Semi-online neural-Q_leaming for real-time robot learning

Reinforcement learning (RL) is a very suitable technique for robot learning, as it can learn in unknown environments and in real-time computation. The main difficulties in adapting classic RL algorithms to robotic systems are the generalization problem and the correct observation of the Markovian state. This paper attempts to solve the generalization problem by proposing the semi-online neural-Q_learning algorithm (SONQL). The algorithm uses the classic Q_learning technique with two modifications. First, a neural network (NN) approximates the Q_function allowing the use of continuous states and actions. Second, a database of the most representative learning samples accelerates and stabilizes the convergence. The term semi-online is referred to the fact that the algorithm uses the current but also past learning samples. However, the algorithm is able to learn in real-time while the robot is interacting with the environment. The paper shows simulated results with the "mountain-car" benchmark and, also, real results with an underwater robot in a target following behavior

Policy gradient based Reinforcement Learning for real autonomous underwater cable tracking

This paper proposes a field application of a high-level reinforcement learning (RL) control system for solving the action selection problem of an autonomous robot in cable tracking task. The learning system is characterized by using a direct policy search method for learning the internal state/action mapping. Policy only algorithms may suffer from long convergence times when dealing with real robotics. In order to speed up the process, the learning phase has been carried out in a simulated environment and, in a second step, the policy has been transferred and tested successfully on a real robot. Future steps plan to continue the learning process on-line while on the real robot while performing the mentioned task. We demonstrate its feasibility with real experiments on the underwater robot ICTINEU AUV

Towards a real-time vision-based navigation system for a small-class UUV

This paper deals with the problem of navigation for an unmanned underwater vehicle (UUV) through image mosaicking. It represents a first step towards a real-time vision-based navigation system for a small-class low-cost UUV. We propose a navigation system composed by: (i) an image mosaicking module which provides velocity estimates; and (ii) an extended Kalman filter based on the hydrodynamic equation of motion, previously identified for this particular UUV. The obtained system is able to estimate the position and velocity of the robot. Moreover, it is able to deal with visual occlusions that usually appear when the sea bottom does not have enough visual features to solve the correspondence problem in a certain area of the trajectory

A new approach to real time traffic sign recognition based on colour information

The automatic interpretation of conventional traffic signs is very complex and time consuming. The paper concerns an automatic warning system for driving assistance. It does not interpret the standard traffic signs on the roadside; the proposal is to incorporate into the existing signs another type of traffic sign whose information will be more easily interpreted by a processor. The type of information to be added is profuse and therefore the most important object is the robustness of the system. The basic proposal of this new philosophy is that the co-pilot system for automatic warning and driving assistance can interpret with greater ease the information contained in the new sign, whilst the human driver only has to interpret the "classic" sign. One of the codings that has been tested with good results and which seems to us easy to implement is that which has a rectangular shape and 4 vertical bars of different colours. The size of these signs is equivalent to the size of the conventional signs (approximately 0.4 m2). The colour information from the sign can be easily interpreted by the proposed processor and the interpretation is much easier and quicker than the information shown by the pictographs of the classic signs

The Estrada Real project and endemic diseases: the case of schistosomiasis, geoprocessing and tourism

Geographical Information System (GIS) is a tool that has recently been applied to better understand spatial disease distributions. Using meteorological, social, sanitation, mollusc distribution data and remote sensing variables, this study aimed to further develop the GIS technology by creating a model for the spatial distribution of schistosomiasis and to apply this model to an area with rural tourism in the Brazilian state of Minas Gerais (MG). The Estrada Real, covering about 1,400 km, is the largest and most important Brazilian tourism project, involving 163 cities in MG with different schistosomiasis prevalence rates. The model with three variables showed a R² = 0.34, with a standard deviation of risk estimated adequate for public health needs. The main variables selected for modelling were summer vegetation, summer minimal temperature and winter minimal temperature. The results confirmed the importance of Remote Sensing data and the valuable contribution of GIS in identifying priority areas for intervention in tourism regions which are endemic to schistosomiasis.

The prevalence of schistosomiasis in school-aged children as an appropriate indicator of its prevalence in the community

School-aged children (6-15 years) from the endemic area of Pernambuco were evaluated both as a target group for and an indicator of schistosomiasis control in the community. Parasitological data were drawn from baseline stool surveys of whole populations that were obtained to diagnose Schistosoma mansoni infection. Nineteen representative localities were selected for assessing the prevalence of schistosomiasis among individuals in the following age groups: 0-5, 6-15, 16-25, 26-40 and 41-80 years. For each locality, the prevalence in each age group was compared to that of the overall population using contingency table analysis. To select a reference group, the operational difficulties of conducting residential surveys were considered. School-aged children may be considered to be the group of choice as the reference group for the overall population for the following reasons: (i) the prevalence of schistosomiasis in this age group had the highest correlation with the prevalence in the overall population (r = 0.967), (ii) this age group is particularly vulnerable to infection and plays an important role in parasite transmission and (iii) school-aged children are the main target of the World Health Organization in terms of helminth control. The Schistosomiasis Control Program should consider school-aged children both as a reference group for assessing the need for intervention at the community level and as a target group for integrated health care actions of the Unified Health System that are focused on high-risk groups.

Increased expression of urokinase-type plasminogen activator mRNA determines adverse prognosis in ErbB2-positive primary breast cancer.

PURPOSE: To evaluate and validate mRNA expression markers capable of identifying patients with ErbB2-positive breast cancer associated with distant metastasis and reduced survival. PATIENTS AND METHODS: Expression of 60 genes involved in breast cancer biology was assessed by quantitative real-time PCR (qrt-PCR) in 317 primary breast cancer patients and correlated with clinical outcome data. Results were validated subsequently using two previously published and publicly available microarray data sets with different patient populations comprising 295 and 286 breast cancer samples, respectively. RESULTS: Of the 60 genes measured by qrt-PCR, urokinase-type plasminogen activator (uPA or PLAU) mRNA expression was the most significant marker associated with distant metastasis-free survival (MFS) by univariate Cox analysis in patients with ErbB2-positive tumors and an independent factor in multivariate analysis. Subsequent validation in two microarray data sets confirmed the prognostic value of uPA in ErbB2-positive tumors by both univariate and multivariate analysis. uPA mRNA expression was not significantly associated with MFS in ErbB2-negative tumors. Kaplan-Meier analysis showed in all three study populations that patients with ErbB2-positive/uPA-positive tumors exhibited significantly reduced MFS (hazard ratios [HR], 4.3; 95% CI, 1.6 to 11.8; HR, 2.7; 95% CI, 1.2 to 6.2; and, HR, 2.8; 95% CI, 1.1 to 7.1; all P < .02) as compared with the group with ErbB2-positive/uPA-negative tumors who exhibited similar outcome to those with ErbB2-negative tumors, irrespective of uPA status. CONCLUSION: After evaluation of 898 breast cancer patients, uPA mRNA expression emerged as a powerful prognostic indicator in ErbB2-positive tumors. These results were consistent among three independent study populations assayed by different techniques, including qrt-PCR and two microarray platforms.

Real-time PCR-based quantification of Toxoplasma gondii in tissue samples of serologically positive outdoor chickens

This study aimed to quantify Toxoplasma gondii in tissue samples of serologically positive chickens using real-time polymerase chain reaction (PCR). Of 65 chickens evaluated, 28 were positive for T. gondii antibodies. Brain and heart samples were collected from 26 seropositive chickens and DNA was extracted using Trizol® and amplified using real-time PCR with SYBR® Green. Parasite DNA was detected in 24 of the 26 samples analyzed; the number of positive tissue samples and the parasite quantity did not differ between tissue types. The results confirmed the analytical sensitivity of parasite detection in chicken tissue samples and demonstrated the possibility of using other molecular systems for genotypic analysis.

Using 3D surface datasets to understand landslide evolution: From analogue models to real case study

Early detection of landslide surface deformation with 3D remote sensing techniques, as TLS, has become a great challenge during last decade. To improve our understanding of landslide deformation, a series of analogue simulation have been carried out on non-rigid bodies coupled with 3D digitizer. All these experiments have been carried out under controlled conditions, as water level and slope angle inclination. We were able to follow 3D surface deformation suffered by complex landslide bodies from precursory deformation still larger failures. These experiments were the basis for the development of a new algorithm for the quantification of surface deformation using automatic tracking method on discrete points of the slope surface. To validate the algorithm, comparisons were made between manually obtained results and algorithm surface displacement results. Outputs will help in understanding 3D deformation during pre-failure stages and failure mechanisms, which are fundamental aspects for future implementation of 3D remote sensing techniques in early warning systems.

Peripheral eosinophilia as an indicator of meningitic angiostrongyliasis in exposed individuals

The diagnosis of meningitic angiostrongyliasis (MA) is based on clinical criteria. A lumbar puncture is used as a diagnostic tool, but it is an invasive procedure. The blood eosinophil levels are also assessed and used in the diagnosis of this disease. We enrolled 47 patients with serologically proven MA and 131 controls with intestinal parasite infections. An absolute eosinophil count model was found to be the best marker for MA. An eosinophil count of more than 798 cells led to sensitivity, specificity, positive predictive and negative predictive values of 76.6%, 80.2%, 58.1% and 90.5%, respectively. These data support the use of testing for high blood eosinophil levels as a diagnostic tool for MA in individuals that are at risk for this disease.

Detection of human bocavirus and human metapneumovirus by real-time PCR from patients with respiratory symptoms in Southern Brazil

The introduction of newer molecular methods has led to the discovery of new respiratory viruses, such as human metapneumovirus (hMPV) and human bocavirus (hBoV), in respiratory tract specimens. We have studied the occurrence of hMPV and hBoV in the Porto Alegre (PA) metropolitan area, one of the southernmost cities of Brazil, evaluating children with suspected lower respiratory tract infection from May 2007-June 2008. A real-time polymerase chain reaction method was used for amplification and detection of hMPV and hBoV and to evaluate coinfections with respiratory syncytial virus (RSV), influenza A and B, parainfluenza 1, 2 and 3, human rhinovirus and human adenovirus. Of the 455 nasopharyngeal aspirates tested, hMPV was detected in 14.5% of samples and hBoV in 13.2%. A unique causative viral agent was identified in 46.2% samples and the coinfection rate was 43.7%. For hBoV, 98.3% of all positive samples were from patients with mixed infections. Similarly, 84.8% of all hMPV-positive results were also observed in mixed infections. Both hBoV and hMPV usually appeared with RSV. In summary, this is the first confirmation that hMPV and hBoV circulate in PA; this provides evidence of frequent involvement of both viruses in children with clinical signs of acute viral respiratory tract infection, although they mainly appeared as coinfection agents.

Efficacy, safety and pharmacokinetic of once-daily boosted saquinavir (1500/100 mg) together with 2 nucleos(t)ide reverse transcriptase inhibitors in real life: a multicentre prospective study.

BACKGROUND. Ritonavir-boosted saquinavir (SQVr) is nowadays regarded as an alternative antiretroviral drug probably due to several drawbacks, such as its high pill burden, twice daily dosing and the requirement of 200 mg ritonavir when given at the current standard 1000/100 mg bid dosing. Several once-daily SQVr dosing schemes have been studied with the 200 mg SQV old formulations, trying to overcome some of these disadvantages. SQV 500 mg strength tablets became available at the end of 2005, thus facilitating a once-daily regimen with fewer pills, although there is very limited experience with this formulation yet. METHODS. Prospective, multicentre study in which efficacy, safety and pharmacokinetics of a regimen of once-daily SQVr 1500/100 mg plus 2 NRTIs were evaluated under routine clinical care conditions in either antiretroviral-naïve patients or in those with no previous history of antiretroviral treatments and/or genotypic resistance tests suggesting SQV resistance. Plasma SQV trough levels were measured by HPLV-UV. RESULTS. Five hundred and fourteen caucasian patients were included (47.2% coinfected with hepatitis C and/or B virus; 7.8% with cirrhosis). Efficacy at 52 weeks (plasma RNA-HIV <50 copies/ml) was 67.7% (CI95: 63.6 - 71.7%) by intention-to-treat, and 92.2% (CI95: 89.8 - 94.6%) by on-treatment analysis. The reasons for failure were: dropout or loss to follow-up (18.4%), virological failure (7.8%), adverse events (3.1%), and other reasons (4.6%). The high rate of dropout may be explained by an enrollment and follow-up under routine clinical care condition, and a population with a significant number of drug users. The median SQV Cmin (n = 49) was 295 ng/ml (range, 53-2172). The only variable associated with virological failure in the multivariate analysis was adherence (OR: 3.36; CI95, 1.51-7.46, p = 0.003). CONCLUSIONS. Our results suggests that SQVr (1500/100 mg) once-daily plus 2 NRTIs is an effective regimen, without severe clinical adverse events or hepatotoxicity, scarce lipid changes, and no interactions with methadone. All these factors and its once-daily administration suggest this regimen as an appropriate option in patients with no SQV resistance-associated mutations.

Real-time molecular epidemiology of tuberculosis by direct genotyping of smear-positive clinical specimens.

We applied MIRU-VNTR (mycobacterial interspersed repetitive-unit-variable-number tandem-repeat typing) to directly analyze the bacilli present in 61 stain-positive specimens from tuberculosis patients. A complete MIRU type (24 loci) was obtained for all but one (no amplification in one locus) of the specimens (98.4%), and the allelic values fully correlated with those obtained from the corresponding cultures. Our study is the first to demonstrate that real-time genotyping of Mycobacterium tuberculosis can be achieved, fully transforming the way in which molecular epidemiology techniques can be integrated into control programs.

Malaria diagnosis from pooled blood samples: comparative analysis of real-time PCR, nested PCR and immunoassay as a platform for the molecular and serological diagnosis of malaria on a large-scale

Malaria diagnoses has traditionally been made using thick blood smears, but more sensitive and faster techniques are required to process large numbers of samples in clinical and epidemiological studies and in blood donor screening. Here, we evaluated molecular and serological tools to build a screening platform for pooled samples aimed at reducing both the time and the cost of these diagnoses. Positive and negative samples were analysed in individual and pooled experiments using real-time polymerase chain reaction (PCR), nested PCR and an immunochromatographic test. For the individual tests, 46/49 samples were positive by real-time PCR, 46/49 were positive by nested PCR and 32/46 were positive by immunochromatographic test. For the assays performed using pooled samples, 13/15 samples were positive by real-time PCR and nested PCR and 11/15 were positive by immunochromatographic test. These molecular methods demonstrated sensitivity and specificity for both the individual and pooled samples. Due to the advantages of the real-time PCR, such as the fast processing and the closed system, this method should be indicated as the first choice for use in large-scale diagnosis and the nested PCR should be used for species differentiation. However, additional field isolates should be tested to confirm the results achieved using cultured parasites and the serological test should only be adopted as a complementary method for malaria diagnosis.

Molecular detection of Schistosoma japonicum in infected snails and mouse faeces using a real-time PCR assay with FRET hybridisation probes

A real-time polymerase chain reaction (PCR) assay with fluorescence resonance energy transfer (FRET) hybridisation probes combined with melting curve analysis was developed to detect Schistosoma japonicum in experimentally infected snails and in faecal samples of infected mice. This procedure is based on melting curve analysis of a hybrid between an amplicon from the S. japonicum internal transcribed spacer region 2 sequence, which is a 192-bp S. japonicum-specific sequence, and fluorophore-labelled specific probes. Real-time FRET PCR could detect as little as a single cercaria artificially introduced into a pool of 10 non-infected snails and a single egg inoculated in 100 mg of non-infected mouse faeces. All S. japonicum-infected snails and all faecal samples from infected mice were positive. Non-infected snails, non-infected mouse faeces and genomic DNA from other parasites were negative. This assay is rapid and has potential for epidemiological S. japonicum surveys in snails, intermediate hosts and faecal samples of final hosts.