Humberto de Jesús Ochoa Domínguez

Automated system for lung nodules classification based on wavelet feature descriptor and support vector machine

BackgroundLung cancer is a leading cause of death worldwide; it refers to the uncontrolled growth of abnormal cells in the lung. A computed tomography (CT) scan of the thorax is the most sensitive method for detecting cancerous lung nodules. A lung nodule is a round lesion which can be either non-cancerous or cancerous. In the CT, the lung cancer is observed as round white shadow nodules. The possibility to obtain a manually accurate interpretation from CT scans demands a big effort by the radiologist and might be a fatiguing process. Therefore, the design of a computer-aided diagnosis (CADx) system would be helpful as a second opinion tool.MethodsThe stages of the proposed CADx are: a supervised extraction of the region of interest to eliminate the shape differences among CT images. The Daubechies db1, db2, and db4 wavelet transforms are computed with one and two levels of decomposition. After that, 19 features are computed from each wavelet sub-band. Then, the sub-band and attribute selection is performed. As a result, 11 features are selected and combined in pairs as inputs to the support vector machine (SVM), which is used to distinguish CT images containing cancerous nodules from those not containing nodules.ResultsThe clinical data set used for experiments consists of 45 CT scans from ELCAP and LIDC. For the training stage 61 CT images were used (36 with cancerous lung nodules and 25 without lung nodules). The system performance was tested with 45 CT scans (23 CT scans with lung nodules and 22 without nodules), different from that used for training. The results obtained show that the methodology successfully classifies cancerous nodules with a diameter from 2 mm to 30 mm. The total preciseness obtained was 82%; the sensitivity was 90.90%, whereas the specificity was 73.91%.ConclusionsThe CADx system presented is competitive with other literature systems in terms of sensitivity. The system reduces the complexity of classification by not performing the typical segmentation stage of most CADx systems. Additionally, the novelty of the algorithm is the use of a wavelet feature descriptor.

Smart Multi-Level Tool for Remote Patient Monitoring Based on a Wireless Sensor Network and Mobile Augmented Reality

Technological innovations in the field of disease prevention and maintenance of patient health have enabled the evolution of fields such as monitoring systems. One of the main advances is the development of real-time monitors that use intelligent and wireless communication technology. In this paper, a system is presented for the remote monitoring of the body temperature and heart rate of a patient by means of a wireless sensor network (WSN) and mobile augmented reality (MAR). The combination of a WSN and MAR provides a novel alternative to remotely measure body temperature and heart rate in real time during patient care. The system is composed of (1) hardware such as Arduino microcontrollers (in the patient nodes), personal computers (for the nurse server), smartphones (for the mobile nurse monitor and the virtual patient file) and sensors (to measure body temperature and heart rate), (2) a network layer using WiFly technology, and (3) software such as LabView, Android SDK, and DroidAR. The results obtained from tests show that the system can perform effectively within a range of 20 m and requires ten minutes to stabilize the temperature sensor to detect hyperthermia, hypothermia or normal body temperature conditions. Additionally, the heart rate sensor can detect conditions of tachycardia and bradycardia.

Lung nodule classification in frequency domain using support vector machines

In this paper a computational alternative to classify lung nodules inside CT thorax images in the frequency domain is presented. After image acquisition, a region of interest is manually selected. Then, the spectrums of the two dimensional Discrete Cosine Transform (2D-DCT) and the two dimensional Fast Fourier Transform (2D-FFT) were calculated. Later, two statistical texture features were extracted from the histogram computed from the spectrum of each CT image. Finally, a support vector machine with a radial basis function as a kernel was used as the classifier. Seventy five tests with different diagnosis and number of images were used to validate the methodology presented. After experimentation and results, ten false negatives (FN) and two false positives (FP) were obtained, and a sensitivity and specificity of 96.15% and 52.17% respectively. The total preciseness obtained with the methodology proposed was 82.66%.

Lung Nodule Classification in CT Thorax Images Using Support Vector Machines

In this paper a computational alternative to classify lung nodules using computed tomography (CT) thorax images is presented. The novelty of the method is the elimination of the segmentation stage. The contribution consist of several steps. After image acquisition, eight texture features were extracted from the histogram and the gray level coocurrence matrix (with four different angles) for each CT image. The features were used to train a non-parametric classifier called support vector machine (SVM), used to classify lung tissues into two classes: with lung nodules and without lung nodules. A total of 128 public clinical data set (ELCAP, NBIA) with different number of slices and diagnoses were used to train and evaluate the performance of the methodology presented. After the tests stage, five false negative (FN) and seven false positive (FP) results were obtained. The results obtained were validated by a radiologist to finally obtain a reliability index of 84%.

Noise reduction in small-animal PET images using a multiresolution transform.

In this paper, we address the problem of denoising reconstructed small animal positron emission tomography (PET) images, based on a multiresolution approach which can be implemented with any transform such as contourlet, shearlet, curvelet, and wavelet. The PET images are analyzed and processed in the transform domain by modeling each subband as a set of different regions separated by boundaries. Homogeneous and heterogeneous regions are considered. Each region is independently processed using different filters: a linear estimator for homogeneous regions and a surface polynomial estimator for the heterogeneous region. The boundaries between the different regions are estimated using a modified edge focusing filter. The proposed approach was validated by a series of experiments. Our method achieved an overall reduction of up to 26% in the %STD of the reconstructed image of a small animal NEMA phantom. Additionally, a test on a simulated lesion showed that our method yields better contrast preservation than other state-of-the art techniques used for noise reduction. Thus, the proposed method provides a significant reduction of noise while at the same time preserving contrast and important structures such as lesions.

The Nonsubsampled Contourlet Transform for Enhancement of Microcalcifications in Digital Mammograms

Microcalcifications detection plays a crucial role in the early detection of breast cancer. The enhancement of the mammographic images is one of the most important tasks during the detection process. This paper presents an algorithm for the enhancement of microcalcifications in digital mammograms. The main novelty is the application of the nonsubsampled contourlet transform and a specific edge filter to enhance the directional structures of the image in the contourlet domain. The inverse contourlet transform is applied to recover an approximation of the mammogram with the microcalcifications enhanced. Results show that the proposed method outperforms the current method based on the discrete wavelet transform.

The H.264 Video Coding Standard

Since 1990, the Video Coding Expert Group (VCEG) and the Moving Pictures Expert Group (MPEG) have focused their research in video coding techniques for different applications. The International Telecommunication Union (ITU-T) and the MPEG groups joined in a single group to form the Joint Video Team (JVT) and worked together on a new video standard called H.264, H.264/AVC Advanced Video Coding, or MPEG-4 part 10, which achieves more coding efficiency with simpler syntax specifications than the previous standards, more integration with all network protocols, and multiplex architectures. The H.264 is the state-ofthe-art codec and covers a wide range of applications with excellent results such as videoconferencing, video streaming, and video transmission over fixed and wireless networks with different transport protocols among others.

Visual perception substitution by the auditory sense

This paper presents the methodology to develop a sensory substitution system. We deal with the absence of visual inputs in humans, and we substitute the blindness by the auditory sense. Visual to auditory substitution involves delivering information about the visual world using auditory signals. The system allows the transformation of digital images captured from a web cam into sound patterns, using a novel scanning method from the center to the left and right side of the image. We define a robust correspondence between the image features and the sound features. The results provided by the system seems high enough to address many practical situations that normally require the sense of sight. Navigation experiments with blind people, have demonstrated the ability of the system to offer a permanent sensory substitution device in the future.

Biometric Human Identification of Hand Geometry Features Using Discrete Wavelet Transform

Since the security factor became a basic need for civilization, a lot of systems have been developed. Those systems, try to ensure the safety in all the things that driving a certain degree of exclusivity. Historically, keys, cards and passwords were used as security systems; however, these methods are vulnerable to loss and theft. As a result biometric identification methods emerge in order to tackle the disadvantages of the non biometric classical methods. Biometrics, is an emerging technology that addresses the automated identification of individuals, based on their physiological and behavioral traits. The main advantage of biometric methods is the ability to recognize, which is made by means of a physical feature or a unique pattern (Jain et al. (2008)). With these methods and individual can hardly be victim of plagiarism. There exist several biometrics cues such as iris (Abhyankara & Schuckersa (2010)), face (Abatea et al. (2007)), fingerprint (Jimenez et al. (2010)), voice (Andicsa et al. (2010)), but one of the cheapest is the hand geometry. Hand geometry, as the name suggests, refers to the geometric structure of the hand (Singh et al. (2009)). Hand geometry measurement is non intrusive and the verification involves a simple processing of the resulting features. Usually the hand geometry identification involves a digital picture acquisition and translation of the nodal points like: space between fingers, curvature, length and width of the hand into numerical representations used to cross reference with other hand prints stored in a database for a match. The schemes which uses geometrical features of the hand, focused on characteristics as widths of fingers at articulations, finger and palm lengths, finger deviations and the angles of the inter-finger valleys with respect to the horizontal. The number of features obtained varied in the range of 20-30, and usually the acquisition stage need pegs to define the accurate finger position (Yoruk et al. (2006)). The ability of associating an identity with an individual is called identification. Hand geometry measurements are easily collectible due to both the dexterity of the hand and due 13

A Novel Evolutionary Face Recognition Algorithm Using Particle Swarm Optimization

In this paper a novel algorithm to solve the problem of automatic face recognition is presented. The novelty of the algorithm is the ability to combine the computer vision tasks with Particle Swarm Optimization (PSO) to improve the execution time and to obtain better recognition results. The crucial stage of a typical system of face recognition is improved by using a fitness function to measure the similarity of an input face compared with a database of faces. The use of the fitness function helps to obtain more accurate results in a faster way. The results obtained are excellent even when the system was tested in uncontrolled environments. A comparison of the results obtained with the algorithm without PSO versus the algorithm using PSO is also presented.