Problem Of Automatic Segmentation Research Articles

Deep learning muscle segmentation model for CT images in DICOM format

This work solves the problem of automatic segmentation of medical images in DICOM format using machine learning methods. A new developed tool is used in the form of a separate module for labeling medical data in the DICOM format. The trained model, proposed in the paper, can be useful in the tasks of muscle segmentation. One can apply it in different ways, but some of the most common include assessment of diseases related to muscles, and sarcopenia is one of them. The further applications of the muscle segmentation model may include examining various medical cases with patients, that tend to have muscle-related diseases. For instance, detecting cachexia may be one of the extensions of the model’s application field.

Methodical approaches to annotation and labeling of histological images in order to automatically detect the layers of the stomach wall and the depth of invasion of gastric cancer

Development of original methodological approaches to annotation and labeling of histological images in relation to the problem of automatic segmentation of the layers of the stomach wall. Three image collections were used in the study: NCT-CRC-HE-100K, CRC-VAL-HE-7K, and part of the PATH-DT-MSU collection. The used part of the original PATH-DT-MSU collection contains 20 histological images obtained using a high performance digital scanning microscope. A labeling system is proposed that includes 5 classes (tissue types): areas of gastric adenocarcinoma (TUM), unchanged areas of the lamina propria (LP), unchanged areas of the muscular lamina of the mucosa (MM), a class of underlying tissues (AT), including areas of the submucosa, own muscular layer of the stomach and subserous sections, image background (BG). The advantage of this marking technique is to ensure high efficiency of recognition of the muscularis lamina (MM) - a natural «line» separating the lamina propria of the mucous membrane and all other underlying layers of the stomach wall. The disadvantages of the technique include a small number of classes, which leads to insufficient detailing of automatic segmentation. In the course of the study, an original technique for labeling and annotating images was developed, including 5 classes (types of tissues). This technique is effective at the initial stages of teaching mathematical algorithms for the classification and segmentation of histological images. Further stages in the development of a real diagnostic algorithm to automatically determine the depth of invasion of gastric cancer require the correction and development of the presented method of marking and annotation.

Analysis of the latent space of pre-trained deep convolutional neural networks in the problem of automatic segmentation of color images

The paper presents a primary study of the latent space structure of neural networks trained for semantic segmentation. Segmentation was performed in a controlled environment of three classes of colored rectangular shapes. The classic autoencoder and U-net like architectures were chosen as reference architectures. To study the structure of the space, a combination of a perceptron that linearly separates classes and the compression algorithms UMAP and PCA was used. As a result, a tool was obtained for evaluating the quality of a neural network based on the degree of separability of classes in the latent space of the network.

Generation of echocardiographic 2D images of the heart using cGAN

One of the most significant tasks of echocardiography is the automatic delineation of the cardiac structures from 2D echocardiographic images. Over the past decades, the automation of this task has been the subject of intense research. One of the most effective approaches is based on the deep convolutional neural networks. Nonetheless, it is necessary to use echocardiogram frames of the cardiac muscle, which show the boundaries of the cardiac structures labeled/annotated by experts/cardiologists to train it. However, the number of databases containing the necessary information is relatively small. Therefore, generated echocardiogram frames are used to increase the amount of training samples. This process is based on the ultrasound images of the heart, annotated by experts. The article proposes an improved method for generating echocardiograms using a generative adversarial neural network (GAN) with a patch-based conditional discriminator. It has been demonstrated that it is possible to improve the quality of generated echocardiogram frames in both two and four chamber views (AP4C, AP2C) using the masks of cardiac segmentation with sub-pixel convolution layer (pixel shuffle). It is demonstrated that the proposed approach makes it possible to generate ultrasound images, the structure of which corresponds to the specified segmentation masks. It is expected that this method will improve the accuracy of solving the direct problem of automatic segmentation of the left ventricle.

Guaranteed Significance Level Criterion in Automatic Speech Signal Segmentation

The article considers the problem of automatic segmentation of a speech signal into phonetic units in conditions of their a priori uncertain spectral composition and correlation properties. A guaranteed significance level criterion is developed based on the information–theoretic approach. An example of practical application of this criterion is considered; a full-scale experiment is set up and conducted. It is shown that the proposed criterion can guarantee a stable significance level when processing speech frames of short duration.

The state of the art in kidney and kidney tumor segmentation in contrast-enhanced CT imaging: Results of the KiTS19 challenge.

There is a large body of literature linking anatomic and geometric characteristics of kidney tumors to perioperative and oncologic outcomes. Semantic segmentation of these tumors and their host kidneys is a promising tool for quantitatively characterizing these lesions, but its adoption is limited due to the manual effort required to produce high-quality 3D segmentations of these structures. Recently, methods based on deep learning have shown excellent results in automatic 3D segmentation, but they require large datasets for training, and there remains little consensus on which methods perform best. The 2019 Kidney and Kidney Tumor Segmentation challenge (KiTS19) was a competition held in conjunction with the 2019 International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI) which sought to address these issues and stimulate progress on this automatic segmentation problem. A training set of 210 cross sectional CT images with kidney tumors was publicly released with corresponding semantic segmentation masks. 106 teams from five continents used this data to develop automated systems to predict the true segmentation masks on a test set of 90 CT images for which the corresponding ground truth segmentations were kept private. These predictions were scored and ranked according to their average Sørensen-Dice coefficient between the kidney and tumor across all 90 cases. The winning team achieved a Dice of 0.974 for kidney and 0.851 for tumor, approaching the inter-annotator performance on kidney (0.983) but falling short on tumor (0.923). This challenge has now entered an "open leaderboard" phase where it serves as a challenging benchmark in 3D semantic segmentation.

Full anatomical labeling of magnetic resonance images of human brain by registration with multiple atlases

The problem of automatic segmentation of magnetic resonance (MR) images of human brain into anatomical structures is considered. Currently, the most popular segmentation algorithms are based on the registration (matching) of the input image with (to) an atlas--an image for which an expert labeling is known. Segmentation on the basis of registration with multiple atlases allows one to better take into account anatomical variability and thereby to compensate, to some extent, for the errors of matching to each individual atlas. In this work, a more efficient (in speed and memory) implementation is proposed of one of the best multiatlas label fusion algorithms in order to obtain a labeling of the input image. The algorithm is applied to the problem of segmentation of brain MR images into 43 anatomical regions with the use of the publicly available IBSR database, in contrast to the original work, where the authors provide test results for the problem of extraction of a single anatomical structure, the hippocampus.

Usul and Makam driven automatic melodic segmentation for Turkish music

Automatic melodic segmentation is a topic studied extensively, aiming at developing systems that perform grouping of musical events. Here, we consider the problem of automatic segmentation via supervised learning from a dataset containing segmentation labels of an expert. We present a statistical classification-based segmentation system developed specifically for Turkish makam music. The proposed system uses two novel features, a makam-based and an usul-based feature, together with features commonly used in literature. The makam-based feature is defined as the probability of a note to appear at the phrase boundary, computed from the distributions of boundaries with respect to the piece’s makam pitches. Likewise, the usul-based feature is computed from the distributions of boundaries with respect to beats in the rhythmic cycle, usul of the piece. Several experimental setups using different feature groups are designed to test the contribution of the proposed features on three datasets. The results show that the new features carry complementary information to existing features in the literature within the Turkish makam music segmentation context and that the inclusion of new features resulted in statistically significant performance improvement.

3D Segmentation for Multi-Organs in CT Images

The study addresses the challenging problem of automatic segmentation of the human anatomy needed for radiation dose calculations.Three-dimensional extensions of two well-known state-of-the art segmentation techniques are proposed and tested for usefulness on a set of clinical CT images.The new techniques are 3D Statistical Region Merging (3D-SRM) and 3D Efficient Graph-based Segmentation (3D-EGS). Segmentations of eight representative tissues (lungs, stomach, liver, heart, kidneys, spleen, bones and the spinal cord)were tested for accuracy using the Dice index, the Hausdorff distance and the $H_t$ index. The 3D-SRM outperformed 3D-EGS producing the average(across the 8 tissues) Dice index, the Hausdorff distance, and the $H_2$ of $0.89$, $12.5$~mm and $0.93$, respectively.

Biological time series segmentation using dynamic neural network model

The problem of automatic segmentation of the laboratory animal behavior into behavioral acts is considered as a problem of multidimensional time series segmentation. Formal statement of the problem with use of variables characterizing reliability of accepted decisions is expounded. A neural model based on NARX-networks has been developed for automatic segmentation of multidimensional time series. Experimental results received with use of biological test data are discussed.

A Two-stage Prosodic Structure Generation Strategy for Mandarin Text-to-speech Systems

Prosodic structure generation is the key component in improving the intelligibility and naturalness of synthetic speech for a text-to-speech (TTS) system. This paper investigates the problem of automatic segmentation of prosodic word and prosodic phrase,which are two fundamental layers in the hierarchical prosodic structure of Mandarin,and presents a two-stage prosodic structure generation strategy. Conditional random fields (CRF) models are built for both prosodic word and prosodic phrase prediction at the front end with diflerent feature selections. Besides,a transformation-based error-driven learning (TBL) modification module is introduced in the back end to amend the initial prediction. Experiment results show that the approach combining CRF and TBL achieves an F-score of 94.66%.

A Two-stage Prosodic Structure Generation Strategy for Mandarin Text-to-speech Systems

Abstract Prosodic structure generation is the key component in improving the intelligibility and naturalness of synthetic speech for a text-to-speech (TTS) system. This paper investigates the problem of automatic segmentation of prosodic word and prosodic phrase, which are two fundamental layers in the hierarchical prosodic structure of Mandarin, and presents a two-stage prosodic structure generation strategy. Conditional random fields (CRF) models are built for both prosodic word and prosodic phrase prediction at the front end with different feature selections. Besides, a transformation-based error-driven learning (TBL) modification module is introduced in the back end to amend the initial prediction. Experiment results show that the approach combining CRF and TBL achieves an F-score of 94.66 %.

Automatic segmentation of stochastic time series using a whitening filter

The problem of automatic segmentation of a stochastic time series into homogeneous data segments one cluster long is posed and solved according to the general formulation of the disorder problem. A new algorithm with cluster normalization by the variance of the generating noise is developed using an autoregression model and the minimum information divergence criterion. It is shown that the main advantage of the proposed algorithm over existing analogs is high dynamic properties. The algorithm was tested by analyzing stock market dynamics in the United States and Russia. Estimates of the admissible (threshold) level of disorder in a time series within one cluster are obtained using the Kullback-Leibler information measure.

Model-based segmentation of short-axis MR cardiac images

Reliable automatic segmentation of MR cardiac images is still an important problem in medical image processing. Although image data quality has improved considerably during the last years, this segmentation is still considered a difficult problem. Manual segmentation is hardly an option as this is extremely labour intensive. This paper describes a combined bottom-up and top-down approach to the problem of automatic segmentation of short-axis MR cardiac images, i.e. extract the boundaries of the left and right ventricle and the myocardium. The method is evaluated using 69 datasets for which manual segmentations were available.

Automatic segmentation of news items based on video and audio features

The automatic segmentation of news items is a key for implementing the automatic cataloging system of news video. This paper presents an approach which manages audio and video feature information to automatically segment news items. The integration of audio and visual analyses can overcome the weakness of the approach using only image analysis techniques. It makes the approach more adaptable to various situations of news items. The proposed approach detects silence segments in accompanying audio, and integrates them with shot segmentation results, as well as anchor shot detection results, to determine the boundaries among news items. Experimental results show that the integration of audio and video features is an effective approach to solving the problem of automatic segmentation of news items.

Modelling and segmentation of lip area in face images

The problem of automatic segmentation of face images to extract the lip area is considered in the context of lip tracking. A novel segmentation method is proposed for lip tracker initialisation which is based on a Gaussian mixture model of the pixel RGB values. The model is built using the predictive validation technique advocated by Kittler which has been modified to allow modelling with full, rather than just diagonal covariance matrices. A subsequent grouping of the mixture components provides the basis for a Bayesian rule labelling of the pixels as lip or non-lip. The proposed method is tested on a database of 145 images and demonstrates that its accuracy is significantly better than the segmentation obtained by k-means clustering. Moreover, the proposed method does not require the number of segments to be specified a priori.

CURVES: Curve evolution for vessel segmentation

The vasculature is of utmost importance in neurosurgery. Direct visualization of images acquired with current imaging modalities, however, cannot provide a spatial representation of small vessels. These vessels, and their branches which show considerable variations, are most important in planning and performing neurosurgical procedures. In planning they provide information on where the lesion draws its blood supply and where it drains. During surgery the vessels serve as landmarks and guidelines to the lesion. The more minute the information is, the more precise the navigation and localization of computer guided procedures. Beyond neurosurgery and neurological study, vascular information is also crucial in cardiovascular surgery, diagnosis, and research. This paper addresses the problem of automatic segmentation of complicated curvilinear structures in three-dimensional imagery, with the primary application of segmenting vasculature in magnetic resonance angiography (MRA) images. The method presented is based on recent curve and surface evolution work in the computer vision community which models the object boundary as a manifold that evolves iteratively to minimize an energy criterion. This energy criterion is based both on intensity values in the image and on local smoothness properties of the object boundary, which is the vessel wall in this application. In particular, the method handles curves evolving in 3D, in contrast with previous work that has dealt with curves in 2D and surfaces in 3D. Results are presented on cerebral and aortic MRA data as well as lung computed tomography (CT) data.

Automatic segmentation of speech recorded in unknown noisy channel characteristics

This paper investigates the problem of automatic segmentation of speech recorded in noisy channel corrupted environments. Using an HMM-based speech segmentation algorithm, speech enhancement and parameter compensation techniques previously proposed for robust speech recognition are evaluated and compared for improved segmentation in colored noise. Speech enhancement algorithms considered include: Generalized Spectral Subtraction, Nonlinear Spectral Subtraction, Ephraim–Malah MMSE enhancement, and Auto-LSP Constrained Iterative Wiener filtering. In addition, the Parallel Model Combination (PMC) technique is also compared for additive noise compensation. In telephone environments, we compare channel normalization techniques including Cepstral Mean Normalization (CMN) and Signal Bias Removal (SBR) and consider the coupling of channel compensation with front-end speech enhancement for improved automatic segmentation. Compensation performance is assessed for each method by automatically segmenting TIMIT degraded by additive colored noise (i.e., aircraft cockpit, automobile highway, etc.), telephone transmitted NTIMIT, and cellular telephone transmitted CTIMIT databases.

A mathematical approach to the segmentation of sea-ice SAR images

Abstract This paper discusses basic mathematical formulations to clarify the problem of automatic segmentation of sea-ice remote sensing imagery. The work is illustrated with a SAR image. The main issue is a robust and automatic analysis of sea-ice conditions from remotely sensed data. The paper reveals some of the potential and essential features that are hidden in a class of filters often known as simple ‘shrink and expand operators’. The intensity of radiation from the ground is idealized as a measure which may have singularities. Absolutely calibrated digital images represent this measure where the measurable set are restricted to unions of picture elements. A related proposed martingale stopping time scheme makes image segmentation robust to changes in resolution. Tunable segmentations based on mathematical morphology can be controlled to give optimal ‘physically proper’ sets which are independent of calibration. This approach introduces physical knowledge, time aspects, and random geometry to the pr...