Embedding Clustering Research Articles

MPC-14 BRAF V600E MUTANT OLIGODENDROGLIOMA-LIKE TUMORS WITH CHROMOSOMAL INSTABILITY IN ADOLESCENTS AND YOUNG ADULTS

We performed genome-wide methylation analysis on 136 pediatric low grade gliomas, identifying a unique cluster consisting of oligodendroglioma-like BRAF V600E mutant tumors with Recurrent gain of Chromosome 7 and loss of Chromosome 10 (OLIVER). Hierarchical clustering and t-stochastic neighbor embedding analyses cluster them with previously described pediatric-type low grade gliomas, separate from adult gliomas. OLIVERS exhibit distinct clinical behavior as temporal lobe lesions in adolescents and young adults, prolonged history of seizures and all are alive with no recurrence (follow-up 3.2 to 13.2 years). Morphogically, all showed oligodendroglioma-like features, including round nuclei with perinuclear halos, a chicken-wire pattern of branching capillaries and microcalcification. None showed astrocytic features or characteristics suggestive of high-grade tumors including necrosis or mitotic figures. All tumors harbored multiple chromosomal copy number abnormalities (more than 10 chromosomes per OLIVER), but none showed 1p/19q co-deletion or IDH1 mutation. Interestingly, one tumor showed a TERT promoter mutation. Although the series is small, OLIVER may represent a new category of IDH wild-type low grade gliomas which may be confused with molecular GBM. Further, they highlight the heterogeneity of IDH wild-type gliomas and the relatively indolent behavior of pediatric-type gliomas.

A Novel Radar HRRP Recognition Method with Accelerated T-Distributed Stochastic Neighbor Embedding and Density-Based Clustering.

High-resolution range profile (HRRP) has attracted intensive attention from radar community because it is easy to acquire and analyze. However, most of the conventional algorithms require the prior information of targets, and they cannot process a large number of samples in real time. In this paper, a novel HRRP recognition method is proposed to classify unlabeled samples automatically where the number of categories is unknown. Firstly, with the preprocessing of HRRPs, we adopt principal component analysis (PCA) for dimensionality reduction of data. Afterwards, t-distributed stochastic neighbor embedding (t-SNE) with Barnes–Hut approximation is conducted for the visualization of high-dimensional data. It proves to reduce the dimensionality, which has significantly improved the computation speed. Finally, it is exhibited that the recognition performance with density-based clustering is superior to conventional algorithms under the condition of large azimuth angle ranges and low signal-to-noise ratio (SNR).

BRAF V600E mutant oligodendroglioma-like tumors with chromosomal instability in adolescents and young adults.

We performed genome-wide methylation analysis on 136 pediatric low-grade gliomas, identifying a unique cluster consisting of three tumors with oligodendroglioma-like histology, BRAF p.V600E mutations and recurrent whole chromosome gains of 7 and loss of 10. Morphologically, all showed similar features, including a diffusely infiltrative glioma composed of round nuclei with perinuclear halos, a chicken-wire pattern of branching capillaries and microcalcification. None showed astrocytic features or characteristics suggestive of high-grade tumors including necrosis or mitotic figures. All tumors harbored multiple chromosomal copy number abnormalities (>10 chromosomes altered), but none showed 1p/19q co-deletion or IDH1 p.R132H mutation. Hierarchical clustering and t-stochastic neighbor embedding analyses from DNA methylation data cluster them more closely to previously described pediatric-type low-grade gliomas and separate from adult gliomas. These tumors exhibit distinct clinical features; they are temporal lobe lesions occurring in adolescents and young adults with a prolonged history of seizures and all are alive with no recurrence (follow-up 3.2 to 13.2years). We encountered another young adult case with quite similar pathological appearance and molecular status except for TERT promoter mutation. Although the series is small, these may represent a new category of IDH wild-type low-grade gliomas which may be confused with "molecular GBM." Further, they highlight the heterogeneity of IDH wild-type gliomas and the relatively indolent behavior of "pediatric-type" gliomas.

Application of functional deep belief network for estimating daily global solar radiation: A case study in China

Solar energy plays an essential role in environment governance and resource protection, as it is totally pollution-free and extensively accessed. An accurate knowledge of solar radiation is beneficial to the deployments of solar energy constructions, photovoltaic and thermal solar systems. In this study, a deep learning method is proposed for estimating daily global solar radiation, which is constituted by embedding clustering (EC) and functional deep belief network (DBN). Based on the curve shapes of daily solar radiation, EC divides the overall dataset into different subsets, which can be modeled separately. Knowledge from empirical radiation models is also merged as the input of functional DBN. The model can be directly applied to solar estimation in various stations due to its strong nonlinear representation. The case study in China is adopted that involves radiation data from a total of 30 stations to validate the practicability and accuracy of the proposed method. From the results, the method obtains better estimation precision with empirical knowledge, achieving 1.706 MJ/m2 of mean absolute error (MAE), 2.352 MJ/m2 of root mean square error (RMSE) and 13.71% of mean absolute percentage error (MAPE) according to the average values at the 30 stations.

Seismic facies analysis based on deep convolutional embedded clustering

Seismic facies classification takes a two-step approach: attribute extraction and seismic facies analysis by using clustering algorithms, sequentially. In general, it is clear that the choice of feature extraction is critical for successful seismic facies analysis. However, the choice of features is customarily determined by the seismic interpreters, and so the clustering result is affected by the difference in the seismic interpreters’ experience levels. It becomes challenging to extract features and identify seismic facies simultaneously. We have introduced deep convolutional embedded clustering (DCEC), which aims to simultaneously learn feature representations and cluster assignments by using deep neural networks. Our method learns mapping from the data space to a lower dimensional feature space in which it iteratively optimizes a clustering objective by building a specific loss function. We apply the method to the Modified National Institute of Standards and Technology (MNIST) data, geophysical model data, and field seismic data. In the MNIST data, the DCEC method shows better latent space of clustering results than traditional clustering methods. In the geophysical model data, the accuracy of waveform classification based on DCEC method is higher than traditional clustering methods. The results from the seismic data demonstrate that selection of input data and method has an important effect on the clustering result. In addition, our method is helpful for improving the resolution of seismic facies edges and offers the richer depositional information than the traditional clustering methods.

Semi-Supervised Deep Time-Delay Embedded Clustering for Stress Speech Analysis

Real stressed speech is affected by various aspects (individual characteristics and environment) so that the stress patterns are diverse and different on each individual. To this end, in our previous work, we performed an unsupervised clustering method that able to self-learning manner by mapping the feature representations of the stress speech and clustering tasks simultaneously, called deep time-delay embedded clustering (DTEC). However, DTEC has not confirmed yet the compatibility between the output class and informational classes. Therefore, we proposed semi-supervised time-delay embedded clustering (SDTEC) as a new framework of semi-supervised in DTEC. SDTEC incorporates the prior information of pairwise constraints in the embedding layer and simultaneously learns the feature representation and the clustering assignments. The prior information was used to guide the clustering procedure so that the points that belong to the incorrect cluster can be corrected. The effectiveness of the proposed SDTEC was evaluated by comparing it with some baseline methods in terms of the clustering error rate (CER). Moreover, to demonstrate SDTEC’s capabilities, we conducted a comprehensive ablation study. Based on experiment results, SDTEC outperformed the baseline methods and achieves state-of-the-art results in semi-supervised clustering.

Acoustic Scene Clustering Using Joint Optimization of Deep Embedding Learning and Clustering Iteration

Recent efforts have been made on acoustic scene classification in the audio signal processing community. In contrast, few studies have been conducted on acoustic scene clustering, which is a newly emerging problem. Acoustic scene clustering aims at merging the audio recordings of the same class of acoustic scene into a single cluster without using prior information and training classifiers. In this study, we propose a method for acoustic scene clustering that jointly optimizes the procedures of feature learning and clustering iteration. In the proposed method, the learned feature is a deep embedding that is extracted from a deep convolutional neural network (CNN), while the clustering algorithm is the agglomerative hierarchical clustering (AHC). We formulate a unified loss function for integrating and optimizing these two procedures. Various features and methods are compared. The experimental results demonstrate that the proposed method outperforms other unsupervised methods in terms of the normalized mutual information and the clustering accuracy. In addition, the deep embedding outperforms many state-of-the-art features.

Hands-Free User Interface for AR/VR Devices Exploiting Wearer's Facial Gestures Using Unsupervised Deep Learning.

Developing a user interface (UI) suitable for headset environments is one of the challenges in the field of augmented reality (AR) technologies. This study proposes a hands-free UI for an AR headset that exploits facial gestures of the wearer to recognize user intentions. The facial gestures of the headset wearer are detected by a custom-designed sensor that detects skin deformation based on infrared diffusion characteristics of human skin. We designed a deep neural network classifier to determine the user’s intended gestures from skin-deformation data, which are exploited as user input commands for the proposed UI system. The proposed classifier is composed of a spatiotemporal autoencoder and deep embedded clustering algorithm, trained in an unsupervised manner. The UI device was embedded in a commercial AR headset, and several experiments were performed on the online sensor data to verify operation of the device. We achieved implementation of a hands-free UI for an AR headset with average accuracy of 95.4% user-command recognition, as determined through tests by participants.

Automatic Identification of Product Usage Contexts from Online Customer Reviews

AbstractThere are three product design contexts that may significantly affect the design of a product and customer preferences towards product attributes, i.e. customer context, market context, and usage context factors. The conventional methods to gather product usage contexts may be costly and time consuming to conduct. As an alternative, this paper aims to automatically identify product usage contexts from publicly available online customer reviews. The proposed methodology consists of Preprocessing, Word Embedding, and Usage Context Clustering stages. The methodology is applied to identify usage contexts from laptop customer reviews, which results in 16 clusters of usage contexts. Furthermore, analyzing the review sentences explains the separation of “playing games” –which is more related to casual gaming, and “gaming rig” –which implies high computing power requirements. Finally, comparing customer review with manufacturer's product description may reveal a discrepancy to be investigated further by product designer, e.g. a customer suggests a laptop for basic use, although the manufacturer's description describes it for heavy use.

Clustering Continuous Wavelet Transform Characteristics of Heart Rate Variability through Unsupervised Learning.

The analysis and interpretation of physiological signals acquired non-invasively are increasingly important in Smart Health, precision medicine, and medical research. However, this analysis is hampered due to the length, complexity, and inter-subject variation of these signals, and, consequently, dimensionality reduction and clustering offer substantial benefits. Machine learning, used widely in biomedicine, is increasingly being applied to physiological time series. Among the applications of unsupervised learning, clustering is one of the most important. In this paper, an unsupervised autoen-coder architecture, deep convolutional embedded clustering, is presented as a data-driven approach to study time-frequency characteristics of heart rate variability records. An autoen-coder network is trained on continuous wavelet transforms of heart rate variability signals calculated from publicly-available annotated ECG records with a wide variety of conditions. The latent variables learned by the clustering autoencoder are low-dimensional representations of wavelet transform characteristics that can be visualized and further analyzed. The results indicate that the learned clusters correspond to beat morphologies in the electrocardiogram in many cases, but also that the reduced dimensions of the time-frequency features can potentially provide additional insights into cardiac activity and the autonomic nervous system.

Pattern detection from seating pressure distribution during wheelchair motion using deep embedded clustering.

To minimize the occurrence of ulcers in high-risk mobile individuals such as wheelchair users, it is necessary to detect all typical distribution patterns and to indentify the patterns that may be associated with pressure ulcers. However, pattern detection is difficult because the pressure distribution during motion includes a variety of patterns compared to those of static postures. Thus, the establishment of a method to detect typical patterns based on distribution patterns is important. We utilized deep embedded clustering for identification purposes. This clustering technique extracts features using auto-encoder and simultaneously optimizes data points into the clusters, which might realize good clustering performance due to the detected optimal features. We used a pressure distribution dataset that was pre-labeled by nursing experts. The dataset consisted of a total of 26944 distribution images with ten class annotations. The clustering method including traditional approaches (k-means and principal component analysis plus k-means) were compared with deep embedded clustering while the threshold to noise reduction was changing. The deep embedded clustering with 80 mmHg threshold achieved the best performance. This approach also tended to be less dependent on the threshold values.

Abstract 1635: A scalable, cloud-based, unsupervised deep learning system for identification, extraction, and summarization of potentially imperceptible patterns in whole-slide images of breast cancer tissue

Abstract Digital pathology images potentially contain novel patterns that may be perceived by modern deep learning models, but not humans. Prior unsupervised pattern recognition approaches have been used to reveal prognostically-relevant subtypes of glioblastoma (PMID: 28984190) and breast density segmentation (PMID: 26915120), and may complement supervised machine learning models trained using labeled data. In the Cancer Prevention Study II (CPS-II) cohort (PMID: 12015775), high-resolution, digitized hemotoxylin and eosin diagnostic slides are available for approximately 1,700 breast cancer cases providing an opportunity to perform unsupervised pattern recognition image analysis for epidemiologic breast cancer studies. Given the size of the dataset and complexity of the models, we constructed an end-to-end analytical pipeline, including preprocessing, feature engineering, and clustering, using cloud-based technologies that enable analysis at scale. Prior to training the unsupervised models, we faced issues converting raw images with open-source software. Specifically, OpenSlides could not open the Leica Versa SCN files due to their proprietary format while BioFormats inverted colors. To fix these issues, we altered the BioFormats library to successfully convert the files into a TIFF format. Since this issue likely affects other researchers, we are in discussions to provide the fix under a public license. TIFF formatted images were then denoised through color normalization to reduce hue variance and artifact detection to remove unwanted features such as pathologist annotations. Due to the computational complexity of analyzing the full image, images were padded with white space to ensure divisibility and broken into nine tiles of a predefined size. To further reduce computation time, uninformative tiles were filtered based on a predetermined threshold of artifact and white space composition. The remaining tiles were input to the unsupervised models. We used convolutional autoencoders, specifically a modified VGG-16 model without pretrained weights and a deep embedded clustering algorithm. These models learn representations of the images called ‘feature vectors’ and encode the images’ salient patterns. The final model was chosen based on iterative testing on a subsample of 100 images (N=21,472 tiles) and performance comparison of various VGG-inspired autoencoders. The feature vectors were clustered by K-means to summarize the information in a format suitable for statistical analyses. Our initial results show that the system captures macro-scale tissue patterns at lower magnifications (1x and 5x) and produces clusters that can be integrated into epidemiological studies of breast cancer etiology and prognosis. Citation Format: Jacob L. Evans, William Seo, Mary Macheski-Preston, Michelle Fritz, Samantha Puvanesarajah, James Hodge, Ted Gansler, Susan Gapstur, Mia M. Gaudet, Michelle Yi. A scalable, cloud-based, unsupervised deep learning system for identification, extraction, and summarization of potentially imperceptible patterns in whole-slide images of breast cancer tissue [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 1635.

Static malware clustering using enhanced deep embedding method

SummaryMalware refers to any software, programs, or files that are intentionally utilised to compromise the system and cause unexpected losses to end‐users such as economical losses or privacy breaches. The rapid growth of malware makes it impossible to keep up with its progress merely via human interventions or manual analysis. One of the challenges for the human‐oriented approaches is they will cause backlog and inability to keep up with the development traces of the malware. Hence, an efficient method is needed urgently to analyse effectively and identify accurately the malware in their domain. Malware clustering has been extensively studied in the machine learning area with regards to distance functions, grouping algorithm and cluster validation. A large number of research studies have been done via behavioral analysis for clustering to achieve high performance of malware detections. However, there is a trade‐off for better detection performance between behaviorial approaches and high computational forces. Up to date, little work focuses on the deep learning representations for malware clustering. Therefore, in this paper, we propose an enhanced deep embedded clustering method to facilitate an effective and efficient malware clustering process. The new method takes advantage of linear dimensionality reduction and a customised deep neural network to learn malware representations in an orthogonal space and performs cluster assignments. Our experimental results demonstrate that the proposed clustering model outperforms the traditional K‐means method with regards to the enhanced features using various auto‐encoder, pre‐trained weight and principle component analysis (PCA).

Clustering single-cell RNA-seq data with a model-based deep learning approach

Single-cell RNA sequencing (scRNA-seq) promises to provide higher resolution of cellular differences than bulk RNA sequencing. Clustering transcriptomes profiled by scRNA-seq has been routinely conducted to reveal cell heterogeneity and diversity. However, clustering analysis of scRNA-seq data remains a statistical and computational challenge, due to the pervasive dropout events obscuring the data matrix with prevailing ‘false’ zero count observations. Here, we have developed scDeepCluster, a single-cell model-based deep embedded clustering method, which simultaneously learns feature representation and clustering via explicit modelling of scRNA-seq data generation. Based on testing extensive simulated data and real datasets from four representative single-cell sequencing platforms, scDeepCluster outperformed state-of-the-art methods under various clustering performance metrics and exhibited improved scalability, with running time increasing linearly with sample size. Its accuracy and efficiency make scDeepCluster a promising algorithm for clustering large-scale scRNA-seq data. Clustering groups of cells in single-cell RNA sequencing datasets can produce high-resolution information for complex biological questions. However, it is statistically and computationally challenging due to the low RNA capture rate, which results in a high number of false zero count observations. A deep learning approach called scDeepCluster, which efficiently combines a model for explicitly characterizing missing values with clustering, shows high performance and improved scalability with a computing time increasing linearly with sample size.

Unsupervised classification of multi-omics data during cardiac remodeling using deep learning.

Integration of multi-omics in cardiovascular diseases (CVDs) presents high potentials for translational discoveries. By analyzing abundance levels of heterogeneous molecules over time, we may uncover biological interactions and networks that were previously unidentifiable. However, to effectively perform integrative analysis of temporal multi-omics, computational methods must account for the heterogeneity and complexity in the data. To this end, we performed unsupervised classification of proteins and metabolites in mice during cardiac remodeling using two innovative deep learning (DL) approaches. First, long short-term memory (LSTM)-based variational autoencoder (LSTM-VAE) was trained on time-series numeric data. The low-dimensional embeddings extracted from LSTM-VAE were then used for clustering. Second, deep convolutional embedded clustering (DCEC) was applied on images of temporal trends. Instead of a two-step procedure, DCEC performes a joint optimization for image reconstruction and cluster assignment. Additionally, we performed K-means clustering, partitioning around medoids (PAM), and hierarchical clustering. Pathway enrichment analysis using the Reactome knowledgebase demonstrated that DL methods yielded higher numbers of significant biological pathways than conventional clustering algorithms. In particular, DCEC resulted in the highest number of enriched pathways, suggesting the strength of its unified framework based on visual similarities. Overall, unsupervised DL is shown to be a promising analytical approach for integrative analysis of temporal multi-omics.

Machine learning analysis of gene expression data reveals novel diagnostic and prognostic biomarkers and identifies therapeutic targets for soft tissue sarcomas.

Based on morphology it is often challenging to distinguish between the many different soft tissue sarcoma subtypes. Moreover, outcome of disease is highly variable even between patients with the same disease. Machine learning on transcriptome sequencing data could be a valuable new tool to understand differences between and within entities. Here we used machine learning analysis to identify novel diagnostic and prognostic markers and therapeutic targets for soft tissue sarcomas. Gene expression data was used from the Cancer Genome Atlas, the Genotype-Tissue Expression project and the French Sarcoma Group. We identified three groups of tumors that overlap in their molecular profiles as seen with unsupervised t-Distributed Stochastic Neighbor Embedding clustering and a deep neural network. The three groups corresponded to subtypes that are morphologically overlapping. Using a random forest algorithm, we identified novel diagnostic markers for soft tissue sarcoma that distinguished between synovial sarcoma and MPNST, and that we validated using qRT-PCR in an independent series. Next, we identified prognostic genes that are strong predictors of disease outcome when used in a k-nearest neighbor algorithm. The prognostic genes were further validated in expression data from the French Sarcoma Group. One of these, HMMR, was validated in an independent series of leiomyosarcomas using immunohistochemistry on tissue micro array as a prognostic gene for disease-free interval. Furthermore, reconstruction of regulatory networks combined with data from the Connectivity Map showed, amongst others, that HDAC inhibitors could be a potential effective therapy for multiple soft tissue sarcoma subtypes. A viability assay with two HDAC inhibitors confirmed that both leiomyosarcoma and synovial sarcoma are sensitive to HDAC inhibition. In this study we identified novel diagnostic markers, prognostic markers and therapeutic leads from multiple soft tissue sarcoma gene expression datasets. Thus, machine learning algorithms are powerful new tools to improve our understanding of rare tumor entities.

On the use of 2D moment invariants in the classification of additive manufacturing powder feedstock

The shapes of near-spherical individual powder particles used as additive manufacturing feedstock are analyzed by means of moment invariants in combination with dimensionality reduction techniques, including t-distributed stochastic neighbor embedding (t-SNE) and hierarchical density-based clustering with noise (HDBSCAN). Affine Cartesian invariants up to the 12th order are found to be the most effective shape descriptors, in particular when used with t-SNE mapping. The methodology described in this paper can capture outlier particle shapes, can distinguish between very similar particle shapes, and has the potential to be implemented as a fully automated classification process to assess to what extent the distribution of the invariants deviates from the expected distribution for a powder that is known to have good additive manufacturing properties.

Semi Supervised Learning with Deep Embedded Clustering for Image Classification and Segmentation.

Deep neural networks usually require large labeled datasets to construct accurate models; however, in many real-world scenarios, such as medical image segmentation, labelling data is a time-consuming and costly human (expert) intelligent task. Semi-supervised methods leverage this issue by making use of a small labeled dataset and a larger set of unlabeled data. In this article, we present a flexible framework for semi-supervised learning that combines the power of supervised methods that learn feature representations using state-of-the-art deep convolutional neural networks with the deep embedded clustering algorithm that assigns data points to clusters based on their probability distributions and feature representations learned by the networks. Our proposed semi-supervised learning algorithm based on deep embedded clustering (SSLDEC) learns feature representations via iterations by alternatively using labeled and unlabeled data points and computing target distributions from predictions. During this iterative procedure the algorithm uses labeled samples to keep the model consistent and tuned with labeling, as it simultaneously learns to improve feature representation and predictions. SSLDEC requires few hyper-parameters and thus does not need large labeled validation sets, which addresses one of the main limitations of many semi-supervised learning algorithms. It is also flexible and can be used with many state-of-the-art deep neural network configurations for image classification and segmentation tasks. To this end, we implemented and tested our approach on benchmark image classification tasks as well as in a challenging medical image segmentation scenario. In benchmark classification tasks, SSLDEC outperformed several state-of-the-art semi-supervised learning methods, achieving 0.46% error on MNIST with 1000 labeled points, and 4.43% error on SVHN with 500 labeled points. In the iso-intense infant brain MRI tissue segmentation task, we implemented SSLDEC on a 3D densely connected fully convolutional neural network where we achieved significant improvement over supervised-only training as well as a semi-supervised method based on pseudo-labelling. Our results show that SSLDEC can be effectively used to reduce the need for costly expert annotations, enhancing applications such as automatic medical image segmentation.

Deep Embedded Clustering With Adversarial Distribution Adaptation

Clustering methods based on deep neural networks have been extensively studied and applied in data mining. Most existing unsupervised deep embedded clustering methods jointly learn feature representations and cluster assignments with deep neural networks. However, to the best of our knowledge, most of them ignore the relevance between representation learning and clustering, so that fail to couple feature representation learning and cluster assignments effectively, which leads to non-representative features for clustering and this in turn hurts clustering performance. In this paper, we propose Adversarial Deep Embedded Clustering (ADEC), a novel unsupervised clustering method based on adversarial auto-encoder (AAE) and k -means clustering method. Specifically, ADEC matches distribution of feature representations with the given prior distribution to preserve the data structure with AAE, and k -means based on distribution distance metrics is conducted for clustering with these feature representations, simultaneously. ADEC optimizes a clustering objective iteratively with backpropagation algorithm in learning AAE from data space to feature space. Extensive experiments have demonstrated the effectiveness of our proposed method on images clustering compared to several strong baseline methods.

A graph-based lesion characterization and deep embedding approach for improved computer-aided diagnosis of nonmass breast MRI lesions.

Nonmass-like enhancements are a common but diagnostically challenging finding in breast MRI. Nonmass-like lesions can be described as clusters of spatially and temporally inter-connected regions of enhancements, so they can be modeled as networks and their properties characterized via network-based connectivity. In this work, we represented nonmass lesions as graphs using a link formation energy model that favors linkages between regions of similar enhancement and closer spatial proximity. However, adding graph features to an existing computer-aided diagnosis (CAD) pipeline incurs an increase of feature space dimensionality, which poses additional challenges to traditional supervised machine learning techniques due to the inability to increase accordingly the number of training datasets. We propose the combination of unsupervised dimensionality reduction and embedded space clustering followed by a supervised classifier to improve the performance of a CAD system for nonmass-like lesions in breast MRI. Our work extends a previoulsy proposed framework for deep embedded unsupervised clustering (DEC) to embedding space classification, with the joint optimization of objective functions for DEC and supervised multi-layered perceptron (MLP) classification. The strength of the method lies in the ability to learn and further optimize an embedded feature representation of lower dimensionality that maximizes the diagnostic accuracy of a CAD lesion classifier to discriminate between benign and malignant lesions. We identified 792 nonmass-like enhancements (267 benign, 110 malignant and 415 unknown) in 411 patients undergoing breast MRI at our institution. The diagnostic performance of the proposed method was evaluated and compared to the performance of a conventional supervised MLP classifier in original feature space. A statistically significant increase in diagnostic area under the ROC curve (AUC) was achieved. Generalization AUC increased from 0.67 ± 0.08 to 0.81 ± 0.10 (21% increase, p-value=4.2×10-8) with the proposed graph-based lesion characterization and deep embedding framework.