In this paper, we propose a novel approach to extract primary object segments in videos in the 'object proposal' domain. The extracted primary object regions are then used to build object models for optimized video segmentation. The proposed approach has several contributions: First, a novel layered Directed Acyclic Graph (DAG) based framework is presented for detection and segmentation of the primary object in video. We exploit the fact that, in general, objects are spatially cohesive and characterized by locally smooth motion trajectories, to extract the primary object from the set of all available proposals based on motion, appearance and predicted-shape similarity across frames. Second, the DAG is initialized with an enhanced object proposal set where motion based proposal predictions (from adjacent frames) are used to expand the set of object proposals for a particular frame. Last, the paper presents a motion scoring function for selection of object proposals that emphasizes high optical flow gradients at proposal boundaries to discriminate between moving objects and the background. The proposed approach is evaluated using several challenging benchmark videos and it outperforms both unsupervised and supervised state-of-the-art methods.

Purpose In vivo range verification in proton therapy is a critical step to help minimize range and dose uncertainty. We propose to employ a time reversal (TR)‐based approach using proton‐induced acoustics (protoacoustics) to reconstruct pressure/dose distribution in heterogeneous tissues. Methods The dose distribution of mono‐energetic proton pencil beam in a CT‐based patient phantom was calculated by Monte Carlo simulation. K‐wave toolbox was used to investigate protoacoustic pressurization, propagation and reconstruction in 2D. To address the tissue heterogeneity effect, a number of physical parameters, including mass density (ρ), speed of sound (c), volumetric thermal expansion coefficient (αV), isobaric specific heat capacity (Cp) and attenuation power law prefactor (α0), were empirically converted from CT number. The performance was evaluated using two figures of merit: mean square error (MSE) of pressure profiles and Bragg peak localization error (ΔBP). The impact of six parameters of the TR inversion was examined, including number of sensors, sampling duration, sampling timestep, spill time, noise level and number of iterations. Results The quantitative accuracy of TR reconstruction and its dependency on the selected parameters is presented. Under optimum conditions, the positioning accuracy of the Bragg peak can be controlled below 1 mm. For instance, MSE is 0.0123 and ΔBP is 0.59 mm under the following conditions (32 sensors, sampling duration: 600 µs, sampling timestep: 40 ns, spill time: 1 µs, no noise). Conclusions The feasibility of TR‐based protoacoustic reconstruction in 2D for proton range verification was first demonstrated. The approach is not only applicable to pencil beam, but also has potential to be extended to passive scattering systems.

No abstract

No abstract

No abstract