VIEWS: 7 PAGES: 31 POSTED ON: 8/1/2011
Real-Time Prediction and Control Yusheng Feng, Ph.D Computational Bioengineering and Nanomechanics Lab SiViRT Computation Center External Advisory Board Review Meeting March 5th, 2010 Real-Time Defined Real time can refer to events simulated by a computer at the same speed that they would occur in real life. http://www.webopedia.com real-time computing (RTC), or "reactive computing", is the study of hard ware and software systems that are subject to a “real- time constraint” – i.e. operational deadlines from event to system response. http://www.wikipedia.com Project Team Computational Cancer Research and Real-time Surgical Control Yusheng Feng, Ph.D Gerald Dodd III, M.D UTSA (Team Lead) UC Denver Computational Neuroscience and Real-time Cellular Control Allan Coop, Ph.D Hugo Cornelis, Ph.D UTHSCSA UTHSCSA Unmanned Aerial Vehicle (UVA) and Real-Time Flight C-J Qian, Ph.D Control UTSA OUTLINE Why we are here? Where we were? Where we are? Where we are going? SiViRT = Si + Vi + RT Three Components: • Imaging • Real-Time Computing • Uncertainty Quantification Computational Cancer Research and Real Time Surgical Control Cancer Treatment Surgery Radiation Chemo-, Gene-, Diagnosis Immuno- Prognosis Viro And Thermo- And Therapies Treatment Outcome Planning And Prediction Image- Guided Intervention 7 Basic Issues Medical Imaging and data processing Mathematical Modeling of Laser-Tissue Interaction Bioheat transfer in tissue and cellular response Mesh generation and numerical simulation Cell Damage and tissue Characterization Inverse problem and parallel computing Treatment planning and outcome prediction Real-time surgical monitoring and control Application of nano-technology with laser therapy Model validation: in vitro and in vivo experimentation MRTI-Guided Laser Surgery RT Control Optimal MRI Control Scan hp-adaptive 980-nm Diode Laser finite element patch Data Imaging process (co- Acquisition registration, segmentation, lofting) and mesh generation. Mathematical Models Bio-Heat Transfer Model Laser Tissue Interaction Model Nanoshell Model and Effective Tissue Properties Cell Damage Model Heat Shock Protein Models Laser-Tissue Interaction Ablation: With enough photon energy, laser can be used to ionize molecules in biological tissue. There exists a threshold at which the absorbed energy is high enough to cause decomposition of tissue. Diffusion Theory Monte Carlo Method – Beer-Lambert’s law – Easy to implement – Inexpensive – Very expensive – Less accurate – Not suitable for heterogeneous media Cell Damage and Heat Shock Protein Damage can be determined by Arrhenius law Ea T t ln Co C A e dt 0 Heat shock proteins (HSP) assist in refolding and repairing other Green Fluorescence HSP27 Red Fluorescence HSP70 denatured proteins, and facilitating synthesis of new proteins in response to damage. HSP Empirical Model: >1). H(t,T) Aexp( t t ) Major Problems with Arrhenius Law for Tissue Damage Experimental data May need two sets of parameters, Arrhenius model Ea / RT C(T,t) e At e Cannot predict “shoulder” effect, T=46oC Model parameters are extremely large and very Cv% sensitive to measurement data, It is not so clear what is the biophysical and physiological meaning of model parameters Heating Time (min) Results for Human Prostate PC3 Cells T=44oC T=50oC Experimental data E / RT T=56oC At e Arrhenius Model C(T,t) e ( H T S ) /T New Model* e C (T , t ) ( H T S ) /T 1 e * Feng, Oden, and Rylander, “A Two State model for cell damage: Theory and its validation in vitro” J. Biomech Eng., 2008 Biological Quantities of Interest How to measure effectiveness of Cancer Treatment? Cell Damage Fraction Index. How to quantify cancer recurrence? Heat Shock Protein Expression. Control parameter: Temperature. Objective of Optimization Nonlinear Transient Bio-Heat Transfer Equation Parameters Estimation Heterogeneous and nonlinear thermal properties: ki(x, t, T), wi(x, t, T) Heterogeneous tissue optical properties: a(x), s(x), … Laser parameters: , P, location, … Note: A domain was ki(x, t, T) wi(x, t, T) discretized into 3700 elements to characterize heterogeneity. Optimal Control Parallel Implementation One master processor One communication processor } Optimization processors: 60 Calibration Processors: 60 190 CPU Visualization Processors: 70 Simulation/real time ratio is 10:1 (e.g. 1 sec ~ 10 sec real time) Typical laser surgery time is 3 ~ 5 min. Processors numbers and allocation is dynamically controlled. Results Remarks Imaged guided real time predictive control Capability to characterize heterogeneous and nonlinear tissue properties on the fly Feasibility demonstrated in canine experiment Computational framework can be generalized to other applications Computational Neuroscience And Real-Time Cellular Control Goal: Developing neurological pathways of computational models and real systems CBI Simulator Framework CBI Simulator Framework (Cont’d) Software & Hardware Validation Experiment Design We have designed and implemented a RTXI plugin module that controls the magnitude of a single channel conductance. The conductance magnitude is read from a pre-generated data file. The standard model of conductance in an equivalent circuit is used to calculate the magnitude of current injected into a real cell (Iinj) based on the supplied conductance value (g), the recorded membrane potential (Vm) and channel current reversal potential (Er) of a real cell. UAV and Real-Time Control CJ Qian, Ph.D. UTSA Activities Two major research directions Research team on UAV: Hardware (undergraduate students), control algorithms (graduate students) Research team on real-time estimation and control over the network Outcomes Development of fast convergent estimators/controllers with applications to real-time estimation and control of PVTOL aircraft Fast convergence of our estimated states (dashed) to the real states(solid) Future Plan Unified Real-Time Computational Framework Integrated SW/HW Control System Thank You! Questions and Comments?
Pages to are hidden for
"Toward Real Time Optimal Control in Laser Surgery for Cancer "Please download to view full document