Search

Automated image analysis in cellular microscopy has acquired a great importance, as a consequence of the large amount of visual information that is currently generated in numerous applications. Computer programs and algorithms to perform these tasks are in constant evolution, and evaluating both their performance and the validity of the results obtained through them, has a primary importance. In the specific case of image segmentation, evaluating the results by means of comparison to reference (ground-truth) images, demand the availability of images in which the segmentation outcome could be considered perfect. This requisite is very hard to accomplish with real images using manual segmentation, due to the limitations inherent to human analysts. As a consequence, in order to evaluate segmentation algorithms in cellular microscopy applications, simulation of realistic images that can serve as an adequate ground-truth has acquired a significant importance. In this work, an algorithm is described and applied to simulate synthetic images in cellular microscopy, specifically those composed by erythrocytes in the field of haematology. In order to generate the erythrocytes, a basic algorithm based in ellipses was developed, that can experience various types of random variations. These variations allow adding realistic details to them, like noise and blur. As a result, simulated images were obtained, which can be used to evaluate the effectiveness of segmentation algorithms in this application, thanks to the fact that at the same time binary ideal ground-truth images are generated. Finally, some examples in evaluating segmentation algorithms using synthetic images are presented, to illustrate their application.

Segmentation of microscopy images of blood smears is a source of some challenging digital image processing (DIP) tasks, among them that of segmenting erythrocyte aggregates (cells that are touching or overlapping). When addressing this problem, it is mandatory to make a good selection of an appropriate segmentation technique, capable of separating the aggregates in their constituent cells, despite the variability in morphological characteristics that they can exhibit. In this paper two alternatives are proposed and evaluated to segment erythrocyte aggregates, which combine the watershed and the distance transforms with minima imposition in the internal markers. The segmentation quality was evaluated by means of the Dice coefficient, calculated using manually segmented images as the groundtruth. The results obtained are illustrated through examples of the resulting images and shown quantitatively in tabulated form for the values of the Dice coefficient obtained.

The Papanicolaou test is used for early prediction of cervical cancer. Computer vision techniques for automating the microscopy analysis of cervical cells in this test have received great attention. Cell segmentation is needed here in order to obtain appropriate features for classification of abnormal cells. However, accurate segmentation of the cell cytoplasm is difficult, due to cell overlapping and variability of color and intensity. This has determined a growing interest in classifying cells using only features from the nuclei, which are easier to segment. In this work, we classified cells in the pap-smear test using a combination of morphometric and Haralick texture features, obtained from the nucleus gray-level co-occurrence matrix. A comparison was made among various classifiers using these features and data dimensionality reduction through PCA. The results obtained showed that this combination can be a promising alternative in order to automate the analysis of cervical cells.

Fluorescent signals are widely used in life sciences to investigate cellular behaviors. Using a fluorometer is a reliable approach to measure the amount of a specific fluorescent radiation emitted by a cellular population. On the other hand, measuring fluorescent signals at the single-cell level requires the use of properly equipped microscopes to acquire images for further analysis. Here we present a digital image processing (DIP) approach to analyze these images. More in detail, a DIP system was developed to analyze photographic images of fluorescent Escherichia coli expressing a Green Fluorescent Protein (GFP). This system includes stages of noise filtering, uneven illumination correction, segmentation, detection of cell aggregates and measurement of the level of fluorescence of the selected cells. To accomplish this task, well established DIP algorithms for the different functions were adapted to the specific conditions of the experiment under study. In the results section, images are presented that reveal the behavior of the algorithms employed, as well as the calculated distribution of single cell fluorescence. In regard of the adequate selection of valid cells (i.e. deletion of aggregates), the system effectiveness was evaluated through comparison to the results obtained by human analysis, and expressed quantitatively. An assessment on the perspectives of further improvements to the developed system is finally proposed.

A study on image quality was developed using 15 tomography images collected over a head anthropomorphic phantom. The physical region of the phantom was used to analyze contrast resolution. The technology used for tomographic acquisitions was a Siemens Sensation 64. Tube current (mA), kVp, tomographic slice thickness and acquisition time per view were the scan variables changed among tomographies, looking for the best scan conditions which permit correct contrast resolution with the lowest dose. Kerma-air in air was also measured and its index Ca,100 (mGy) was calculated as dose indicator. Contrast resolution and detection limits of the CT machine were studied with the Rose Model application. Objective image quality metrics as contrast to noise ratio (CNR), image contrast and global signal to noise ratio (SNR) were calculated over the tomographic images and a visual subjective expert analysis were done to compare with the model results.

Complex images from different processes are often acquired with a low signal to noise ratio, as it is the case with Magnetic Resonance Imaging. Noise filtering is used to recover the associated phase images, mitigating negative effects such as loss of contrast and the introduction of phase residues, which constitute a major drawback for phase unwrapping processes. In this work, a group of algorithms combining nonlinear filters and wavelet de-noising were developed and applied to MRI images, in order to recover the phase information. The results obtained with the two algorithms that exhibited the best performance when applied to both phantom and real images, are shown. Application of these algorithms resulted in improvements both in terms of SNR and of the decrement in the number of phase residues.

Identifying protein coding regions in DNA sequences is a basic step in the location of genes. Several approaches based on signal processing tools have been applied to solve this problem, trying to achieve more accurate predictions. This paper presents a new predictor that improves the efficacy of three ones that use the Fourier Transform to predict coding regions, and that could be computed using an algorithm that reduces the computation load. ROC curves are used to demonstrate the efficacy of the proposed predictor, based on the computation of 25 DNA sequences from three different organisms.

The Hilbert Transform (HT) and the analytic signal (AS) are widely used in their one-dimensional version for various applications. However, in the bi-dimensional (2D) case as occur for images, the definition of the 2D-HT is not unique and several approaches to it have been developed, having as one of the main goals to obtain a meaningful 2D-AS or analytic image, which can be used for various practical applications. In this work, one particular approach to the 2D-HT is introduced that allowed the calculation of analytic images which satisfy the basic properties that these functions have in the 1D case, and that produces a 2D spectrum equal to zero in one quadrant. The methods for calculation of the discrete version of the 2D-HT and the associated AS are presented and analyzed, as well as two applications, for edge detection and for envelope detection in a 2D AM modulated radial chirp.

Complex images with low signal to noise ratio (SNR) appear in various applications. The recovery of the associated phase images needs to consider noise influence, including loss of contrast and phase residues that can deteriorate the phase unwrapping process. Noise reduction in complex images has been addressed by various methods; however most of them deal only with the magnitude image. Only few works have been devoted to phase image denoising, despite the existence of important applications like Current Density Imaging (CDI) and Magnetic Resonance Imaging (MRI) that require their use. In this work, a group of denoising algorithms in the wavelet domain were developed and applied to the complex image in Magnetic Resonance Imaging (MRI), in order to recover the phase information. These filtering algorithms were applied to tomographic and phantom images. Significant decrement in the number of remaining phase residues were achieved by using the proposed filters.

Time-frequency representations (TFR) convey relevant information about systems that can not be obtained under stationary conditions. In this paper, a methodology to classify systems using the information obtained from time-frequency representations during transient phenomena is described and tested experimentally. The study includes an assessment of the features to be extracted from the TFR, which are relevant for the desired classification, as well as the construction of the appropriate discriminant functions using them. The methodology is tested by means of a biomedical example related to patient's classification.

Complex images contaminated by noise appear in various applications. To improve these phase images, noise effects, as loss of contrast and phase residues that deteriorate the phase unwrapping process, should be reduced. Noise reduction in complex images has been addressed by various methods, most of them dealing only with the magnitude image. Few works have been devoted to phase image de-noising, despite the existence of important applications like Interferometric Synthetic Aperture Radar (IFSAR), Current Density Imaging (CDI) and Magnetic Resonance Imaging (MRI). In this work, several de-noising algorithms in the wavelet packets domain were applied to complex images to recover the phase information. These filtering algorithms were applied to simulated images contaminated by three different noise models, including mixtures of Gaussian and Impulsive noise. Significant improvements in SNR for low initial values (SNR

Ventricular late potential detection can be used as a non-invasive diagnostic tool, but traditional detection techniques need around 300 heartbeats and fail to obtain the beat-to-beat information. This paper combines a modified signal averaging and an adaptive enhancer to deal with non-stationary environments and get beat-to-beat information from as little as 60 beats. In the ventricular late potential region of the recovered signal, discernible patterns indicate the presence or not of such waveforms. A maximum absolute value “averaging” can emphasize the boundaries of the QRS complex even further to successfully detect ventricular late potentials.