INESC TEC Repository
Not a member yet
2130 research outputs found
Sort by
Fabrication of Multimode-Single Mode Polymer Fiber Tweezers for Single Cell Trapping and Identification with Improved Performance
Get PDFOptical fiber tweezers have been gaining prominence in several applications in Biology and Medicine. Due to their outstanding focusing abilities, they are able to trap and manipulate microparticles, including cells, needing any physical contact and with a low degree of invasiveness to the trapped cell. Recently, we proposed a fiber tweezer configuration based on a polymeric micro-lens on the top of a single mode fiber, obtained by a self-guided photopolymerization process. This configuration is able to both trap and identify the target through the analysis of short-term portions of the back-scattered signal. In this paper, we propose a variant of this fabrication method, capable of producing more robust fiber tips, which produce stronger trapping effects on targets by as much as two to ten fold. These novel lenses maintain the capability of distinguish the different classes of trapped particles based on the back-scattered signal. This novel fabrication method consists in the introduction of a multi mode fiber section on the tip of a single mode (SM) fiber. A detailed description of how relevant fabrication parameters such as the length of the multi mode section and the photopolymerization laser power can be tuned for different purposes (e.g., microparticles trapping only, simultaneous trapping and sensing) is also provided, based on both experimental and theoretical evidences.</jats:p
A Weakly-Supervised Framework for Interpretable Diabetic Retinopathy Detection on Retinal Images
Get PDFDiabetic retinopathy (DR) detection is a critical retinal image analysis task in the context of early blindness prevention. Unfortunately, in order to train a model to accurately detect DR based on the presence of different retinal lesions, typically a dataset with medical expert's annotations at the pixel level is needed. In this paper, a new methodology based on the multiple instance learning (MIL) framework is developed in order to overcome this necessity by leveraging the implicit information present on annotations made at the image level. Contrary to previous MIL-based DR detection systems, the main contribution of the proposed technique is the joint optimization of the instance encoding and the image classification stages. In this way, more useful mid-level representations of pathological images can be obtained. The explainability of the model decisions is further enhanced by means of a new loss function enforcing appropriate instance and mid-level representations. The proposed technique achieves comparable or better results than other recently proposed methods, with 90% area under the receiver operating characteristic curve (AUC) on Messidor, 93% AUC on DR1, and 96% AUC on DR2, while improving the interpretability of the produced decisions
Multi-Temporal Vineyard Monitoring through UAV-Based RGB Imagery
Get PDFThis study aimed to characterize vineyard vegetation thorough multi-temporal monitoring using a commercial low-cost rotary-wing unmanned aerial vehicle (UAV) equipped with a consumer-grade red/green/blue (RGB) sensor. Ground-truth data and UAV-based imagery were acquired on nine distinct dates, covering the most significant vegetative growing cycle until harvesting season, over two selected vineyard plots. The acquired UAV-based imagery underwent photogrammetric processing resulting, per flight, in an orthophoto mosaic, used for vegetation estimation. Digital elevation models were used to compute crop surface models. By filtering vegetation within a given height-range, it was possible to separate grapevine vegetation from other vegetation present in a specific vineyard plot, enabling the estimation of grapevine area and volume. The results showed high accuracy in grapevine detection (94.40%) and low error in grapevine volume estimation (root mean square error of 0.13 m and correlation coefficient of 0.78 for height estimation). The accuracy assessment showed that the proposed method based on UAV-based RGB imagery is effective and has potential to become an operational technique. The proposed method also allows the estimation of grapevine areas that can potentially benefit from canopy management operations.</jats:p
Loading constraints for a multi-compartment vehicle routing problem
No full textMulti-compartment vehicles (MCVs) can deliver several product segments jointly. Separate compartments are necessary as each product segment has its own specific characteristics and segments cannot be mixed during transportation. The size and position of the compartments can be adjusted for each tour with the use of flexible compartments. However, this requires that the compartments can be accessed for loading/unloading. The layout of the compartments is defined by the customer and segment sequence, and it needs to be organized in a way that no blocking occurs during loading/unloading processes. Routing and loading layouts are interdependent for MCVs. This paper addresses such loading/unloading issues raised in the distribution planning when using MCVs with flexible compartments, loading from the rear, and standardized transportation units. The problem can therefore be described as a two-dimensional loading and multi-compartment vehicle routing problem (2L-MCVRP). We address the problem of obtaining feasible MCV loading with minimal routing, loading and unloading costs. We define the loading problem that configures the compartment setup. Consequently, we develop a branch-and-cut (B&C) algorithm as an exact approach and extend a large neighborhood search (LNS) as a heuristic approach. In both cases, we use the loading model in order to verify the feasibility of the tours and to assess the problem as a routing and loading problem. The loading model dictates the cuts to be performed in the B&C, and it is used as a repair mechanism in the LNS. Numerical studies show that the heuristic reaches the optimal solution for small instances and can be applied efficiently to larger problems. Additionally, further tests on large instances enable us to derive general rules regarding the influence of loading constraints. Our results were validated in a case study with a European retailer. We identified that loading constraints matter even for small instances. Feasible loading can often be achieved only through minor changes to the routing solution and therefore with limited additional costs. Further, the importance to integrate loading constraints grows as the problem size increases, especially when a heterogeneous mix of segments is ordered. © 2018 Springer-Verlag GmbH Germany, part of Springer Natur
Load balance recovery for multi-drop distribution problems: A mixed integer linear programming approach
Get PDFIn road freight transport, a loaded vehicle with a distribution route and a compliant load balance at the depot can become non-compliant during the route, since the total weight of the cargo and its centre of gravity change with each delivery. Nowadays, vehicles circulating on our roads either undermine safety regulations or lack operational efficiency when these regulations are taken into account and cargo is extensively rearranged after each delivery. This issue has been completely ignored both in the vehicle routing literature and in the container loading literature. The aim of this work is to provide tools capable of ensuring that a cargo arrangement is load balanced along the complete distribution trip. It proposes a multi-drop load balance recovery algorithm (MDLBRA), which seeks to ensure that, when both a complete route and the respective cargo arrangement are provided, the boxes to be removed from the cargo arrangement at the depot and the boxes to be rearranged at each customer are identified, allowing the cargo to remain balanced after every delivery. It is important to notice that a MDLBRA is not a container loading algorithm: a MDLBRA modifies solutions generated by any container loading algorithm so that load balance is guaranteed when the truck leaves the depot and during the entire distribution route. A mixed integer linear programming (MILP) model is proposed to balance the cargo at each customer stop. The MILP model incorporates load distribution diagram constraints in order to determine the feasible domain for the location of the centre of gravity of the cargo arrangement, taking into account the regulatory requirements and the technical characteristics of the vehicle. Extensive computational experiments show that a MDLBRA can be used in practical contexts, as the MILP model was able to find a solution in less than ten minutes in 93% of the unbalanced test instances. © 2018 Elsevier Lt
A diffusion-based connectivity map of the GPi for optimised stereotactic targeting in DBS
No full textBackground: The GPi (globus pallidus internus) is an important target nucleus for Deep Brain Stimulation (DBS) in medically refractory movement disorders, in particular dystonia and Parkinson's disease. Beneficial clinical outcome critically depends on precise electrode localization. Recent evidence indicates that not only neurons, but also axonal fibre tracts contribute to promoting the clinical effect. Thus, stereotactic planning should, in the future, also take the individual course of fibre tracts into account. Objective: The aim of this project is to explore the GPi connectivity profile and provide a connectivity based parcellation of the GPi. Methods: Diffusion MRI sequences were performed in sixteen healthy, right-handed subjects. Connectivity-based parcellation of the GPi was performed applying two independent methods: 1) a hypothesis-driven, seed-to-target approach based on anatomic priors set as connectivity targets and 2) a purely data-driven approach based on k-means clustering of the GPi. Results: Applying the hypothesis-driven approach, we obtained five major parcellation clusters, displaying connectivity to the prefrontal cortex, the brainstem, the GPe (globus pallidus externus), the putamen and the thalamus. Parcellation clusters obtained by both methods were similar in their connectivity profile. With the data-driven approach, we obtained three major parcellation clusters. Inter individual variability was comparable with results obtained in thalamic parcellation. Conclusion: The three parcellation clusters obtained by the purely data-driven method might reflect GPi subdivision into a sensorimotor, associative and limbic portion. Clinical and physiological studies indicate greatest clinical DBS benefit for electrodes placed in the postero-ventro-lateral GPi, the region displaying connectivity to the thalamus in our study and generally attributed to the sensorimotor system. Clinical studies relating DBS electrode positions to our GPi connectivity map would be needed to complement our findings
A control strategy for a multi-terminal HVDC network integrating wind farms to the AC grid
No full textThis paper describes a novel control strategy of voltage source converters (VSC) in large-scale wind farm applications, in order to achieve a constant DC voltage for connecting to the long transmission system based on multi-terminal high voltage direct current (HVDC) technology. The control strategy is based on a multi-loop current and voltage control for tracking the predetermined value of the DC link voltage in the rectifier side station to achieve proper performance in the irregular circumstances of wind farm operation. In addition, the control strategy is able to transmit the maximum input power to the consumption side in different situations in grid side converter. Grid connection of the HVDC system is analysed under two conditions: balanced AC grid, and connection of unbalanced nonlinear load into the AC grid. The control strategy guarantees injection of minimum harmonic current components from the grid to the loads. MATLAB simulation results are presented to demonstrate the effectiveness of the proposed strategy for different types of variations in AC voltage amplitude and frequency of wind turbines output voltage