Search

Melanoma is one of the most lethal and rapidly growing cancers, causing many deaths each year. This cancer can be treated effectively if it is detected quickly. For this reason, many algorithms and systems have been developed to support automatic or semiautomatic detection of neoplastic skin lesions based on the analysis of optical images of individual moles. Recently, full-body systems have gained attention because they enable the analysis of the patient’s entire body based on a set of photos. This paper presents a prototype of such a system, focusing mainly on assessing the effectiveness of algorithms developed for the detection and segmentation of lesions. Three detection algorithms (and their fusion) were analyzed, one implementing deep learning methods and two classic approaches, using local brightness distribution and a correlation method. For fusion of algorithms, detection sensitivity = 0.95 and precision = 0.94 were obtained. Moreover, the values of the selected geometric parameters of segmented lesions were calculated and compared for all algorithms. The obtained results showed a high accuracy of the evaluated parameters (error of area estimation &lt, 10%), especially for lesions with dimensions greater than 3 mm, which are the most suspected of being neoplastic lesions.

This chapter addresses object measurements from 2D microscopy images. Our motivation is to introduce readers to image-based object measurements and to quantify numerical variability of image features and feature-based classification outcomes. The objective is to highlight the sources of digital image processing variability when deriving image-based scientific conclusions. By characterizing feature variations across Python scikit-image, CellProfiler, MaZda, ImageJ, and in-house Java libraries, we concluded 15.6% of 32 intensity features, 47.9% of 71 shape features, and 88.2% of 68 textural features differ in values. Shape and textural feature variations had a negative impact on classification outcomes in 52.9% and 97.1% of all single feature-based classifications, respectively. All these numerical results are available at https://isg.nist.gov/deepzoomweb/resources/featureVariability/index.html and are traceable to image objects. They document the amount of digital image processing noise and the importance of gathering computational provenance information.

Background: Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is an imaging technique which helps in visualizing and quantifying perfusion&mdash, one of the most important indicators of an organ&rsquo, s state. This paper focuses on perfusion and filtration in the kidney, whose performance directly influences versatile functions of the body. In clinical practice, kidney function is assessed by measuring glomerular filtration rate (GFR). Estimating GFR based on DCE-MRI data requires the application of an organ-specific pharmacokinetic (PK) model. However, determination of the model parameters, and thus the characterization of GFR, is sensitive to determination of the arterial input function (AIF) and the initial choice of parameter values. Methods: This paper proposes a multi-layer perceptron network for PK model parameter determination, in order to overcome the limitations of the traditional model&rsquo, s optimization techniques based on non-linear least-squares curve-fitting. As a reference method, we applied the trust-region reflective algorithm to numerically optimize the model. The effectiveness of the proposed approach was tested for 20 data sets, collected for 10 healthy volunteers whose image-derived GFR scores were compared with ground-truth blood test values. Results: The achieved mean difference between the image-derived and ground-truth GFR values was 2.35 mL/min/1.73 m2, which is comparable to the result obtained for the reference estimation method (&minus, 5.80 mL/min/1.73 m2). Conclusions: Neural networks are a feasible alternative to the least-squares curve-fitting algorithm, ensuring agreement with ground-truth measurements at a comparable level. The advantages of using a neural network are twofold. Firstly, it can estimate a GFR value without the need to determine the AIF for each individual patient. Secondly, a reliable estimate can be obtained, without the need to manually set up either the initial parameter values or the constraints thereof.

Image texture is a very important component in many types of images, including medical images. Medical images are often corrupted by noise and affected by artifacts. Some of the texture-based features that should describe the structure of the tissue under examination may also reflect, for example, the uneven sensitivity of the scanner within the tissue region. This in turn may lead to an inappropriate description of the tissue or incorrect classification. To limit these phenomena, the analyzed regions of interest are normalized. In texture analysis methods, image intensity normalization is usually followed by a reduction in the number of levels coding the intensity. The aim of this work was to analyze the impact of different image normalization methods and the number of intensity levels on texture classification, taking into account noise and artifacts related to uneven background brightness distribution. Analyses were performed on four sets of images: modified Brodatz textures, kidney images obtained by means of dynamic contrast-enhanced magnetic resonance imaging, shoulder images acquired as T2-weighted magnetic resonance images and CT heart and thorax images. The results will be of use for choosing a particular method of image normalization, based on the types of noise and distortion present in the images.

Texture can be a very rich source of information about the image. Texture analysis finds applications, among other things, in biomedical imaging. One of the widely used methods of texture analysis is the Gray Level Co-occurrence Matrix (GLCM). Texture analysis using the GLCM method is most often carried out in several stages: determination of areas of interest, normalization, calculation of the GLCM, extraction of features, and finally, the classification. Values of the GLCM based features depend on the choice of the normalization method, which was examined in this work. The normalization is necessary, since acquired images often suffer from noise and intensity artifacts. Certainly, the normalization will not eliminate these two effects, however it was demonstrated, that its application improves texture analysis accuracy. The aim of the work was to analyze the influence of different normalization methods on the discriminating ability of features estimated from the GLCM. The analysis was performed both for Brodatz textures and real magnetic resonance data. Brodatz textures were corrupted by three types of distortion: intensity nonuniformity, Gaussian noise and Rician Noise. Three types of normalizations were tested: min- max, 1–99% and $+/-3\sigma$ .

A novel interpolation-based model for the computation of the Gray Level Co-occurrence Matrix (GLCM) is presented. The model enables GLCM computation for any real-valued angles and offsets, as opposed to the traditional, lattice-based model. A texture directionality estimation algorithm is defined using the GLCM-derived correlation feature. The robustness of the algorithm with respect to image blur and additive Gaussian noise is evaluated. It is concluded that directionality estimation is robust to image blur and low noise levels. For high noise levels, the mean error increases but remains bounded. The performance of the directionality estimation algorithm is illustrated on fluorescence microscopy images of fibroblast cells. The algorithm was implemented in C++ and the source code is available in an openly accessible repository.

Texture is present in a large number of medical images. Its structure codes selected properties of visualized organ and tissues so texture can be rich source of information regarding their condition. Quantitative texture analysis plays significant role in imaging diagnosis support systems, enabling segmentation of analyzed organs, detection of lesions, and assessment of the degree of their pathological change. Unfortunately, medical images are often corrupted by noise which affect texture based image features. One of the steps of texture feature extraction is reduction of gray levels number which is performed after a normalization of pixel intensities inside a region of interest. This reduces the noise effect on texture feature values. We demonstrated, based on analysis of natural and MR images, that such reduction improves classification accuracy while reducing the computational costs.

Total hip replacement is the accepted treatment procedure of the end stage degeneration of the hip joint. Instability of the prosthesis might be recognized on the radiographic images as area of bone radio - lucency adjacent to the prosthesis pin. However, the very important issue of radiological recognition of periprosthetic lucent areas reflecting the lysis remains a challenge. Small dimensions and fuzzy borders of the lytic areas makes them difficult regions to recognize. Additional factors as high BMI of the patients and/or radiograms taken through a mattress can make the evaluation even more difficult, while small lucent areas might be additionally blurred and of very low contrast. The paper presents a new approach for quantitative recognition of preprothetic lytic areas. We have proposed a multistep algorithm utilizing Statistical Dominance Transform for detection of lytic areas on digital radiograms. Preliminary results are quite promising. It was demonstrated that location and shape of the detected lytic region is in good agreement with assessment by radiologists.

In brewing industry, quality of barley accepted for malt production is essential. The visual inspection of grain for malting is performed by a qualified expert. The process is time-consuming, expensive, and still may yield unreproducible results. Therefore, there is a need for automatic systems, based on computer vision, able to verify grain properties. We present a concept of such the system, which implements image preprocessing, texture, color and shape feature extraction, supervised learning and selected classification algorithms. The results of classification are presented and discussed.

Qmazda is a package of software tools for digital image analysis. They compute shape, color and texture attributes in arbitrary regions of interest, implement selected algorithms of discriminant analysis and machine learning, and enable texture based image segmentation. The algorithms generalize a concept of texture to three-dimensional data to enable analysis of volumetric images from magnetic resonance imaging or computed tomography scanners. The tools support a complete workflow — from image examples as an input to classification rules as an output. The extracted knowledge can be further used in custom made image analysis systems. Here we also present an application of QMaZda to identify defective barley kernels. The cereal seeds variability is high, therefore, characterization and discriminant analysis of such the biological objects is challenging and non-trivial. The software is available free of charge and open source, with executables for Windows, Linux and OS X platforms.

Summary We present a new method for segmenting phase contrast images of NIH 3T3 fibroblast cells that is accurate even when cells are physically in contact with each other. The problem of segmentation, when cells are in contact, poses a challenge to the accurate automation of cell counting, tracking and lineage modelling in cell biology. The segmentation method presented in this paper consists of (1) background reconstruction to obtain noise-free foreground pixels and (2) incorporation of biological insight about dividing and nondividing cells into the segmentation process to achieve reliable separation of foreground pixels defined as pixels associated with individual cells. The segmentation results for a time-lapse image stack were compared against 238 manually segmented images (8219 cells) provided by experts, which we consider as reference data. We chose two metrics to measure the accuracy of segmentation: the ‘Adjusted Rand Index’ which compares similarities at a pixel level between masks resulting from manual and automated segmentation, and the ‘Number of Cells per Field’ (NCF) which compares the number of cells identified in the field by manual versus automated analysis. Our results show that the automated segmentation compared to manual segmentation has an average adjusted rand index of 0.96 (1 being a perfect match), with a standard deviation of 0.03, and an average difference of the two numbers of cells per field equal to 5.39% with a standard deviation of 4.6%.

There is no suitable standard for the detection of blue whitish veil atypical vascular pattern and regression structures applied to skin lesion images. This information however is important in assessment of melanoma in skin dermatoscopic images. Thus there is a need for development of image analysis techniques that satisfy at least subjective criteria required by dermatologists. In this paper the application of color based image features for detection of blue whitish veil and atypical vas-cular pattern is presented. Preliminary test results are promising; for analyzed melanoma images the accuracy of developed methods provides 78 % correctly detected blue whitish veils, 84 % correctly detected atypical vascular pattern, and 86,5 % correctly detected regression structures. This paper is a contribution to the computer aided diagnostic system implementing the ELM 7-point check-list aimed at melanoma detection.

There is no suitable golden standard for the detection of atypical pigment network and irregular streaks applied to skin lesion images. This information however is important in assessment of melanoma in skin dermatoscopic images. Thus there is a need for development of image analysis techniques that satisfy at least subjective criteria defined by dermatologists. In this paper we present the application of histogram based features for detection of atypical pigment network and shape based features supplemented by artificial neural network for detection of irregular streaks. Preliminary test results are promising, for analyzed melanoma images we get 97,7% correctly detected pigmentation networks and 94,8% correctly detected irregular streaks. This paper constitutes the part of our efforts to implement the ELM 7-point checklist in order to support melanoma diagnosis and to automate this process.

The analysis of fluorescence microscopy of cells often requires the determination of cell edges. This is typically done using segmentation techniques that separate the cell objects in an image from the surrounding background. This study compares segmentation results from nine different segmentation techniques applied to two different cell lines and five different sets of imaging conditions. Significant variability in the results of segmentation was observed that was due solely to differences in imaging conditions or applications of different algorithms. We quantified and compared the results with a novel bivariate similarity index metric that evaluates the degree of underestimating or overestimating a cell object. The results show that commonly used threshold-based segmentation techniques are less accurate than k-means clustering with multiple clusters. Segmentation accuracy varies with imaging conditions that determine the sharpness of cell edges and with geometric features of a cell. Based on this observation, we propose a method that quantifies cell edge character to provide an estimate of how accurately an algorithm will perform. The results of this study will assist the development of criteria for evaluating interlaboratory comparability.

The goal of this survey paper is to overview cellular measurements using optical microscopy imaging followed by automated image segmentation. The cellular measurements of primary interest are taken from mammalian cells and their components. They are denoted as two- or three-dimensional (2D or 3D) image objects of biological interest. In our applications, such cellular measurements are important for understanding cell phenomena, such as cell counts, cell-scaffold interactions, cell colony growth rates, or cell pluripotency stability, as well as for establishing quality metrics for stem cell therapies. In this context, this survey paper is focused on automated segmentation as a software-based measurement leading to quantitative cellular measurements. We define the scope of this survey and a classification schema first. Next, all found and manually filteredpublications are classified according to the main categories: (1) objects of interests (or objects to be segmented), (2) imaging modalities, (3) digital data axes, (4) segmentation algorithms, (5) segmentation evaluations, (6) computational hardware platforms used for segmentation acceleration, and (7) object (cellular) measurements. Finally, all classified papers are converted programmatically into a set of hyperlinked web pages with occurrence and co-occurrence statistics of assigned categories. The survey paper presents to a reader: (a) the state-of-the-art overview of published papers about automated segmentation applied to optical microscopy imaging of mammalian cells, (b) a classification of segmentation aspects in the context of cell optical imaging, (c) histogram and co-occurrence summary statistics about cellular measurements, segmentations, segmented objects, segmentation evaluations, and the use of computational platforms for accelerating segmentation execution, and (d) open research problems to pursue. The novel contributions of this survey paper are: (1) a new type of classification of cellular measurements and automated segmentation, (2) statistics about the published literature, and (3) a web hyperlinked interface to classification statistics of the surveyed papers at https://isg.nist.gov/deepzoomweb/resources/survey/index.html .

is no suitable golden standard for assessment and comparison of segmentation methods applied to skin lesions images. Thus there is a need for development of image analysis techniques that satisfy at least subjective criteria defined by dermatologists. We present a model based approach for melanocytic image segmentation as a tool to improve computer aided diagnosis. During the research it was necessary to correct non-uniform image illumination caused by dermatoscope lightning. The correction algorithm based on dermatoscope light intensity estimation was used. The proposed segmentation method is based on histogram skin modeling. Preliminary test results are promising, for the analyzed melanoma images mean Jaccard index of 89.48% and mean sensitivity of 92.45% were obtained (when compared to expert assessment).