Apoptosis is a tightly regulated cellular process that plays an essential role in the development, aging, cancer biology, immune response, and pathogenesis of various diseases. Herein, we report a new SERS sensing strategy for in vitro sensitive detection of early apoptotic cells. The principle of this method is to in situ synthesize silver nanoparticles (AgNPs) on the phosphatidylserine (PS) of the apoptotic cell mem- brane during the early apoptosis, which enables distinguishing normal and apoptotic cells. The total assay time of the presented method is only 10 min, thus being faster, cheaper and simpler than current tech- niques for the detection of apoptosis. The intrinsic mechanism was veri ﬁed by diﬀerent approaches based on externalized phosphatidylserine. In addition, the detection process is real-time and label-free; i.e., the intrinsic SERS spectra from the cellular membrane are directly employed for apoptosis real-time detec- tion, which avoids using additional chemical or biological reagents as external signal indicators. Therefore, our SERS approach may serve as a potentially practical tool for sensitive and real-timedetection of early cell apoptosis, complementing the state-of-the-art strategies, e.g. ﬂow cytometry. While further investi- gation is required to better understand the intrinsic mechanism of the in situ coating method, the current results may provide another choice for real-timedetection of early apoptosis.
methodologies are available. It turns out that simple non-parametric procedures, such as the famous Bry and Boschan (1971) procedure still used by the Dating Committee of the NBER, are more convenient for this kind of work (see Harding and Pagan, 2001, or Anas and Ferrara, 2002b, for a discussion on this issue). Real-timedetection refers mainly to short-term economic analysis, which is not an easy task for practitioners. Indeed, several economic indicators are released on a regularly monthly basis, or even on a daily basis as regards the financial sector, adding volatility to the existing volatility and thus leading to an inflation of the available information set. Moreover, the data are often strongly revised and the diverse statistical methods, such as seasonal adjustment or filtering techniques, lead to edge effects. In this framework, the statistician has a crucial role to play which consists in extracting the right signal to help the short-term economic diagnosis. The too often quoted word ”data miner” seems to be here well appropriate. Therefore, the real-time economic analysis asks for methods with strong statistical content.
V. C ONCLUSION
Successful development of a micro-total-analysis system for the continuous detection of the low-limits gaseous formaldehyde is highly desired since this carcinogenic substance largely used in the fabrication of household products is continually released indoors. This project aims to study the possible paths towards on-chip real-timedetection of low-limits indoor formaldehyde concentrations, a laboratory prototype being today developed based on the Hantzsch reaction coupled to the optical fluorescence detection method. The gas-liquid micro-reactor relies on a disposable PMMA gas-liquid contacting chip that uses as separation medium a hydrophobic polymer membrane. The fluorescence optical detection system combines the contact sensing of a disposable quartz/SU-8 interrogation fluidic cell with the CMOS time-resolved spectroscopy. After the
Researchers showed a lot of interest in studying whispering gallery microcavities as a tool for biosensing in the last decade. Optical microcavities are structures that confine light at the microscale due to total internal reflection of light at the interface between the cavity and its surrounding medium. If a molecule binds to the surface of the microcavity, light can interact with it several times, making optical microcavities very sensitive tools for label-free sensing. During this Ph.D. project, optical microdisks are used to detect the presence of Staphylococ- cus aureus (S. aureus) bacteria. To our knowledge, this is the first time optical microdisks are used to specifically detect bacteria. In order to have a reliable and efficient biosensor, it needs to be highly specific. Specificity is achieved by choosing an appropriate functionaliza- tion process. The functionalization process uses the antibody that is specific to the antigen of interest. In this case, the choice of a specific bacteriophage to bind S. aureus bacteria is crucial to obtain a specific sensor, and many experiences were done in order to identify the most appropriate. However, the purification of bacteriophages can be long and complex. An alternative to working with whole bacteriophages is the use of purified protein phages that can be easier to prepare. The functionalization process used in this thesis was developed in collaboration with professor Jay L. Nadeau’s group from the biomedical engineering depart- ment at McGill university. LysK protein phage is added to the microdisk and will attach S. aureus bacteria during the real-timedetection experiments. In order to demonstrate the specificity of the functionalization process, LysK was used with E. coli bacteria. As predicted, since LysK is only specific to S. aureus strains, it did not attach any E. coli.
50 minutes after injection of LS102.9 cells (300 μL at 10 4
cells.mL -1 ), 4 and 5 cells could be counted on the two anti-CD19 spots (Figure 2b) whereas 0 to 2 cells could barely be detected on four control spots. By taking into account the reactor size that is 1.5 mm in height, the spot diameter (c.a. 700 μm) and the cell concentration, one may estimate the number of cells suspended above each spot to about 6 cells. Thus, the designed biochip allowed the binding and SPR detection of approximately 3/4 of the
UV-LIF spectrometers such as the WIBS have many ad- vantages over traditional bioaerosol sampling methods, e.g. online single-particle detection and high time resolution; however, some non-biological fluorescent interferent parti- cles can also show weak auto-fluorescence and so can be a source of false positives resulting in potential artefacts when interpreting biological materials. This means there can be difficulties discriminating some classes of biological par- ticles unambiguously. Generally the majority of identified interferent non-biological fluorescent aerosols have fluores- cence levels similar to the detection limit of the instrument; for example PAHs and PAH-containing soot particles of small diameter (< 1 µm) have been demonstrated to fluoresce only weakly in FL1 (Toprak and Schnaiter, 2013; Pöhlker et al., 2012). However, we would not expect to see signifi- cant concentrations of PAHs or soot particles at this remote site outside of long-range transport events. Mineral dusts also contain a small subset of very weakly fluorescent particles due to the presence of luminescence centres within the min- erals. These are often associated with rare earth elements, but their observed fluorescent intensity is considerably weaker than observed for primary bio-fluorophores (Pöhlker et al., 2012). Given the ubiquitous nature of mineral dusts, these weakly fluorescent dust sub-categories may therefore present a significant, even dominant, fraction of recorded fluorescent material at the measurement site, particularly during long- range transport events. Thus, they would likely form their own population clusters, as demonstrated in Crawford et al. (2016).
analysis .  For the present study, that quaternization of pyridyls to produce pyridinium groups enhances anion binding affinity by electrostatic interactions.
We report herein an efficient and portable sensing platform for the transduction of anion binding events to create electrical signals, which makes use of squaramide-based selectors enhanced by proximality to a cationic pyridinium moieties that wrap the SWCNTs. This platform has capability for real-timedetection of multiple anions in small samples (2 μL) by wireless communication methods.
This servoing task requires a robust real-timedetection of the needle. Most of the existing needle localization methods in the literature are based on parallel projections, which are designed to find imperfect instances of parameterized shapes by optimizing the integral of the image along parallel curves. Hong et al.  use the Hough transform (HT) to detect straight needles in two-dimensional US images. Aboofazeli et al.  adapt this method to the detection of mildly curved needles in a 3D volume, by first projecting the volume onto 2D planes by a ray casting process. Several variants of the Hough transform allow to reduce the computational time, using coarse-fine strategies  or randomization . Barva et al.  use the parallel integral projection (PIP) to localize straight needles in 3D volumes. Uherˇcik et al.  develop a multi-resolution scheme (MR-PIP) to speed-up the com- putation. Novotny et al.  optimize this algorithm for an implementation on graphics processing unit (GPU) to achieve real-timedetection. Although the Hough transform was first designed for the detection of straight lines, it is applicable to any parameterized curve. Neshat and Patel  represent a curved needle by a B´ezier polynomial, and optimize its parameters using the Radon transform (equivalent to PIP). The algorithm is also implemented on GPU for real-timedetection.
We describe a simple way to reduce the amount of required training data in context-based models of real- time object detection. We demonstrate the feasibility of our approach in a very challenging vehicle detection scenario comprising multiple weather, environment and light conditions such as rain, snow and darkness (night). The investigation is based on a real-timedetection system effectively composed of two trainable components: an exhaustive multiscale object detector (”signal-driven detection”), as well as a module for generating object-specific visual attention (”context models”) controlling the signal-driven detection process. Both parts of the system require a significant amount of ground-truth data which need to be generated by human annotation in a time-consuming and costly process.
Ubiquitous cardiac care (UCC) is one of pervasive healthcare services, which needs a dedicated real- time AED (Ambulatory Electrocardiograph Detection) algorithm that should have at least two significant characteristics: real-timedetection capability for cardiac arrhythmia events; a small resource requirement for its computation and storage. As the body-surface manifestation (mV) of cardiac electrical potentials, QRS (Q-R-S waves) complexes in ECG signals have a remarkable waveform shape (high potential am- plitude, steep slope) and plentiful waveform information, which are thus utilized in AED algorithms as an essential basis for QRS complex detection.
5.4 On-line Change Point Detection
The above change point detection experiments were performed in a off-line mode for bcp and ecp, and for all CPD algorithms using LDA: CPD algorithms read all data points at all time intervals at once. For real-timedetection, however, it is necessary to test the performance in a real on-line fashion. In order to simulate that with bcp and ecp, we use a sliding window, feeding the algorithm overlapping slices of the data, step by step. Although ocp and ocp+ are built as on-line CPD algorithms, processing one new data point at a time, we used the same sliding window to get a fair comparison. On the Olympics and Zika datasets, we tested different sliding window sizes and steps to evaluate the impact of this size on performance. Table 4 shows the result obtained by the four CPD algorithms on either oLDA ∞ t topic scores or baseline message counts. On the Olympics data, oLDA ∞ t performs better than counts in all CPD algorithms except bcp. With smaller steps and window sizes, ecp performs the best. ocp performs best on the largest step and sliding window sizes (100 and 200 hours, respectively). The best performance for different settings (bold) are very close. On the Zika data, the performance of CPD algorithms on counts and oLDA ∞ t are not consistent. ocp+ reaches the best performance on counts using a 150 day window size in 75 day steps, and performs well on oLDA ∞ t in two other settings, while ecp performs well on counts in the remaining situation (100 day window, 50 day steps). Comparing the best performance obtained in off-line and on-line modes on these two datasets (Table 3 vs. Table 4), we see that the on-line CPD algorithms using sliding windows achieve better performance than using the entire data set off-line. 5.5 Error Analysis
2.4.3 Methods based on Pattern Recognition
In the field of computer vision, the problem of linking semantics between high level concepts and low level features has been studied. Pattern recognition is a branch of machine learning that focuses on the recognition of patterns and regularities in data [Bishop 2006]. Classification methods used in computer vision are often based on the extraction of features that may include color, texture, shape, or spatial relation informations. Recently, different sophisticated feature extraction techniques have been proposed. An obstacle detection system based on color and/or texture feature algorithms requires an important amount of calculations. Some optimizations have been proposed in [Cervantes 2008] in order to reduce computation time by performing a color-based segmentation before applying a region based classification. The proposed approach is however not suitable for all applications. In [Manzano 2010], a hardware based architecture based on texture-color classification is proposed in order to detect specific objects on the scene. Classifier parameters are learnt offline from ground examples. Hence, the proposed system is able to detect obstacles on indoor environment as shown in figure 2.8.
In our results, the qAP-16S probe was not completely specific for AP detection in real-time PCR. ESFY isolates were also detected by this probe. It is probably due to the presence of only one mismatch between ESFY sequence and probe. Besides, it seems Baric and Dalla-Via set the threshold line on inflexion point while we set a threshold line corresponding to 10 x S.D. The 10 x S.D. is more popular than inflexion point method to set the threshold line (Ginzinger, 2002). Moreover, it is generally recommended to use a threshold line close to the baseline. Indeed, fluorescent curves may diverge at higher fluorescent level while lower thresholds will minimize the error due to small changes in efficiency (Hunt, 2005). Fitting the intersecting line upon the ten-times value of ground fluorescence standard deviation can be easily automated and is very robust (Pfaffl, 2003).
In the category of state-of-the-art detectors, Convolutional Neural Net- works are now the reference in terms of accuracy. Deep Learning became popular in object classification, when researchers, notably Krizhevsky [ KSH12 ] managed to beat the state-of-the-art on classification in Ima- geNet dataset [ Den+09 ] using CNNs to train the network AlexNet, out- ranking previous methods based on Bag of Words [ Csu+04 ], and reach- ing performances in image classification close to the human [ Rus+14 ]. AlexNet was followed by different CNNs such as Overfeat [ Ser+13 ], VGG [ SZ14 ]... Girshick [ Gir+14 ] managed to beat state-of-the-art detectors on Pascal VOC datasets [ Eve+07 ]; [ Eve+10 ] by proposing R-CNN. From an image, Girshick extracts many search windows. Each candidate window is then tested using CNNs to determine if the image belongs to one class (Pascal VOC contains several classes). Girshick also considers two CNNs: one for classification, using ImageNet for training [ Den+09 ], one for lo- calization (using a subset of Pascal VOC for training). Many works based on R-CNN aimed to enhance training and detection speed [ Ren+15 ]; [ Red+16 ], or to enhance accuracy [ ZD14 ] (for this category, [ HBS14 ] proposed a survey of different contributions).
The past decade has been marked with important strides towards clinical translation of RS in vivo. Technological advances in CCD detectors, lasers, and fibered optics have allowed for spectral acquisitions with performance and time scales suitable for in vivo use. Zhao et al conducted a large study on 289 patients where they performed measurements on suspect skin lesions with acquisition times under 1 second. Using partial least squares regression and linear discriminant analysis on the spectra, they achieved 91% sensitivity and 75% specificity for differentiating skin cancers from benign legions and 97% sensitivity/78% specificity in differentiating malignant melanoma from benign pigmented lesions. For more difficult to reach organs, RS has been integrated to a variety of endoscopic systems for investigative work . Draga et al. used an RS probe to obtain Raman spectra from ‘suspicious’ and ‘nonsuspicious’ bladder locations with acquisition times under 5 seconds. They reported a sensitivity of 85% and 79% in distinguishing normal from cancerous bladder locations. In brain cancer, invasive cancer cells frequently remain after surgery, leading to disease recurrence and a negative impact on overall survival. Our research group at Polytechnique Montreal has developed a handheld RS probe for intraoperative collection of Raman spectra. The following figure shows a view of the handheld probe used during open cranium glioblastoma resection.
Figure 12: Left: T emplate for a single node M i . Right: Template for the overall spe
an forward the token by outputting on rec i+1 , but only after a delay between d and D time units. Fig. 11 (right) illustrates a ring of su
h nodes M i in whi
h some nodes have been grouped together. This grouping exemplies
This study is focused on the fabrication of an aeronautical sensing coating. We propose giving a piezoelectric behavior to a polyurethane matrix composite by insertion of BT particles. Polyurethane spray coating technology is well known in the aeronautical industry. Since the tetragonality of BT decreases with particle size [ 8 ], 300 nm diameter particles have been chosen. The challenge is to ensure a synergy between the electroactive properties of the ceramic and the attractive engineering behavior of polyurethane. Furthermore, the feasibility of realtime location on aircraft polyepoxy/carbon fiber reinforced composite (CFRC) with a cross correlation technique [ 12 ] is reported.
During the past decade, the need for performance and the tremendous integration technology progress have allowed inexpensive superscalar processors featuring instruction level parallelism as well as memory hierarchy. Moreover, nowadays, these processors targeting large scale volumes are also featuring hardware thread level parallelism. Economic pressure is pushing for the use of these standard processor components or IPs in every application domain. As a consequence, an application domain can only marginally influence the design of the standard processors. That is the processor manufacturers will consider adding a feature to a standard processor only if the application domain generates huge volume, or if the modification is very marginal and does not impair the rest of the design. This applies for real-time systems.
It should now be easy to see how the elements above match RTMRP (E, π π π), Defini- tion 12 .
Interestingly, the RTMDP is equivalent to a 1-step constant delay MDP ( Walsh et al. [ 2008 ]). However, we believe the different intuitions behind both of them warrant the different names: The constant delay MDP is trying to model external action and observation delays whereas the RTMDP is modelling the time it takes to select an action. The connection makes sense, though: In a framework where the action selection is assumed to be instantaneous, we can apply a delay to account for the fact that the action selection was not instantaneous after all.