RESUMEN
Computed tomography (CT) scan provides first-hand knowledge to doctors to identify an ailment. Deep neural networks help enhance image understanding through segmentation and labeling. In this work, we implement two variants of Pix2Pix generative adversarial networks (GANs) with varying complexities of generator and discriminator networks for plane invariant segmentation of CT scan images and subsequently propose an effective generative adversarial network with a suitably weighted binary cross-entropy loss function followed by image processing layer necessary for getting high-quality output segmentation. Our conditional GAN is powered by a unique set of an encoder-decoder network that coupled with the image processing layer produces enhanced segmentation. The network can be extended to the complete set of Hounsfield units and can also be implemented on smartphones. Furthermore, we also demonstrate effects on accuracy, F-1 score, and Jaccard index by using the conditional GAN networks on the spine vertebrae dataset, thus achieving an average of 86.28 % accuracy, 90.5 % Jaccard index score, and 89.9 % F-1 score in predicting segmented maps for validation input images. In addition, an overall lifting of accuracy, F-1 score, and Jaccard index graph for validation images with better continuity has also been highlighted.
Asunto(s)
Procesamiento de Imagen Asistido por Computador , Redes Neurales de la Computación , Entropía , Procesamiento de Imagen Asistido por Computador/métodos , Tomografía Computarizada por Rayos X/métodosRESUMEN
Internet of Things (IoT) devices usage is increasing exponentially with the spread of the internet. With the increasing capacity of data on IoT devices, these devices are becoming venerable to malware attacks; therefore, malware detection becomes an important issue in IoT devices. An effective, reliable, and time-efficient mechanism is required for the identification of sophisticated malware. Researchers have proposed multiple methods for malware detection in recent years, however, accurate detection remains a challenge. We propose a deep learning-based ensemble classification method for the detection of malware in IoT devices. It uses a three steps approach; in the first step, data is preprocessed using scaling, normalization, and de-noising, whereas in the second step, features are selected and one hot encoding is applied followed by the ensemble classifier based on CNN and LSTM outputs for detection of malware. We have compared results with the state-of-the-art methods and our proposed method outperforms the existing methods on standard datasets with an average accuracy of 99.5%.
Asunto(s)
Aprendizaje Profundo , Internet de las Cosas , Humanos , Internet , InvestigadoresRESUMEN
BACKGROUND: Short birth intervals (SBIs) and long birth intervals (LBIs) have been shown to have serious implications for health of both mothers and their children. This study was aimed to investigate the determinants and reproductive outcome of SBI and LBI in a multiethnic Pakistani population. METHODS: In a cross-sectional prospective study design, 2798 women admitted in a tertiary-care hospital in Islamabad for delivery were recruited and data on second or higher birth order deliveries were collected. Birth intervals were defined as short (<24 months) and long (>36 months). The reproductive outcome was defined in terms of perinatal and neonatal mortalities, and neonatal complications. Univariate and multivariate logistic regression analyses were performed. RESULTS: Pregnancies with SBI and LBI were observed in 20% and 24% of 2798 women, respectively. Women with SBI had increased odds of perinatal death [adjusted odd ratio (AOR): 1.50] and neonatal death (AOR: 1.47) as compared to women with optimal birth intervals, while women with LBI had slightly lower odds of perinatal deaths (AOR: 0.96), but increased odds of neonatal deaths (AOR: 1.12). Further, the pregnancies with both SBI and LBI were associated with increased odds of short body length, low birth weight, small head circumference and low APGAR score. CONCLUSION: Nearly half of all pregnancies do not have optimal birth spacing albeit there is wide heterogeneity in the distribution of BI in various Pakistani ethnicities. Pregnancies with SBI and LBI had high risk of adverse reproductive outcome. Intervention programs for maternal and child health need to emphasize optimal birth spacing.
Birth interval (BI) or interpregnancy interval is the length of time between a birth and conception of the next pregnancy. Short birth intervals (SBIs) as well as long birth intervals (LBIs) have been shown to have serious implications for health of both mothers and their children. WHO recommendation for optimal spacing between 3 and 5 years. In this study, we aimed to investigate the effect of SBI and LBI on pregnancy outcome in the Pakistani population. A total of 2798 pregnant women admitted in a tertiary-care hospital in Islamabad for delivery were recruited and data on BI and pregnancy outcomes, i.e. perinatal and neonatal mortalities, and neonatal complications, were obtained. Results revealed that pregnancies with SBI and LBI were 20% and 24% of the total pregnancies, respectively. Women with SBI had higher likelihood of perinatal and neonatal death as compared to women with optimal birth intervals. Similarly, the women with LBI had higher likelihood of neonatal deaths. Furthermore, the pregnancies with both SBI and LBI were associated neonatal complications like short body length, low birth weight, small head circumference and low APGAR score. In conclusion, nearly half of all pregnancies do not have optimal birth spacing. Intervention programs for maternal and child health need to emphasize optimal birth spacing.
Asunto(s)
Intervalo entre Nacimientos , Muerte Perinatal , Niño , Estudios Transversales , Femenino , Humanos , Mortalidad Infantil , Recién Nacido , Pakistán/epidemiología , Embarazo , Resultado del Embarazo/epidemiología , Estudios ProspectivosRESUMEN
Clustering is the most common method for organizing unlabeled data into its natural groups (called clusters), based on similarity (in some sense or another) among data objects. The Partitioning Around Medoids (PAM) algorithm belongs to the partitioning-based methods of clustering widely used for objects categorization, image analysis, bioinformatics and data compression, but due to its high time complexity, the PAM algorithm cannot be used with large datasets or in any embedded or real-time application. In this work, we propose a simple and scalable parallel architecture for the PAM algorithm to reduce its running time. This architecture can easily be implemented either on a multi-core processor system to deal with big data or on a reconfigurable hardware platform, such as FPGA and MPSoCs, which makes it suitable for real-time clustering applications. Our proposed model partitions data equally among multiple processing cores. Each core executes the same sequence of tasks simultaneously on its respective data subset and shares intermediate results with other cores to produce results. Experiments show that the computational complexity of the PAM algorithm is reduced exponentially as we increase the number of cores working in parallel. It is also observed that the speedup graph of our proposed model becomes more linear with the increase in number of data points and as the clusters become more uniform. The results also demonstrate that the proposed architecture produces the same results as the actual PAM algorithm, but with reduced computational complexity.
Asunto(s)
Algoritmos , Análisis por Conglomerados , Biología Computacional/estadística & datos numéricos , Procesamiento de Imagen Asistido por Computador/estadística & datos numéricos , ComputadoresRESUMEN
Computer Vision has provided immense support to medical diagnostics over the past two decades. Analogous to Non Destructive Testing of mechanical parts, advances in medical imaging has enabled surgeons to determine root cause of an illness by consulting medical images particularly 3-D imaging. 3-D modeling in medical imaging has been pursued using surface rendering, volume rendering and regularization based methods. Tomographic reconstruction in 3D is different from camera based scene reconstruction which has been achieved using various techniques including minimal surfaces, level sets, snakes, graph cuts, silhouettes, multi-scale approach, patchwork etc. In tomography limitations of image aquisition method i-e CT Scan, X Rays and MRI as well as non availability of camera parameters for calibration restrict the quality of final reconstruction. In this work, a comprehensive study of related approaches has been carried out with a view to provide a summary of state of the art 3D modeling algorithms developed over the past four decades and also to provide a foundation study for our future work which will include precise 3D reconstruction of human spine.