Quang Trần - Academia.edu (original) (raw)

Papers by Quang Trần

arXiv (Cornell University), Oct 4, 2021

Deformable registration is a crucial step in many medical procedures such as image-guided surgery... more Deformable registration is a crucial step in many medical procedures such as image-guided surgery and radiation therapy. Most recent learning-based methods focus on improving the accuracy by optimizing the nonlinear spatial correspondence between the input images. Therefore, these methods are computationally expensive and require modern graphic cards for real-time deployment. In this paper, we introduce a new Light-weight Deformable Registration network that significantly reduces the computational cost while achieving competitive accuracy. In particular, we propose a new adversarial learning with distilling knowledge algorithm that successfully leverages meaningful information from the effective but expensive teacher network to the student network. We design the student network such as it is light-weight and well suitable for deployment on a typical CPU. The extensively experimental results on different public datasets show that our proposed method achieves state-of-the-art accuracy while significantly faster than recent methods. We further show that the use of our adversarial learning algorithm is essential for a time-efficiency deformable registration method. Finally, our source code and trained models are available at .

medRxiv (Cold Spring Harbor Laboratory), May 10, 2022

Mathematics in Industry, 2008

Numerical Mathematics and Advanced Applications

Hyperbolic Problems: Theory, Numerics, Applications, 2008

2010 IEEE RIVF International Conference on Computing & Communication Technologies, Research, Innovation, and Vision for the Future (RIVF), 2010

 Apply deep learning to improve image ads' quality. Graduate Research Assistant, Georgia Tech Au... more  Apply deep learning to improve image ads' quality. Graduate Research Assistant, Georgia Tech August 2012-Present  Modeling the timing and physical state of human activity: implemented a unified system for predicting human movement by combining low level dynamic of the movements and high level temporal structure of the activity.

Multiscale Modeling & Simulation, 2010

PloS one, 2018

To examine the effects of socioecological factors on multiple communicable diseases across Vietna... more To examine the effects of socioecological factors on multiple communicable diseases across Vietnam. We used the Moran's I tests to evaluate spatial clusters of diseases and applied multilevel negative binomial regression models using the Bayesian framework to analyse the association between socioecological factors and the diseases queried by oral, airborne, vector-borne, and animal transmission diseases. The study found that oral-transmission diseases were spatially distributed across the country; whereas, the airborne-transmission diseases were more clustered in the Northwest and vector-borne transmission diseases were more clustered in the South. Most of diseases were sensitive with climatic factors. For instance, a 1°C increase in average temperature is significantly associated with 0.4% (95CI, 0.3-0.5), 2.5% (95%CI, 1.4-3.6), 0.9% (95%CI, 0.6-1.4), 1.1% (95%CI), 5% (95%CI, 3-.7.4), 0.4% (95%CI, 0.2-0.7), and 2% (95%CI, 1.5-2.8) increase in risk of diarrhoea, shigellosis, mum...

ArXiv, 2020

Different approaches have been proposed to Visual Question Answering (VQA). However, few works ar... more Different approaches have been proposed to Visual Question Answering (VQA). However, few works are aware of the behaviors of varying joint modality methods over question type prior knowledge extracted from data in constraining answer search space, of which information gives a reliable cue to reason about answers for questions asked in input images. In this paper, we propose a novel VQA model that utilizes the question-type prior information to improve VQA by leveraging the multiple interactions between different joint modality methods based on their behaviors in answering questions from different types. The solid experiments on two benchmark datasets, i.e., VQA 2.0 and TDIUC, indicate that the proposed method yields the best performance with the most competitive approaches.

PAMM, 2007

We are concerned with the numerical simulation of two‐phase flows representing oil transportation... more We are concerned with the numerical simulation of two‐phase flows representing oil transportation along a 1‐D pipeline. More specifically, we wish to show that, for this kind of problems, it is highly recommended to make use of the Lagrange‐Projection formalism. Indeed, this approach naturally splits the effects of fast (acoustic) and slow (kinematic) waves, thus enabling us to design numerical schemes with many desirable properties: explicit with respect to slow waves and implicit with respect to fast waves. The design of the overall explicit‐implicit Lagrange‐Projection method involves many interesting features, among which the positivity of the partial densities is of utmost interest to us. In this paper, the emphasis will be laid on the motivation and the general philosophy of the Lagrange‐Projection method we propose. Extensive numerical results for realistic test cases are then presented in order to illustrate the capabilities of the method

20th AIAA International Communication Satellite Systems Conference and Exhibit, 2002

This paper presents the implementation and initial test results of an Orthogonal Frequency Divisi... more This paper presents the implementation and initial test results of an Orthogonal Frequency Division Multiplexing (OFDM) digital modem (modulator and demodulator) with an aggregate information throughput of 622 megabits per second (Mbps). The OFDM waveform is constructed by dividing an incoming data stream into four channels, each channel using either a 16-ary Quadrature Amplitude Modulation (16QAM) scheme or an 8-Phase Shift Keying (8PSK) scheme. The generation and detection of the composite waveform are performed using Discrete Fourier Transform (DFT) and polyphase filtering, to digitally stack and band-limit the individual carriers respectively. The four-channel OFDM approach enables the implementation of a modem that can be both power and bandwidth efficient, with sufficient parallelism to meet higher data rate goals. As a result, the OFDM modem requires only a 240 MHz bandwidth to transmit 622 Mbps. Hardware and simulation results in the form of spectrum diagrams and bit-error-rate (BER) curves are also presented in this paper.

arXiv (Cornell University), Oct 4, 2021

Deformable registration is a crucial step in many medical procedures such as image-guided surgery... more Deformable registration is a crucial step in many medical procedures such as image-guided surgery and radiation therapy. Most recent learning-based methods focus on improving the accuracy by optimizing the nonlinear spatial correspondence between the input images. Therefore, these methods are computationally expensive and require modern graphic cards for real-time deployment. In this paper, we introduce a new Light-weight Deformable Registration network that significantly reduces the computational cost while achieving competitive accuracy. In particular, we propose a new adversarial learning with distilling knowledge algorithm that successfully leverages meaningful information from the effective but expensive teacher network to the student network. We design the student network such as it is light-weight and well suitable for deployment on a typical CPU. The extensively experimental results on different public datasets show that our proposed method achieves state-of-the-art accuracy while significantly faster than recent methods. We further show that the use of our adversarial learning algorithm is essential for a time-efficiency deformable registration method. Finally, our source code and trained models are available at .

medRxiv (Cold Spring Harbor Laboratory), May 10, 2022

Mathematics in Industry, 2008

Numerical Mathematics and Advanced Applications

Hyperbolic Problems: Theory, Numerics, Applications, 2008

2010 IEEE RIVF International Conference on Computing & Communication Technologies, Research, Innovation, and Vision for the Future (RIVF), 2010

 Apply deep learning to improve image ads' quality. Graduate Research Assistant, Georgia Tech Au... more  Apply deep learning to improve image ads' quality. Graduate Research Assistant, Georgia Tech August 2012-Present  Modeling the timing and physical state of human activity: implemented a unified system for predicting human movement by combining low level dynamic of the movements and high level temporal structure of the activity.

Multiscale Modeling & Simulation, 2010

PloS one, 2018

To examine the effects of socioecological factors on multiple communicable diseases across Vietna... more To examine the effects of socioecological factors on multiple communicable diseases across Vietnam. We used the Moran's I tests to evaluate spatial clusters of diseases and applied multilevel negative binomial regression models using the Bayesian framework to analyse the association between socioecological factors and the diseases queried by oral, airborne, vector-borne, and animal transmission diseases. The study found that oral-transmission diseases were spatially distributed across the country; whereas, the airborne-transmission diseases were more clustered in the Northwest and vector-borne transmission diseases were more clustered in the South. Most of diseases were sensitive with climatic factors. For instance, a 1°C increase in average temperature is significantly associated with 0.4% (95CI, 0.3-0.5), 2.5% (95%CI, 1.4-3.6), 0.9% (95%CI, 0.6-1.4), 1.1% (95%CI), 5% (95%CI, 3-.7.4), 0.4% (95%CI, 0.2-0.7), and 2% (95%CI, 1.5-2.8) increase in risk of diarrhoea, shigellosis, mum...

ArXiv, 2020

Different approaches have been proposed to Visual Question Answering (VQA). However, few works ar... more Different approaches have been proposed to Visual Question Answering (VQA). However, few works are aware of the behaviors of varying joint modality methods over question type prior knowledge extracted from data in constraining answer search space, of which information gives a reliable cue to reason about answers for questions asked in input images. In this paper, we propose a novel VQA model that utilizes the question-type prior information to improve VQA by leveraging the multiple interactions between different joint modality methods based on their behaviors in answering questions from different types. The solid experiments on two benchmark datasets, i.e., VQA 2.0 and TDIUC, indicate that the proposed method yields the best performance with the most competitive approaches.

PAMM, 2007

We are concerned with the numerical simulation of two‐phase flows representing oil transportation... more We are concerned with the numerical simulation of two‐phase flows representing oil transportation along a 1‐D pipeline. More specifically, we wish to show that, for this kind of problems, it is highly recommended to make use of the Lagrange‐Projection formalism. Indeed, this approach naturally splits the effects of fast (acoustic) and slow (kinematic) waves, thus enabling us to design numerical schemes with many desirable properties: explicit with respect to slow waves and implicit with respect to fast waves. The design of the overall explicit‐implicit Lagrange‐Projection method involves many interesting features, among which the positivity of the partial densities is of utmost interest to us. In this paper, the emphasis will be laid on the motivation and the general philosophy of the Lagrange‐Projection method we propose. Extensive numerical results for realistic test cases are then presented in order to illustrate the capabilities of the method

20th AIAA International Communication Satellite Systems Conference and Exhibit, 2002

This paper presents the implementation and initial test results of an Orthogonal Frequency Divisi... more This paper presents the implementation and initial test results of an Orthogonal Frequency Division Multiplexing (OFDM) digital modem (modulator and demodulator) with an aggregate information throughput of 622 megabits per second (Mbps). The OFDM waveform is constructed by dividing an incoming data stream into four channels, each channel using either a 16-ary Quadrature Amplitude Modulation (16QAM) scheme or an 8-Phase Shift Keying (8PSK) scheme. The generation and detection of the composite waveform are performed using Discrete Fourier Transform (DFT) and polyphase filtering, to digitally stack and band-limit the individual carriers respectively. The four-channel OFDM approach enables the implementation of a modem that can be both power and bandwidth efficient, with sufficient parallelism to meet higher data rate goals. As a result, the OFDM modem requires only a 240 MHz bandwidth to transmit 622 Mbps. Hardware and simulation results in the form of spectrum diagrams and bit-error-rate (BER) curves are also presented in this paper.