtained using the External gating method and Image PCA. Quantitatively, the gated images generated by the proposed Deep Clustering method showed larger center of mass (COM) displacement and higher lesion contrast than those obtained using the other two methods.

CONCLUSIONS

The effectiveness of our proposed method was validated using physical phantom and real patient data. The results showed our proposed framework could provide superior gating than the conventional External method and Image PCA.

Functional Imaging Derived ADHD Biotypes Based on Deep Clustering May Guide Personalized Medication Therapy

Attention deficit hyperactivity disorder (ADHD) is one prevalent neurodevelopmental disorder with childhood onset, however, there is no clear correspondence established between clinical ADHD subtypes and primary medications. Identifying objective and reliable neuroimaging markers for categorizing ADHD biotypes may lead to more individualized, biotype-guided treatment. Here we proposed graph convolutional network plus deep clustering for ADHD biotype detection using functional network connectivity (FNC), resulting in two biotypes based on 1069 ADHD patients selected from Adolescent Brain and Cognitive Development (ABCD) study, which were well replicated on independent ADHD adolescents undergoing longitudinal medication treatment (n=130). Interestingly, in addition to differences in cognitive performance and hyperactivity/impulsivity symptoms, biotype 1 treated with methylphenidate demonstrated significantly better recovery than biotype 2 treated with atomoxetine (p<0.05, FDR corrected). This imaging-driven, biotype-guided approach holds promise for facilitating personalized treatment of ADHD, exploring possible boundaries through innovative deep learning algorithms aimed at improving medication treatment effectiveness.

Deep enhanced constraint clustering based on contrastive learning for scRNA-seq data

Single-cell RNA sequencing (scRNA-seq) measures transcriptome-wide gene expression at single-cell resolution. Clustering analysis of scRNA-seq data enables researchers to characterize cell types and states, shedding new light on cell-to-cell heterogeneity in complex tissues. Recently, self-supervised contrastive learning has become a prominent technique for underlying feature representation learning. However, for the noisy, high-dimensional and sparse scRNA-seq data, existing methods still encounter difficulties in capturing the intrinsic patterns and structures of cells, and seldom utilize prior knowledge, resulting in clusters that mismatch with the real situation. To this end, we propose scDECL, a novel deep enhanced constraint clustering algorithm for scRNA-seq data analysis based on contrastive learning and pairwise constraints. Specifically, based on interpolated contrastive learning, a pre-training model is trained to learn the feature embedding, and then perform clustering according to the constructed enhanced pairwise constraint. In the pre-training stage, a mixup data augmentation strategy and interpolation loss is introduced to improve the diversity of the dataset and the robustness of the model. In the clustering stage, the prior information is converted into enhanced pairwise constraints to guide the clustering. To validate the performance of scDECL, we compare it with six state-of-the-art algorithms on six real scRNA-seq datasets. The experimental results demonstrate the proposed algorithm outperforms the six competing methods. In addition, the ablation studies on each module of the algorithm indicate that these modules are complementary to each other and effective in improving the performance of the proposed algorithm. Our method scDECL is implemented in Python using the Pytorch machine-learning library, and it is freely available at https://github.com/DBLABDHU/scDECL.

Deep clustering for abdominal organ classification in ultrasound imaging

Purpose

The purpose of this study is to examine the utilization of unlabeled data for abdominal organ classification in multi-label (non-mutually exclusive classes) ultrasound images, as an alternative to the conventional transfer learning approach.

Approach

We present a new method for classifying abdominal organs in ultrasound images. Unlike previous approaches that only relied on labeled data, we consider the use of both labeled and unlabeled data. To explore this approach, we first examine the application of deep clustering for pretraining a classification model. We then compare two training methods, fine-tuning with labeled data through supervised learning and fine-tuning with both labeled and unlabeled data using semisupervised learning. All experiments were conducted on a large dataset of unlabeled images (n u = 84967 ) and a small set of labeled images (n s = 2742 ) comprising progressively 10%, 20%, 50%, and 100% of the images.

Results

We show that for supervised fine-tuning, deep clustering is an effective pre-training method, with performance matching that of ImageNet pre-training using five times less labeled data. For semi-supervised learning, deep clustering pre-training also yields higher performance when the amount of labeled data is limited. Best performance is obtained with deep clustering pre-training combined with semi-supervised learning and 2742 labeled example images with an F 1 -score weighted average of 84.1%.

Conclusions

This method can be used as a tool to preprocess large unprocessed databases, thus reducing the need for prior annotations of abdominal ultrasound studies for the training of image classification algorithms, which in turn could improve the clinical use of ultrasound images.

ClusterX: a novel representation learning-based deep clustering framework for accurate visual inspection in virtual screening

Molecular clustering analysis has been developed to facilitate visual inspection in the process of structure-based virtual screening. However, traditional methods based on molecular fingerprints or molecular descriptors limit the accuracy of selecting active hit compounds, which may be attributed to the lack of representations of receptor structural and protein-ligand interaction during the clustering. Here, a novel deep clustering framework named ClusterX is proposed to learn molecular representations of protein-ligand complexes and cluster the ligands. In ClusterX, the graph was used to represent the protein-ligand complex, and the joint optimisation can be used efficiently for learning the cluster-friendly features. Experiments on the KLIFs database show that the model can distinguish well between the binding modes of different kinase inhibitors. To validate the effectiveness of the model, the clustering results on the virtual screening dataset further demonstrated that ClusterX achieved better or more competitive performance against traditional methods, such as SIFt and extended connectivity fingerprints. This framework may provide a unique tool for clustering analysis and prove to assist computational medicinal chemists in visual decision-making.

Denoising adaptive deep clustering with self-attention mechanism on single-cell sequencing data

A large number of works have presented the single-cell RNA sequencing (scRNA-seq) to study the diversity and biological functions of cells at the single-cell level. Clustering identifies unknown cell types, which is essential for downstream analysis of scRNA-seq samples. However, the high dimensionality, high noise and pervasive dropout rate of scRNA-seq samples have a significant challenge to the cluster analysis of scRNA-seq samples. Herein, we propose a new adaptive fuzzy clustering model based on the denoising autoencoder and self-attention mechanism called the scDASFK. It implements the comparative learning to integrate cell similar information into the clustering method and uses a deep denoising network module to denoise the data. scDASFK consists of a self-attention mechanism for further denoising where an adaptive clustering optimization function for iterative clustering is implemented. In order to make the denoised latent features better reflect the cell structure, we introduce a new adaptive feedback mechanism to supervise the denoising process through the clustering results. Experiments on 16 real scRNA-seq datasets show that scDASFK performs well in terms of clustering accuracy, scalability and stability. Overall, scDASFK is an effective clustering model with great potential for scRNA-seq samples analysis. Our scDASFK model codes are freely available at https://github.com/LRX2022/scDASFK.

Subtype-DCC: decoupled contrastive clustering method for cancer subtype identification based on multi-omics data

Due to the high heterogeneity and complexity of cancers, patients with different cancer subtypes often have distinct groups of genomic and clinical characteristics. Therefore, the discovery and identification of cancer subtypes are crucial to cancer diagnosis, prognosis and treatment. Recent technological advances have accelerated the increasing availability of multi-omics data for cancer subtyping. To take advantage of the complementary information from multi-omics data, it is necessary to develop computational models that can represent and integrate different layers of data into a single framework. Here, we propose a decoupled contrastive clustering method (Subtype-DCC) based on multi-omics data integration for clustering to identify cancer subtypes. The idea of contrastive learning is introduced into deep clustering based on deep neural networks to learn clustering-friendly representations. Experimental results demonstrate the superior performance of the proposed Subtype-DCC model in identifying cancer subtypes over the currently available state-of-the-art clustering methods. The strength of Subtype-DCC is also supported by the survival and clinical analysis.

scDCCA: deep contrastive clustering for single-cell RNA-seq data based on auto-encoder network

The advances in single-cell ribonucleic acid sequencing (scRNA-seq) allow researchers to explore cellular heterogeneity and human diseases at cell resolution. Cell clustering is a prerequisite in scRNA-seq analysis since it can recognize cell identities. However, the high dimensionality, noises and significant sparsity of scRNA-seq data have made it a big challenge. Although many methods have emerged, they still fail to fully explore the intrinsic properties of cells and the relationship among cells, which seriously affects the downstream clustering performance. Here, we propose a new deep contrastive clustering algorithm called scDCCA. It integrates a denoising auto-encoder and a dual contrastive learning module into a deep clustering framework to extract valuable features and realize cell clustering. Specifically, to better characterize and learn data representations robustly, scDCCA utilizes a denoising Zero-Inflated Negative Binomial model-based auto-encoder to extract low-dimensional features. Meanwhile, scDCCA incorporates a dual contrastive learning module to capture the pairwise proximity of cells. By increasing the similarities between positive pairs and the differences between negative ones, the contrasts at both the instance and the cluster level help the model learn more discriminative features and achieve better cell segregation. Furthermore, scDCCA joins feature learning with clustering, which realizes representation learning and cell clustering in an end-to-end manner. Experimental results of 14 real datasets validate that scDCCA outperforms eight state-of-the-art methods in terms of accuracy, generalizability, scalability and efficiency. Cell visualization and biological analysis demonstrate that scDCCA significantly improves clustering and facilitates downstream analysis for scRNA-seq data. The code is available at https://github.com/WJ319/scDCCA.

SPICE: Semantic Pseudo-Labeling for Image Clustering

The similarity among samples and the discrepancy among clusters are two crucial aspects of image clustering. However, current deep clustering methods suffer from inaccurate estimation of either feature similarity or semantic discrepancy. In this paper, we present a Semantic Pseudo-labeling-based Image ClustEring (SPICE) framework, which divides the clustering network into a feature model for measuring the instance-level similarity and a clustering head for identifying the cluster-level discrepancy. We design two semantics-aware pseudo-labeling algorithms, prototype pseudo-labeling and reliable pseudo-labeling, which enable accurate and reliable self-supervision over clustering. Without using any ground-truth label, we optimize the clustering network in three stages: 1) train the feature model through contrastive learning to measure the instance similarity; 2) train the clustering head with the prototype pseudo-labeling algorithm to identify cluster semantics; and 3) jointly train the feature model and clustering head with the reliable pseudo-labeling algorithm to improve the clustering performance. Extensive experimental results demonstrate that SPICE achieves significant improvements (~10%) over existing methods and establishes the new state-of-the-art clustering results on six balanced benchmark datasets in terms of three popular metrics. Importantly, SPICE significantly reduces the gap between unsupervised and fully-supervised classification; e.g. there is only 2% (91.8% vs 93.8%) accuracy difference on CIFAR-10. Our code is made publicly available at https://github.com/niuchuangnn/SPICE.

A Deep Clustering-based Novel Approach for Binning of Metagenomics Data

Background

One major challenge in binning Metagenomics data is the limited availability of reference datasets, as only 1% of the total microbial population is yet cultured. This has given rise to the efficacy of unsupervised methods for binning in the absence of any reference datasets.

Objective

To develop a deep clustering-based binning approach for Metagenomics data and to evaluate results with suitable measures.

Methods

In this study, a deep learning-based approach has been taken for binning the Metagenomics data. The results are validated on different datasets by considering features such as Tetra-nucleotide frequency (TNF), Hexa-nucleotide frequency (HNF) and GC-Content. Convolutional Autoencoder is used for feature extraction and for binning; the K-means clustering method is used.

Results

In most cases, it has been found that evaluation parameters such as the Silhouette index and Rand index are more than 0.5 and 0.8, respectively, which indicates that the proposed approach is giving satisfactory results. The performance of the developed approach is compared with current methods and tools using benchmarked low complexity simulated and real metagenomic datasets. It is found better for unsupervised and at par with semi-supervised methods.

Conclusion

An unsupervised advanced learning-based approach for binning has been proposed, and the developed method shows promising results for various datasets. This is a novel approach for solving the lack of reference data problem of binning in metagenomics.

Learning neural decoders without labels using multiple data streams

OBJECTIVE

Recent advances in neural decoding have accelerated the development of brain-computer interfaces aimed at assisting users with everyday tasks such as speaking, walking, and manipulating objects. However, current approaches for training neural decoders commonly require large quantities of labeled data, which can be laborious or infeasible to obtain in real-world settings. Alternatively, self-supervised models that share self-generated pseudo-labels between two data streams have shown exceptional performance on unlabeled audio and video data, but it remains unclear how well they extend to neural decoding.

APPROACH

We learn neural decoders without labels by leveraging multiple simultaneously recorded data streams, including neural, kinematic, and physiological signals. Specifically, we apply cross-modal, self-supervised deep clustering to train decoders that can classify movements from brain recordings. After training, we then isolate the decoders for each input data stream and compare the accuracy of decoders trained using cross-modal deep clustering against supervised and unimodal, self-supervised models.

MAIN RESULTS

We find that sharing pseudo-labels between two data streams during training substantially increases decoding performance compared to unimodal, self-supervised models, with accuracies approaching those of supervised decoders trained on labeled data. Next, we extend cross-modal decoder training to three or more modalities, achieving state-of-the-art neural decoding accuracy that matches or slightly exceeds the performance of supervised models. Significance: We demonstrate that cross-modal, self-supervised decoding can be applied to train neural decoders when few or no labels are available and extend the cross-modal framework to share information among three or more data streams, further improving self-supervised training.

Cluster activation mapping with application to computed tomography scans of the lung

Purpose: An open question in deep clustering is how to explain what in the image is driving the cluster assignments. This is especially important for applications in medical imaging when the derived cluster assignments may inform decision-making or create new disease subtypes. We develop cluster activation mapping (CLAM), which is methodology to create localization maps highlighting the image regions important for cluster assignment. Approach: Our approach uses a linear combination of the activation channels from the last layer of the encoder within a pretrained autoencoder. The activation channels are weighted by a channelwise confidence measure, which is a modification of score-CAM. Results: Our approach performs well under medical imaging-based simulation experiments, when the image clusters differ based on size, location, and intensity of abnormalities. Under simulation, the cluster assignments were predicted with 100% accuracy when the number of clusters was set at the true value. In addition, applied to computed tomography scans from a sarcoidosis population, CLAM identified two subtypes of sarcoidosis based purely on CT scan presentation, which were significantly associated with pulmonary function tests and visual assessment scores, such as ground-glass, fibrosis, and honeycombing. Conclusions: CLAM is a transparent methodology for identifying explainable groupings of medical imaging data. As deep learning networks are often criticized and not trusted due to their lack of interpretability, our contribution of CLAM to deep clustering architectures is critical to our understanding of cluster assignments, which can ultimately lead to new subtypes of diseases.

Deep structural clustering for single-cell RNA-seq data jointly through autoencoder and graph neural network

Single-cell RNA sequencing (scRNA-seq) permits researchers to study the complex mechanisms of cell heterogeneity and diversity. Unsupervised clustering is of central importance for the analysis of the scRNA-seq data, as it can be used to identify putative cell types. However, due to noise impacts, high dimensionality and pervasive dropout events, clustering analysis of scRNA-seq data remains a computational challenge. Here, we propose a new deep structural clustering method for scRNA-seq data, named scDSC, which integrate the structural information into deep clustering of single cells. The proposed scDSC consists of a Zero-Inflated Negative Binomial (ZINB) model-based autoencoder, a graph neural network (GNN) module and a mutual-supervised module. To learn the data representation from the sparse and zero-inflated scRNA-seq data, we add a ZINB model to the basic autoencoder. The GNN module is introduced to capture the structural information among cells. By joining the ZINB-based autoencoder with the GNN module, the model transfers the data representation learned by autoencoder to the corresponding GNN layer. Furthermore, we adopt a mutual supervised strategy to unify these two different deep neural architectures and to guide the clustering task. Extensive experimental results on six real scRNA-seq datasets demonstrate that scDSC outperforms state-of-the-art methods in terms of clustering accuracy and scalability. Our method scDSC is implemented in Python using the Pytorch machine-learning library, and it is freely available at https://github.com/DHUDBlab/scDSC.

Deep Semi-Supervised Algorithm for Learning Cluster-Oriented Representations of Medical Images Using Partially Observable DICOM Tags and Images

The task of automatically extracting large homogeneous datasets of medical images based on detailed criteria and/or semantic similarity can be challenging because the acquisition and storage of medical images in clinical practice is not fully standardised and can be prone to errors, which are often made unintentionally by medical professionals during manual input. In this paper, we propose an algorithm for learning cluster-oriented representations of medical images by fusing images with partially observable DICOM tags. Pairwise relations are modelled by thresholding the Gower distance measure which is calculated using eight DICOM tags. We trained the models using 30,000 images, and we tested them using a disjoint test set consisting of 8000 images, gathered retrospectively from the PACS repository of the Clinical Hospital Centre Rijeka in 2017. We compare our method against the standard and deep unsupervised clustering algorithms, as well as the popular semi-supervised algorithms combined with the most commonly used feature descriptors. Our model achieves an NMI score of 0.584 with respect to the anatomic region, and an NMI score of 0.793 with respect to the modality. The results suggest that DICOM data can be used to generate pairwise constraints that can help improve medical images clustering, even when using only a small number of constraints.

ScCAEs: deep clustering of single-cell RNA-seq via convolutional autoencoder embedding and soft K-means

Clustering and cell type classification are a vital step of analyzing scRNA-seq data to reveal the complexity of the tissue (e.g. the number of cell types and the transcription characteristics of the respective cell type). Recently, deep learning-based single-cell clustering algorithms become popular since they integrate the dimensionality reduction with clustering. But these methods still have unstable clustering effects for the scRNA-seq datasets with high dropouts or noise. In this study, a novel single-cell RNA-seq deep embedding clustering via convolutional autoencoder embedding and soft K-means (scCAEs) is proposed by simultaneously learning the feature representation and clustering. It integrates the deep learning with convolutional autoencoder to characterize scRNA-seq data and proposes a regularized soft K-means algorithm to cluster cell populations in a learned latent space. Next, a novel constraint is introduced to the clustering objective function to iteratively optimize the clustering results, and more importantly, it is theoretically proved that this objective function optimization ensures the convergence. Moreover, it adds the reconstruction loss to the objective function combining the dimensionality reduction with clustering to find a more suitable embedding space for clustering. The proposed method is validated on a variety of datasets, in which the number of clusters in the mentioned datasets ranges from 4 to 46, and the number of cells ranges from 90 to 30 302. The experimental results show that scCAEs is superior to other state-of-the-art methods on the mentioned datasets, and it also keeps the satisfying compatibility and robustness. In addition, for single-cell datasets with the batch effects, scCAEs can ensure the cell separation while removing batch effects.

Separation of normal and impaired dynamic cerebral autoregulation using deep embedded clustering: a proof-of-concept study

Objective. A previous study has shown that a data-driven approach can significantly improve the discriminative power of transfer function analysis (TFA) used to differentiate between normal and impaired cerebral autoregulation (CA) in two groups of data. The data was collected from both healthy subjects (assumed to have normal CA) and symptomatic patients with severe stenosis (assumed to have impaired CA). However, the sample size of the labeled data was relatively small, owing to the difficulty in data collection. Therefore, in this proof-of-concept study, we investigate the feasibility of using an unsupervised learning model to differentiate between normal and impaired CA on TFA variables without requiring labeled data for learning.Approach. Continuous arterial blood pressure (ABP) and cerebral blood flow velocity (CBFV), which were recorded simultaneously for approximately 10 min, were included from 148 subjects (41 healthy subjects, 31 with mild stenosis, 13 with moderate stenosis, 22 asymptomatic patients with severe stenosis, and 41 symptomatic patients with severe stenosis). Tiecks' model was used to generate surrogate data with normal and impaired CA. A recently proposed unsupervised learning model was optimized and applied to separate the normal and impaired CA for both the surrogate data and real data.Main results. It achieved 98.9% and 74.1% accuracy for the surrogate and real data, respectively.Significance. To our knowledge, this is the first attempt to employ an unsupervised data-driven approach to assess CA using TFA. This method enables the development of a classifier to determine the status of CA, which is currently lacking.

A Deep Clustering Method For Analyzing Uterine Cervix Images Across Imaging Devices

Visual inspection of the cervix with acetic acid (VIA), though error prone, has long been used for screening women and to guide management for cervical cancer. The automated visual evaluation (AVE) technique, in which deep learning is used to predict precancer based on a digital image of the acetowhitened cervix, has demonstrated its promise as a low-cost method to improve on human performance. However, there are several challenges in moving AVE beyond proof-of-concept and deploying it as a practical adjunct tool in visual screening. One of them is making AVE robust across images captured using different devices. We propose a new deep learning based clustering approach to investigate whether the images taken by three different devices (a common smartphone, a custom smartphone-based handheld device for cervical imaging, and a clinical colposcope equipped with SLR digital camera-based imaging capability) can be well distinguished from each other with respect to the visual appearance/content within their cervix regions. We argue that disparity in visual appearance of a cervix across devices could be a significant confounding factor in training and generalizing AVE performance. Our method consists of four components: cervix region detection, feature extraction, feature encoding, and clustering. Multiple experiments are conducted to demonstrate the effectiveness of each component and compare alternative methods in each component. Our proposed method achieves high clustering accuracy (97%) and significantly outperforms several representative deep clustering methods on our dataset. The high clustering performance indicates the images taken from these three devices are different with respect to visual appearance. Our results and analysis establish a need for developing a method that minimizes such variance among the images acquired from different devices. It also recognizes the need for large number of training images from different sources for robust device-independent AVE performance worldwide.

DEEP ATTRACTOR NETWORK FOR SINGLE-MICROPHONE SPEAKER SEPARATION

Despite the overwhelming success of deep learning in various speech processing tasks, the problem of separating simultaneous speakers in a mixture remains challenging. Two major difficulties in such systems are the arbitrary source permutation and unknown number of sources in the mixture. We propose a novel deep learning framework for single channel speech separation by creating attractor points in high dimensional embedding space of the acoustic signals which pull together the time-frequency bins corresponding to each source. Attractor points in this study are created by finding the centroids of the sources in the embedding space, which are subsequently used to determine the similarity of each bin in the mixture to each source. The network is then trained to minimize the reconstruction error of each source by optimizing the embeddings. The proposed model is different from prior works in that it implements an end-to-end training, and it does not depend on the number of sources in the mixture. Two strategies are explored in the test time, K-means and fixed attractor points, where the latter requires no post-processing and can be implemented in real-time. We evaluated our system on Wall Street Journal dataset and show 5.49% improvement over the previous state-of-the-art methods.