Automatic segmentation of nasopharyngeal carcinoma on CT images using efficient UNet-2.5D ensemble with semi-supervised pretext task pretraining.

Nasopharyngeal carcinoma (NPC) is primarily treated with radiation therapy. Accurate delineation of target volumes and organs at risk is important. However, manual delineation is time-consuming, variable, and subjective depending on the experience of the radiation oncologist. This work explores the use of deep learning methods to automate the segmentation of NPC primary gross tumor volume (GTVp) in planning computer tomography (CT) images. A total of sixty-three (63) patients diagnosed with NPC were included in this study. Although a number of studies applied have shown the effectiveness of deep learning methods in medical imaging, their high performance has mainly been due to the wide availability of data. In contrast, the data for NPC is scarce and inaccessible. To tackle this problem, we propose two sequential approaches. First we propose a much simpler architecture which follows the UNet design but using 2D convolutional network for 3D segmentation. We find that this specific architecture is much more effective in the segmentation of GTV in NPC. We highlight its efficacy over other more popular and modern architecture by achieving significantly higher performance. Moreover to further improve performance, we trained the model using multi-scale dataset to create an ensemble of models. However, the performance of the model is ultimately dependent on the availability of labelled data. Hence building on top of this proposed architecture, we employ the use of semi-supervised learning by proposing the use of a combined pre-text tasks. Specifically we use the combination of 3D rotation and 3D relative-patch location pre-texts tasks to pretrain the feature extractor. We use an additional 50 CT images of healthy patients which have no annotation or labels. By semi-supervised pretraining the feature extractor can be frozen after pretraining which essentially makes it much more efficient in terms of the number of parameters since only the decoder is trained. Finally it is not only efficient in terms of parameters but also data, which is shown when the pretrained model with only portion of the labelled training data was able to achieve very close performance to the model trained with the full labelled data.

A3C-GS: Adaptive Moment Gradient Sharing With Locks for Asynchronous Actor-Critic Agents.

We propose an asynchronous gradient sharing mechanism for the parallel actor-critic algorithms with improved exploration characteristics. The proposed algorithm (A3C-GS) has the property of automatically diversifying worker policies in the short term for exploration, thereby reducing the need for entropy loss terms. Despite policy diversification, the algorithm converges to the optimal policy in the long term. We show in our analysis that the gradient sharing operation is a composition of two contractions. The first contraction performs gradient computation, while the second contraction is a gradient sharing operation coordinated by locks. From these two contractions, certain short- and long-term properties result. For the short term, gradient sharing induces temporary heterogeneity in policies for performing needed exploration. In the long term, under a suitably small learning rate and gradient clipping, convergence to the optimal policy is theoretically guaranteed. We verify our results with several high-dimensional experiments and compare A3C-GS against other on-policy policy-gradient algorithms. Our proposed algorithm achieved the highest weighted score. Despite lower entropy weights, it performed well in high-dimensional environments that require exploration due to sparse rewards and those that need navigation in 3-D environments for long survival tasks. It consistently performed better than the base asynchronous advantage actor-critic (A3C) algorithm.

Using deep-belief networks to understand propensity for livelihood change in a rural coastal community to further conservation.

Overharvesting of terrestrial and marine resources may be alleviated by encouraging an alternative configuration of livelihoods, particularly in rural communities in developing countries. Typical occupations in such areas include fishing and farming, and rural households often switch livelihood activities to suit climate and economic conditions. We used a machine-learning tool, deep-belief networks (DBN), and data from surveys of a rural Philippine coastal community to examine household desire to change livelihood. This desire is affected by a variety of factors, such as income, family needs, and feelings of work satisfaction, that are interrelated in complex ways. In farming households, livelihood changes often occur to diversify resources, increase income, and lessen economic risk. The DBN, given its multilayer perceptron structure, has a capacity to model nonlinear relationships among factors while providing an acceptable degree of accuracy. Relative to a set of 34 features (e.g., education, boat ownership, and work satisfaction), we examined the binary response variables desire to change work or not to change work. The best network had a test set accuracy of 97.5%. Among the features, 7 significantly affected desire to shift work: ethnicity, work satisfaction, number of persons in a household in ill health, number of fighting cocks owned, fishing engagement, buy-and-sell revenue, and educational level. A cross-correlation matrix of these 7 features indicated households less inclined to change work were those engaged in fishing and retail buying and selling. For fishing, provision of economic and other incentives should be considered to encourage changing from this occupation to allow recovery of fishery resources.

Uso de Redes de Creencia Profunda para Entender la Predisposición por el Cambio en el Medio de Vida en una Comunidad Rural Costera para Promover la Conservación Resumen La sobrexplotación de los recursos terrestres y marinos puede ser mitigada promoviendo una configuración alternativa de los medios de vida, particularmente en las comunidades rurales de los países en desarrollo. Las ocupaciones típicas en dichas áreas incluyen la pesca y la agricultura, y los hogares rurales comúnmente alternan entre estas actividades de sustento para ajustarse a las condiciones climáticas y económicas. Usamos una herramienta de aprendizaje automático, redes de creencia profunda (RCP) y datos tomados de censos aplicados a una comunidad rural costera en las Filipinas para examinar el deseo de cada hogar por cambiar su medio de vida. Este deseo se ve afectado por una variedad de factores, como el nivel de ingreso, las necesidades familiares y la sensación de satisfacción laboral, que están interrelacionados de maneras complejas. En los hogares agricultores, los cambios en el medio de vida ocurren con frecuencia para diversificar los recursos, aumentar el ingreso y disminuir el riesgo económico. Las RCP, dada la estructura multicapa de su perceptrón, tiene una capacidad para modelar las relaciones no lineales entre los factores mientras proporciona un grado aceptable de precisión. En relación con un conjunto de 34 caracteres (p. Ej.: educación, propiedad de navíos, satisfacción laboral), examinamos las variables de respuesta binaria deseo cambiar de trabajo y no deseo cambiar de trabajo. La mejor red tuvo una precisión de conjunto de pruebas del 97.5%. Entre los caracteres, siete afectaron significativamente el deseo por cambiar de trabajo: la etnicidad, la satisfacción laboral, el número de personas con mala salud en un mismo hogar, el número de gallos de pelea en su propiedad, la participación dentro de la pesca, los ingresos por compra y venta, y el nivel educativo. Una matriz de correlación cruzada de estos siete caracteres indicó que los hogares con una menor tendencia hacia el cambio de trabajo fueron aquellos que participan en la pesca y en la compra y venta al menudeo. Para la pesca, el suministro de incentivos económicos y de otros tipos debería considerarse para promover el cambio de esta ocupación y así permitir la recuperación de los recursos pesqueros.

ESP: an expert system for poisoning diagnosis and management.

We describe a clinical decision support system (CDSS) designed to provide timely information germane to poisoning. The CDSS aids medical decision making through recommendations to clinicians for immediate evaluation. The system is implemented as a rule-based expert system with two major components: the knowledge base and the inference engine. The knowledge base serves as the database which contains relevant poisoning information and rules that are used by the inference engine in making decisions. This expert system accepts signs and symptoms observed from a patient as input and presents a list of possible poisoning types with the corresponding management procedures which may be considered in making the final diagnosis. A knowledge acquisition tool (KAT) that allows toxicological experts to update the knowledge base was also developed. This article describes the architecture of the fully featured system, the design of the CDSS and the KAT as web applications, the utilisation of the inferencing mechanism of C Language Integrated Production System (CLIPS), which is an expert system shell that helps the system in decision-making tasks, the methods used as well as problems encountered. We also present the results obtained after testing the system and propose some recommendations for future work.

Parameter estimation using Simulated Annealing for S-system models of biochemical networks.

MOTIVATION: High-throughput technologies now allow the acquisition of biological data, such as comprehensive biochemical time-courses at unprecedented rates. These temporal profiles carry topological and kinetic information regarding the biochemical network from which they were drawn. Retrieving this information will require systematic application of both experimental and computational methods. RESULTS: S-systems are non-linear mathematical approximative models based on the power-law formalism. They provide a general framework for the simulation of integrated biological systems exhibiting complex dynamics, such as genetic circuits, signal transduction and metabolic networks. We describe how the heuristic optimization technique simulated annealing (SA) can be effectively used for estimating the parameters of S-systems from time-course biochemical data. We demonstrate our methods using three artificial networks designed to simulate different network topologies and behavior. We then end with an application to a real biochemical network by creating a working model for the cadBA system in Escherichia coli. AVAILABILITY: The source code written in C++ is available at http://www.engg.upd.edu.ph/~naval/bioinformcode.html. All the necessary programs including the required compiler are described in a document archived with the source code. SUPPLEMENTARY INFORMATION: Supplementary material is available at Bioinformatics online.