Discriminative Set Research Articles

P1.11-11 Initial Discovery of Exhaled Small Polar Energetics-Related Metabolites by GC-MS for Lung Cancer Risk Assessment

There is a need for non-invasive airway-based biomarkers in lung carcinogenesis for both risk assessment, and earlier diagnosis. Exhaled breath condensate (EBC) contains airway lining fluid molecules, including small molecules of polar and non-polar lipid origin, presumably in part from epithelial cellular origins. Here we pilot a GC-MS strategy for measurement of small, polar molecules in exhaled breath condensate in EBC from lung cancer patients and controls. Exhaled breath condensate (EBC) was collected non-invasively, using a handheld device (RTube) in ambulatory subjects engaged/recruited/consented through our pulmonary and thoracic surgery practices, under IRB protocol. A volume of 50 ul of EBC samples were combined with 200uL of methanol, containing 2 internal standards, 1nmol U13_succinate, 5nmol U13_citrate. Then, the samples were vortexed and 240 μl of supernatant was transferred to a sampling vial. The samples were dried under gentle nitrogen flow and derivatized with a two-step derivatization procedure, including a methoxyamine step for 90 minutes, and a silylation step for 60 minutes. QC sample was run multiple times during the analysis. The samples were analyzed by gas chromatography time-of-flight mass spectrometry (GC-TOFMS premier, Waters, USA). A number of 282 variables were detected after alignment and excluding any known artificial peaks, 49 of them were annotated. The data was normalized to the intensity of the sum of all the metabolites. The data set was then imported into SIMCA-p software (Umeå, Sweden) for multivariate analysis. A multivariate case-control ROC discriminant analysis compared the clinical model (AUC 0.72) to clinical plus exhaled small polar discriminant metabolites combined (AUC 0.92), showed incremental discrimination attributable to these metabolites (p=2.22e-62). ROC comparing clinical model (AUC 0.72) to clinical model plus metabolites (AUC 0.92); the difference attributable to metabolites was significant (p=2.22e-62). This exhaled biomarker platform can yield case-control discriminant small polar molecule sets related to known metabolic pathways, some of which are known to be deranged in cancers. Once further distilled and validated, our goal is to apply this non-invasive biomarker approach to prospective cohorts for non-invasive lung cancer risk assessment of the at-risk epithelium, in order to better select higher risk individuals to undergo effective CT screening. Supported by NIH-R21 CA192168-01; DoD-CDMRP- LC150738, NIH-NCI P60DK020541.

A Multi-Pass GAN for Fluid Flow Super-Resolution

We propose a novel method to up-sample volumetric functions with generative neural networks using several orthogonal passes. Our method decomposes generative problems on Cartesian field functions into multiple smaller sub-problems that can be learned more efficiently. Specifically, we utilize two separate generative adversarial networks: the first one up-scales slices which are parallel to the XY-plane, whereas the second one refines the whole volume along the Z---axis working on slices in the YZ-plane. In this way, we obtain full coverage for the 3D target function and can leverage spatio-temporal supervision with a set of discriminators. Additionally, we demonstrate that our method can be combined with curriculum learning and progressive growing approaches. We arrive at a first method that can up-sample volumes by a factor of eight along each dimension, i.e., increasing the number of degrees of freedom by 512. Large volumetric up-scaling factors such as this one have previously not been attainable as the required number of weights in the neural networks renders adversarial training runs prohibitively difficult. We demonstrate the generality of our trained networks with a series of comparisons to previous work, a variety of complex 3D results, and an analysis of the resulting performance.

Improved approach for LSC detection of 35S aiming at its application as tracer for short groundwater residence times

The knowledge of groundwater residence times in (vulnerable) aquifers is essential for the sustainable management of the associated groundwater resources. A powerful tool for related investigations is the application of naturally occurring radioisotopes as water age indicators. However, due to the limited number of suitable (i.e. omnipresent, short-lived and easily detectable) radionuclides only few studies focus on groundwater ages below one year. A natural radionuclide that does have the potential to cover this time range is 35S (87.4 day half-life). 35S is continually produced in the upper atmosphere and transferred with the rain to the groundwater. Since no natural sources of 35S exist in the subsurface the decrease of the 35S activity concentration in such young groundwater can be used for the determination of its age. Still, 35S activities in precipitation (and hence even more in groundwater) are very low and necessitate appropriate analytical protocols based on liquid scintillation counting (LSC). This turns out to be challenging due to the required large sample volumes and due to potentially high SO42− loads of the samples, both limiting the range of possible applications of 35S as indicator for short groundwater residence times. In the paper we present an improved straightforward LSC based approach for the detection of 35S in natural water samples. We recommend using Insta-Gel Plus as scintillation cocktail for allowing a homogeneous suspension of 35S-containing BaSO4 in the cocktail. The recommended improvements in instrument setting concern the LSC (TriCarb 3170 Tr/SL) counting window, the pulse decay discriminator setting and the delay before burst setting. The settings allow measuring low activity concentrations of 35S, which was previously pre-concentrated from natural water samples, containing SO42− loads of up to 1500 mg with a reasonably high statistical reliability.

Learning 3D joint constraints from vision-based motion capture datasets

Realistic estimation and synthesis of articulated human motion must satisfy anatomical constraints on joint angles. A data-driven approach is used to learn human joint limits from 3D motion capture datasets. We represent joint constraints with a new formulation (s1,s2,τ) using swing-twist representation in exponential maps form. Our parameterization is applied on Human3.6M dataset to create the lookup-map for each joint. These maps enable us to generate ‘synthetic’ datasets in entire joint rotation space of a given joint. A set of neural network discriminators is then trained with synthetic datasets to learn valid/invalid joint rotations. The discriminators achieve accuracy of [94.4−99.4%] for different joints. We validate precision-accuracy trade-off of discriminators and qualitatively evaluate classified poses with an interactive tool. The learned discriminators can be used as ‘priors’ for human pose estimation and motion synthesis.

Quantitative Imaging features Improve Discrimination of Malignancy in Pulmonary nodules

Pulmonary nodules are frequently detected radiological abnormalities in lung cancer screening. Nodules of the highest- and lowest-risk for cancer are often easily diagnosed by a trained radiologist there is still a high rate of indeterminate pulmonary nodules (IPN) of unknown risk. Here, we test the hypothesis that computer extracted quantitative features (“radiomics”) can provide improved risk-assessment in the diagnostic setting. Nodules were segmented in 3D and 219 quantitative features are extracted from these volumes. Using these features novel malignancy risk predictors are formed with various stratifications based on size, shape and texture feature categories. We used images and data from the National Lung Screening Trial (NLST), curated a subset of 479 participants (244 for training and 235 for testing) that included incident lung cancers and nodule-positive controls. After removing redundant and non-reproducible features, optimal linear classifiers with area under the receiver operator characteristics (AUROC) curves were used with an exhaustive search approach to find a discriminant set of image features, which were validated in an independent test dataset. We identified several strong predictive models, using size and shape features the highest AUROC was 0.80. Using non-size based features the highest AUROC was 0.85. Combining features from all the categories, the highest AUROC were 0.83.

Sleep Apnea Detection Using Artificial Bee Colony Optimize Hermite Basis Functions for EEG Signals

Sleep apnea is a sleeping disorder, which adversely affects the health of humans. The diagnosis of sleep apnea is possible by the detection of apnea events using electroencephalogram (EEG) recordings. This paper introduces an adaptive decomposition for the detection of apnea events using EEG signals. In introduced decomposition, the evolutionary techniques (ETs) optimized Hermite functions (HFs) represent the EEG signals. In tested ETs, the artificial bee colony (ABC) algorithm provides the least reconstruction error for the representation of EEG signals. The ABC is considered for Hermite coefficients-based feature extraction. From the extracted features, a highly discriminative feature set is obtained using the Fisher-score ranking test. The apnea detection performance of ranking-based selected features is evaluated using the extreme learning machine and least-squares support vector machine classifiers. The proposed method obtained performance measures, sensitivity 99.47%, specificity 99.58%, and accuracy 99.53%, are better as compared to the state-of-the-art methods.

Temporal progression of gene regulation of peripheral white blood cells explains gender dimorphism of critically ill patients after trauma

BackgroundThe immune response of the critically ill after severe trauma is sex-specific and may explain the different progression of the disease. This may be explained by a different gene regulatory program of their peripheral immune cells. We investigated the progression of the transcription profiles of peripheral immune cells of the patients to elucidate their distinct physiological response and clinical course.MethodsWe compared transcription profiles of whole blood of male and female patients from a larger longitudinal study of critically ill patients after trauma. We developed a statistical analysis pipeline that synchronized the time lapse of the profiles based on the temporal severity score of each patient.ResultsThis enabled to categorize the temporal progression of the disease into two pre-acute, an acute and two post-acute phases. Comparing gene regulation of male and female patients at each phase, we identified distinctively regulated molecular processes mainly in the immune response, but also in the regulation of metabolism allowing to cluster these discriminative gene sets into sets of highly related cellular processes. Compared to male patients and healthy controls, female patients showed upregulation of gene sets of innate immunity in the early phase, upregulation of wound healing processes during the acute phase and upregulation of adaptive immunity in the late phase indicating early recovery. In turn, during the pre-acute and acute phase, male patients showed less suppression of gene sets coding for enzymes of energy metabolism and anabolism, most prominently the tricarboxylic acid cycle and β-oxidation, and cellular maintenance, such as cell cycle, DNA replication and damage response, and RNA metabolism.ConclusionsA stronger innate immune response at the very early phase of the disease may support early clearance of the pathogen and its associated molecular patterns. Upregulation of wound healing processes may explain reduced multiple organ failure during the acute phase. Down regulated energy metabolism during the acute phase may make female patients less susceptible to oxidative stress, the upregulated adaptive immune system reflects an earlier recovery and rebuilding of the adaptive immune system that may protect them from secondary infections. Follow up studies need to be performed confirming these observations experimentally.

Lyapunov-Schmidt Method in Bifurcation Solutions of Nonlinear Fourth Order Differential Equation

this paper studied the bifurcation solutions of elastic beams equation by usingLyapunov-Schmidt method. The bifurcation equation corresponding to the elastic beamsequation has been found. Also, the Discriminant set (bifurcation set) of the elastic beamsequation has been found for some values of parameters.

Novel multispectral face descriptor using orthogonal walsh codes

This paper proposes a novel high-performance multispectral local descriptor that uses orthogonal Walsh codes during the generation of the discriminative feature set. Rotational variance and noise are compelling factors that significantly affect the distinctive performance of a facial identifier. The descriptor proposed in this article handles these challenges favourably sacrificing any distinctive performance. Orthogonal Walsh codes are used during the generation of the local descriptor. A Walsh code is assigned to each neighbour of a reference pixel. Before the assignment of an orthogonal code to each neighbouring pixel, these pixels are sorted in ascending order that consolidates the robustness of the method against rotational variances. Almost all methods proposed so far focus on grayscale images. However, colour bands contain important information about the relationship between pixels. Therefore, authors’ method considers RGB colour bands of pixels to improve distinctive performance. The results of extensive simulations show the remarkable and competitive performance of the proposed method regarding recognition accuracy and robustness against rotational variances as well as noise effects.

Using word embedding technique to efficiently represent protein sequences for identifying substrate specificities of transporters

Membrane transport proteins and their substrate specificities play crucial roles in various cellular functions. Identifying the substrate specificities of membrane transport proteins is closely related to protein-target interaction prediction, drug design, membrane recruitment, and dysregulation analysis, thus being an important problem for bioinformatics researchers. In this study, we applied word embedding approach, the main cause for natural language processing breakout in recent years, to protein sequences of transporters. We defined each protein sequence based on the word embeddings and frequencies of its biological words. The protein features were then fed into machine learning models for prediction. We also varied the lengths of protein sequence's constituent biological words to find the optimal length which generated the most discriminative feature set. Compared to four other feature types created from protein sequences, our proposed features can help prediction models yield superior performance. Our best models reach an average area under the curve of 0.96 and 0.99, respectively on the 5-fold cross validation and the independent test. With this result, our study can help biologists identify transporters based on substrate specificities as well as provides a basis for further research that enriches a field of applying natural language processing techniques in bioinformatics.

Current Strategy for Local- to Global-Level Molecular Epidemiological Characterisation of Global Antimicrobial Resistance Surveillance System Pathogens

AbstractThe prime goal of molecular epidemiology is to identify the origin and evolution of pathogens, which can potentially influence the public health worldwide. Traditional methods provide limited information which is not sufficient for outbreak investigation and studying transmission dynamics. The recent advancement of next-generation sequencing had a major impact on molecular epidemiological studies. Currently, whole-genome sequencing (WGS) has become the gold standard typing method, especially for clinically significant pathogens. Here, we aimed to describe the application of appropriate molecular typing methods for global antimicrobial resistance surveillance system pathogens based on the level of discrimination and epidemiological settings. This shows that sequence-based methods such as multi-locus sequence typing (MLST) are widely used due to cost-effectiveness and database accessibility. However, WGS is the only method of choice for studying Escherichia coli and Shigella spp. WGS is shown to have higher discrimination than other methods in typing Klebsiella pneumoniae, Acinetobacter baumannii and Salmonella spp. due to its changing accessory genome content. For Gram positives such as Streptococcus pneumoniae, WGS would be preferable to understand the evolution of the strains. Similarly, for Staphylococcus aureus, combination of MLST, staphylococcal protein A or SCCmec typing along with WGS could be the choice for epidemiological typing of hospital- and community-acquired strains. This review highlights that combinations of different typing methods should be used to get complete information since no one standalone method is sufficient to study the varying genome diversity.

Perceived discrimination among older adults living in urban and rural areas in Brazil: a national study (ELSI-Brazil)

BackgroundResearch on discrimination and health focused on older adults has been scarce, comparatively with younger and middle-aged adults. Considering where people live matters, accurate measures of perceived discrimination might consider how the place of residence interferes on discriminatory experiences. This study aimed to assess the association between perceived discrimination and urban/rural place of residence among a representative sample of older adults in Brazil.MethodsData came from the baseline of the Brazilian Longitudinal Study of Aging (ELSI-Brazil), conducted in 2015/2016, with individuals aged 50 years and older. Perceived Discrimination was measured by means of the following question: “In the past 12 months have you felt a victim of any type of discrimination” with five possible answers: (1)“when you sought medical services or health care?”, (2)“in social gatherings?”, (3)“in the work place?”, (4)“within the family?”, (5)“due to where you live?”. Participants who answered yes for any of the five domains were coded as having reported an experience of discrimination. The main exposure variable was the urban-rural classification of the households, carried out according to the methods employed by the Brazilian Institute of Geography and Statistics during the 2010 Population Census. Other covariates included: age, sex, skin color, household wealth and education. Multiple Poisson regression was used to estimate prevalence ratios and their respective 95% confidence interval for the association between discrimination and independent variables.ResultsPrevalence of any perceived discrimination among Brazilian older adults was 16.8%. Regardless the place of residence (either urban or rural), participants reported health care settings as the most common domain where discriminatory experiences occurred and the work place as the least common. According to the adjusted model, perceived discrimination was significantly higher among urban dwellers when compared to their rural counterparts, independent of sociodemographic characteristics, health status and neighborhood social environment. The outcome was significant associated with skin color, education and health status.ConclusionsUrban environment plays a core role in perceived discrimination and health care settings constitute the most common domain where discriminatory experiences occurred. Our findings may contribute to fulfill the knowledge gap on discrimination among older adults living in developing countries.

Computation of the Resonance Set of a Polynomial under Constraints on Its Coefficients

Methods for computing the generalized discriminant set of a polynomial the roots of which satisfy a linear relation are considered. Using a q-analog of the classical elimination theory and computer algebra algorithms, methods for computing the parametric representation of this set are described, and these methods are implemented in a Maple library. The operation of these methods is demonstrated by an example.

Combination of hand-crafted and unsupervised learned features for ischemic stroke lesion detection from Magnetic Resonance Images

Detection of ischemic stroke lesions plays a vital role in the assessment of stroke treatments such as thrombolytic therapy and embolectomy. Manual detection and quantification of stroke lesions is a time-consuming and cumbersome process. In this paper, we present a novel automatic method to detect acute ischemic stroke lesions from Magnetic Resonance Image (MRI) volumes using textural and unsupervised learned features. The proposed method proficiently exploits the 3D contextual evidence using a patch-based approach, which extracts patches randomly from the input MR volumes. Textural feature extraction (TFE) using Gray Level Co-occurrence Matrix (GLCM) and unsupervised feature learning (UFL) based on k-means clustering approaches are employed independently to extract features from the input patches. These features obtained from the two feature extractors are then given as input to the Random Forest (RF) classifier to discriminate between normal and lesion classes. A hybrid approach based on the combination of TFE using GLCM and UFL based on the k-means clustering is proposed in this work. Hybrid combination approach results in more discriminative feature set compared with the traditional approaches. The proposed method has been evaluated on the Ischemic Stroke Lesion Segmentation (ISLES) 2015 training dataset. The proposed method achieved an overall dice coefficient (DC) of 0.886, precision of 0.979, recall of 0.831 and accuracy of 0.8201. Quantitative measures show that the proposed approach is 28.4%, 27.14%, and 5.19% higher than the existing methods in terms of DC, precision, and recall, respectively.

Template-based gait authentication through Bayesian thresholding

While gait recognition is the mapping of a gait sequence to an identity known to the system, gait authentication refers to the problem of identifying whether a given gait sequence belongs to the claimed identity. A typical gait authentication system starts with a feature representation such as a gait template, then proceeds to extract its features, and a transformation is ultimately applied to obtain a discriminant feature set. Almost every authentication approach in literature favours the use of Euclidean distance as a threshold to mark the boundary between a legitimate subject and an impostor. This article proposes a method that uses the posterior probability of a Bayes &#x02BC classifier in place of the Euclidean distance. The proposed framework is applied to template-based gait feature representations and is evaluated using the standard CASIA-B gait database. Our study experimentally demonstrates that the Bayesian posterior probability performs significantly better than the de facto Euclidean distance approach and the cosine distance which is established in research to be the current state of the art.

Instantaneous fMRI based cerebral parameters for automatic Alzheimer, mild cognitive impairment and healthy subject classification.

Automatic identification and categorization of Alzheimer's patients and the ability to distinguish between different levels of this disease would be very helpful to the research community studying this disease since non-automatic approaches are both very time-consuming and highly dependent upon the experience of experts. Here, it is proposed that instantaneous cerebral phase and envelope information from functional magnetic resonance imaging data is of use to discriminate between Alzheimer's patients, mild cognitively impaired subjects and healthy individuals. Following a region-of-interest analysis of functional magnetic resonance imaging data, different features including power, entropy, and coherency features are derived from the instantaneous phase and envelope signal sequences. Various sets of features are calculated and fed to a sequential forward floating feature selection algorithm to identify the most discriminative and informative feature sets. A Student's t-test was employed to select the most relevant features from the sets. Finally, a K-nearest neighbor classifier is used to distinguish between classes in a three-class categorization problem. The reported performance in overall accuracy using functional magnetic resonance imaging data of 111 combined participants is 80.1% with 80.0% sensitivity for the distinction of both Alzheimer's and healthy categories. This is comparable to the state-of-the-art approaches recently proposed for this task. The significance of obtained results was statistically confirmed by the evaluation of standard classification performance indicators. Results illustrate that the analytic phase and envelope feature indices derived from the region of interest signals described here are significant discriminators suited to distinguish between Alzheimer patients and healthy subjects.

Hand recognition in gaming

Hand motion acknowledgment gives a savvy and characteristic method for human PC collaboration (HCI). It has applications running from customer hardware to control of therapeutic recovery (for example cell phone). So as to recognize hand motions, different sorts of detecting methods are used to get signals for example acknowledgment. Increasing speed base and electromyogram-based strategies are two research branches in the field of hand signal example acknowledgment. Quickening based (ACC-based) motion control is typically contemplated as a valuable connection methodology. It is appropriate to recognize perceptible, bigger scale signals with various hand directions of lower arm developments. With ACC-based strategies some unobtrusive finger or hand development might be overlooked while electromyogram-based (EMG-based) motion acknowledgment systems use multi-channel EMG signals which contain rich data about hand motions of different size scales. Because of certain issues natural in the EMG estimations, including the distinguishableness and reproducibility of estimation, the size of discriminable hand signal set is as yet restricted to 4-8 classes. So as to understand a characteristic and powerful signal based HCI framework, the determination of info hand motions that are well discriminable from one another is of essential significance. Thinking about the integral highlights of ACC-and EMG-estimations, we accept that their mix will expand the quantity of discriminable hand, wrist and lower arm signals and the precision of the acknowledgment framework. This paper depicts In IP Gaming we are proposing a framework wherein without utilizing sensors and Devices, we are recognizing the hand and signal with Simple Web camera and playing out the Image Processing system in which utilizing those motion, we can play game on reassure. In Image Process Gaming, the movements are identified through a web camera. These pictures are then passed for the picture handling. The procedures utilized for picture handling are hand motion identification, edge recognition, thresholding, form location. Utilizing OpenCV, which gives a library accumulation of capacities for various picture handling systems, these info pictures can be prepared and relating key strokes will be created. As of late, PC vision improvement has extraordinary headways and our everyday life undertakings are fragmented without utilizing PCs. The real info gadgets like Keyboard and mouse are accustomed to communicating with PCs. Among the different communication systems utilization of hands as an information is an appealing technique for building up regular Human Computer Interaction. By utilizing Hand motions client can convey more data in less Timespan. So for improving the interface among clients and PCs human PCs collaboration (HCI) innovation has incredible use.

Live fingerprint detection using magnitude of perceived spatial stimuli and local phase information

Fingerprint recognition systems are widely used for authentication purposes in security systems. However, fingerprint recognition systems can easily be spoofed by imitations of fingerprints using various spoof materials. A compact and discriminative set of features is needed to discriminate between live and spoof fingerprints. We explore combined Shepard magnitude and orientation for live fingerprint detection using independent quantization of global and local features extracted in spatial and frequency domain. The spatial domain features that are extracted comprise of the magnitude of perceived spatial stimuli that is computed from the net variation of perceived edge information. Rotation invariance is achieved by extracting local features based on phase information of significant frequency components in the frequency domain. The concatenated feature vector associated with a fingerprint image is represented as a two-dimensional histogram. The support vector machine classifier is used to classify the fingerprint as either live or spoof. Experiments are performed on three databases, i.e., the fingerprint liveness detection (LivDet) competition databases of 2011, 2013, and 2015. Results showed a reduction in average error rate to 5.8, 2.2, and 5.3 on LivDet 2011, 2013, and 2015, respectively.

P1.11-12 Interrogation of an Exhaled MicroRNA Panel for Lung Cancer Risk Assessment

There is a need for non-invasive airway-based biomarkers in lung carcinogenesis for both risk assessment of the ex-smoker, and earlier diagnosis. Exhaled breath condensate (EBC) contains airway lining fluid molecules, including nucleic acids, presumably in part from epithelial cellular origins. MicroRNAs play important regulatory roles in many processes, including carcinogenesis. Here we further develop and begin validation of the detection of microRNAs in EBC from lung cancer patients and controls. EBC was collected non-invasively, using a handheld commercial RTube® device in a clinical series of ambulatory subjects. We generated a 40 miR panel based on literature-derived microRNAs, bronchial brushed microRNA discovery by collaboration, and a lung parenchymal tissue discovery effort we have performed using microRNA-seq on lung tumors and surrounding non-tumor tissue. The PCR primers were designed using our previously published RNA-specific RT-PCR technique. All samples were run twice with positive and negative controls. Calls was made of individual miR present or absent in a given EBC sample, by one of two replicates being positive. The data were analyzed by random forests. We applied the panel to EBC from 177 individuals, 89 NSCLC predominantly early stage (I and II) cases and 88 controls of similar smoking history. We used an empirically-derived base clinical model that included age, smoking status, pack-years, quit-years, and underlying lung disease. MicroRNA signatures alone discriminated cases from controls with ROC-AUCs of 0.64-0.76. Both analyses revealed a small set of incrementally informative miRs (e.g., miRs-21**, 33b**, 96*,105*, `130b.3p*, 200a*, 200b**, 205, 212***, 221***, 345***, 767**, 944*, 1269a***, 1293*, 1910**, 3648*, 3662*, where number of ‘*’ implies number of appearances in each of three models). Here some modest incremental case-control discriminant capacity was conferred by the respective microRNA signature over and above the base clinical model. The magnitude of the increment was typically 2-3% [RF AUC 0.84 clinical model=>0.86 clinical+microRNA (all lung cancer histologies, Welch p=1.33e-05); 0.82=>0.84 (adenocarcinomas, Welch p=3.13e-03); 0.84=>0.86 (NSCLC, Welch p=4.10e-05). Quantitative RT-PCR using this platform set-up was not sufficiently robust to further analyze. This new exhaled biomarker platform can yield case-control discriminant microRNA sets. albeit modest in incremental impact as formulated so far. Once quantified, further distilled and validated, our goal is to test this non-invasive biomarker approach to prospective cohorts for non-invasive lung cancer risk assessment, in order to non-invasively better select higher risk individuals to undergo effective CT screening.

First-Person Daily Activity Recognition With Manipulated Object Proposals and Non-Linear Feature Fusion

Most previous works on the first-person video recognition focus on measuring the similarity of different actions by using low-level features of objects interacting with humans. However, due to noisy camera motion and frequent changes in viewpoint and scale, they fail to capture and model highly discriminative object features. In this paper, we propose a novel pipeline for the first-person daily activity recognition. Our object feature extraction pipeline is inspired by the recent success of object hypotheses and deep convolutional neural network (CNN)-based detection frameworks. Our key contribution is a simple yet effective manipulated object proposal generation scheme. This scheme leverages motion cues, such as motion boundary and motion magnitude (in contrast, camera motion is usually considered as “noise” for most previous methods), to generate a more compact and discriminative set of object proposals, which are more closely related to the objects, which are being manipulated. Then, we learn more discriminative object detectors from these manipulated object proposals based on region-based CNN. Meanwhile, we develop a non-linear feature fusion scheme, which better combines object and motion features. We show in experiments that the proposed framework significantly outperforms the state-of-the-art recognition performance on a challenging first-person daily activity benchmark.