Showing papers by "Dorin Comaniciu published in 2013"

Method and System for Non-Invasive Functional Assessment of Coronary Artery Stenosis

Abstract: A method and system for non-invasive assessment of coronary artery stenosis is disclosed. Patient-specific anatomical measurements of the coronary arteries are extracted from medical image data of a patient acquired during rest state. Patient-specific rest state boundary conditions of a model of coronary circulation representing the coronary arteries are calculated based on the patient-specific anatomical measurements and non-invasive clinical measurements of the patient at rest. Patient-specific rest state boundary conditions of the model of coronary circulation representing the coronary arteries are calculated based on the patient-specific anatomical measurements and non-invasive clinical measurements of the patient at rest. Hyperemic blood flow and pressure across at least one stenosis region of the coronary arteries are simulated using the model of coronary circulation and the patient-specific hyperemic boundary conditions. Fractional flow reserve (FFR) is calculated for the at least one stenosis region based on the simulated hyperemic blood flow and pressure.

Non-Invasive Hemodynamic Assessment of Aortic Coarctation: Validation with In Vivo Measurements

Abstract: We propose a CFD-based approach for the non-invasive hemodynamic assessment of pre- and post-operative coarctation of aorta (CoA) patients. Under our approach, the pressure gradient across the coarctation is determined from computational modeling based on physiological principles, medical imaging data, and routine non-invasive clinical measurements. The main constituents of our approach are a reduced-order model for computing blood flow in patient-specific aortic geometries, a parameter estimation procedure for determining patient-specific boundary conditions and vessel wall parameters from non-invasive measurements, and a comprehensive pressure-drop formulation coupled with the overall reduced-order model. The proposed CFD-based algorithm is fully automatic, requiring no iterative tuning procedures for matching the computed results to observed patient data, and requires approximately 6-8 min of computation time on a standard personal computer (Intel Core2 Duo CPU, 3.06 GHz), thus making it feasible for use in a clinical setting. The initial validation studies for the pressure-drop computations have been performed on four patient datasets with native or recurrent coarctation, by comparing the results with the invasively measured peak pressure gradients recorded during routine cardiac catheterization procedure. The preliminary results are promising, with a mean absolute error of less than 2 mmHg in all the patients.

System and method for patient specific planning and guidance of ablative procedures for cardiac arrhythmias

Abstract: A method and system for patient-specific planning and guidance of an ablation procedure for cardiac arrhythmia is disclosed. A patient-specific anatomical heart model is generated based on pre-operative cardiac image data. The patient-specific anatomical heart model is registered to a coordinate system of intra-operative images acquired during the ablation procedure. One or more ablation site guidance maps are generated based on the registered patient-specific anatomical heart model and intra-operative patient-specific measurements acquired during the ablation procedure. The ablation site guidance maps may include myocardium diffusion and action potential duration maps. The ablation site guidance maps are generated using a computational model of cardiac electrophysiology which is personalized by fitting parameters of the cardiac electrophysiology model using the intra-operative patient-specific measurements. The ablation site guidance maps are displayed by a display device during the ablation procedure.

Method and System for Advanced Measurements Computation and Therapy Planning from Medical Data and Images Using a Multi-Physics Fluid-Solid Heart Model

Abstract: Method and system for computation of advanced heart measurements from medical images and data; and therapy planning using a patient-specific multi-physics fluid-solid heart model is disclosed. A patient-specific anatomical model of the left and right ventricles is generated from medical image patient data. A patient-specific computational heart model is generated based on the patient-specific anatomical model of the left and right ventricles and patient-specific clinical data. The computational model includes biomechanics, electrophysiology and hemodynamics. To generate the patient-specific computational heart model, initial patient-specific parameters of an electrophysiology model, initial patient-specific parameters of a biomechanics model, and initial patient-specific computational fluid dynamics (CFD) boundary conditions are marginally estimated. A coupled fluid-structure interaction (FSI) simulation is performed using the initial patient-specific parameters, and the initial patient-specific parameters are refined based on the coupled FSI simulation. The estimated model parameters then constitute new advanced measurements that can be used for decision making.

Method and system for patient specific planning of cardiac therapies on preoperative clinical data and medical images

Abstract: A method and system for patient-specific planning of cardiac therapy, such as cardiac resynchronization therapy (CRT), based on preoperative clinical data and medical images, such as ECG data, magnetic resonance imaging (MRI) data, and ultrasound data, is disclosed. A patient-specific anatomical model of the left and right ventricles is generated from medical image data of a patient. A patient-specific computational heart model, which comprises cardiac electrophysiology, biomechanics and hemodynamics, is generated based on the patient-specific anatomical model of the left and right ventricles and clinical data. Simulations of cardiac therapies, such as CRT at one or more anatomical locations are performed using the patient-specific computational heart model. Changes in clinical cardiac parameters are then computed from the patient-specific model, constituting predictors of therapy outcome useful for therapy planning and optimization.

Image-based Co-Registration of Angiography and Intravascular Ultrasound Images

Abstract: In image-guided cardiac interventions, X-ray imaging and intravascular ultrasound (IVUS) imaging are two often used modalities. Interventional X-ray images, including angiography and fluoroscopy, are used to assess the lumen of the coronary arteries and to monitor devices in real time. IVUS provides rich intravascular information, such as vessel wall composition, plaque, and stent expansions, but lacks spatial orientations. Since the two imaging modalities are complementary to each other, it is highly desirable to co-register the two modalities to provide a comprehensive picture of the coronaries for interventional cardiologists. In this paper, we present a solution for co-registering 2-D angiography and IVUS through image-based device tracking. The presented framework includes learning-based vessel detection and device detections, model-based tracking, and geodesic distance-based registration. The system first interactively detects the coronary branch under investigation in a reference angiography image. During the pullback of the IVUS transducers, the system acquires both ECG-triggered fluoroscopy and IVUS images, and automatically tracks the position of the medical devices in fluoroscopy. The localization of tracked IVUS transducers and guiding catheter tips is used to associate an IVUS imaging plane to a corresponding location on the vessel branch under investigation. The presented image-based solution can be conveniently integrated into existing cardiology workflow. The system is validated with a set of clinical cases, and achieves good accuracy and robustness.

Image-Based Computational Models for TAVI Planning: From CT Images to Implant Deployment

Abstract: Transcatheter aortic valve implantation (TAVI) is becoming the standard choice of care for non-operable patients suffering from severe aortic valve stenosis. As there is no direct view or access to the affected anatomy, accurate preoperative planning is crucial for a successful outcome. The most important decision during planning is selecting the proper implant type and size. Due to the wide variety in device sizes and types and non-circular annulus shapes, there is often no obvious choice for the specific patient. Most clinicians base their final decision on their previous experience. As a first step towards a more predictive planning, we propose an integrated method to estimate the aortic apparatus from CT images and compute implant deployment. Aortic anatomy, which includes aortic root, leaflets and calcifications, is automatically extracted using robust modeling and machine learning algorithms. Then, the finite element method is employed to calculate the deployment of a TAVI implant inside the patient-specific aortic anatomy. The anatomical model was evaluated on 198 CT images, yielding an accuracy of 1.30±0.23 mm. In eleven subjects, pre- and post-TAVI CT images were available. Errors in predicted implant deployment were of 1.74±0.40 mm in average and 1.32 mm in the aortic valve annulus region, which is almost three times lower than the average gap of 3 mm between consecutive implant sizes. Our framework may thus constitute a surrogate tool for TAVI planning.

System and Method for Real-Time Ultrasound Guided Prostate Needle Biopsy Based on Biomechanical Model of the Prostate from Magnetic Resonance Imaging Data

Abstract: A method and system for real-time ultrasound guided prostate needle biopsy based on a biomechanical model of the prostate from 3D planning image data, such as magnetic resonance imaging (MRI) data, is disclosed. The prostate is segmented in the 3D ultrasound image. A reference patient-specific biomechanical model of the prostate extracted from planning image data is fused to a boundary of the segmented prostate in the 3D ultrasound image, resulting in a fused 3D biomechanical prostate model. In response to movement of an ultrasound probe to a new location, a current 2D ultrasound image is received. The fused 3D biomechanical prostate model is deformed based on the current 2D ultrasound image to match a current deformation of the prostate due to the movement of the ultrasound probe to the new location.

Method and system for hemodynamic assessment of aortic coarctation from medical image data

Abstract: A method and system for non-invasive hemodynamic assessment of aortic coarctation from medical image data, such as magnetic resonance imaging (MRI) data is disclosed. Patient-specific lumen anatomy of the aorta and supra-aortic arteries is estimated from medical image data of a patient, such as contrast enhanced MRI. Patient-specific aortic blood flow rates are estimated from the medical image data of the patient, such as velocity encoded phase-contrasted MRI cine images. Patient-specific inlet and outlet boundary conditions for a computational model of aortic blood flow are calculated based on the patient-specific lumen anatomy, the patient-specific aortic blood flow rates, and non-invasive clinical measurements of the patient. Aortic blood flow and pressure are computed over the patient-specific lumen anatomy using the computational model of aortic blood flow and the patient-specific inlet and outlet boundary conditions.

Needle enhancement in diagnostic ultrasound imaging

Abstract: A needle is enhanced in a medical diagnostic ultrasound image. The image intensities associated with a needle in an image are adaptively increased and/or enhanced by compounding from a plurality of ultrasound images. Filtering methods and probabilistic methods are used to locate possible needle locations. In one approach, possible needles are found in component frames that are acquired at the same time but at different beam orientations. The possible needles are associated with each other across the component frames and false detections are removed based on the associations. In one embodiment of needle detection in an ultrasound component frame, lines are found first. The lines are then searched to find possible needle segments. In another embodiment, data from different times may be used to find needle motion and differences from a reference, providing the features in additional to features from a single component frame for needle detection.

Fast data-driven calibration of a cardiac electrophysiology model from images and ECG.

Abstract: Recent advances in computational electrophysiology (EP) models make them attractive for clinical use. We propose a novel data-driven approach to calibrate an EP model from standard 12-lead electrocardiograms (ECG), which are in contrast to invasive or dense body surface measurements widely available in clinical routine. With focus on cardiac depolarization, we first propose an efficient forward model of ECG by coupling a mono-domain, Lattice-Boltzmann model of cardiac EP to a boundary element formulation of body surface potentials. We then estimate a polynomial regression to predict myocardium, left ventricle and right ventricle endocardium electrical diffusion from QRS duration and ECG electrical axis. Training was performed on 4,200 ECG simulations, calculated in ≈3s each, using different diffusion parameters on 13 patient geometries. This allowed quantifying diffusion uncertainty for given ECG parameters due to the ill-posed nature of the ECG problem. We show that our method is able to predict myocardium diffusion within the uncertainty range, yielding a prediction error of less than 5ms for QRS duration and 2° for electrical axis. Prediction results compared favorably with those obtained with a standard optimization procedure, while being 60 times faster. Our data-driven model can thus constitute an efficient preliminary step prior to more refined EP personalization.

Learning-Based Detection and Tracking in Medical Imaging: A Probabilistic Approach

Abstract: Medical image processing tools are playing an increasingly important role in assisting the clinicians in diagnosis, therapy planning and image-guided interventions. Accurate, robust and fast tracking of deformable anatomical objects, such as the heart, is a crucial task in medical image analysis. One of the main challenges is to maintain an anatomically consistent representation of target appearance that is robust enough to cope with inherent changes due to target movement, imaging device movement, varying imaging conditions, and is consistent with the domain expert clinical knowledge. To address these challenges, this chapter presents a probabilistic framework that relies on anatomically indexed component-based object models which integrate several sources of information to determine the temporal trajectory of the deformable target. Large annotated imaging databases are exploited to encode the domain knowledge in shape models and motion models and to learn discriminative image classifiers for the target appearance. The framework robustly fuses the prior information with traditional tracking approaches based on template matching and registration. We demonstrate various medical image analysis applications with focus on cardiology such as 2D auto left heart, catheter detection and tracking, 3D cardiac chambers surface tracking, and 4D complex cardiac structure tracking, in multiple modalities including Ultrasound (US), cardiac Computed Tomography (CT), Magnetic Resonance Imaging (MRI), and X-ray fluoroscopy.

Method and System for Fast Patient-Specific Cardiac Electrophysiology Simulations for Therapy Planning and Guidance

Abstract: A method and system for patient-specific cardiac electrophysiology is disclosed. Particularly, a patient-specific anatomical model of a heart is generated from medical image data of a patient, a level-set representation of the patient-specific anatomical model is generated of the heart on a Cartesian grid; and a transmembrane action potential at each node of the level-set representation of the of the patient-specific anatomical model of the heart is computed on a Cartesian grid.

Computation of Hemodynamic Quantities from Angiographic Data

Abstract: Methods for computing hemodynamic quantities include: (a) acquiring angiography data from a patient; (b) calculating a flow and/or calculating a change in pressure in a blood vessel of the patient based on the angiography data; and (c) computing the hemodynamic quantity based on the flow and/or the change in pressure. Systems for computing hemodynamic quantities and computer readable storage media are described.

Patient Specific Planning and Simulation of Ablative Procedures

Abstract: Patient specific temperature distribution in organs, due to an ablative device, is simulated. The effects of ablation are modeled. The modeling is patient specific. The vessel structure for a given patient, segmented from medical images, is accounted for as a heat sink in the model of biological heat transfer. A temperature map is generated to show the effects of ablation in a pre-operative analysis. Temperature maps resulting from different ablation currents and ablation device positions may be used to determine a more optimal location of the ablative device for a given patient. Other models may be included, such as accounting for the tissue damage during the ablation.

Mitral valve detection for transthoracic echocardiography

Abstract: A mitral valve is detected in transthoracic echocardiography. The ultrasound transducer is positioned against the chest of the patient rather than being inserted within the patient. While data acquired from such scanning may be noisier or have less resolution, the mitral valve may still be automatically detected. Using both B-mode data representing tissue as well as flow data representing the regurgitant jet, the mitral valve may be detected automatically with a machine-learnt classifier. A series of classifiers may be used, such as determining a position and orientation of a valve region with one classifier, determining a regurgitant orifice with another classifier, and locating mitral valve anatomy with a third classifier. One or more features for some of the classifiers may be calculated based on the orientation of the valve region.

Robust multi-object tracking using sparse appearance representation and online sparse appearance dictionary update

Abstract: A computer-implemented method for tracking one or more objects in a sequence of images includes generating a dictionary based on object locations in a first image included in the sequence of images. One or more object landmark candidates are identified in the sequence of images and a plurality of tracking hypothesis for the object landmark candidates are generated. A first tracking hypothesis is selected from the plurality of tracking hypothesis based on the dictionary.

Method and System for Model Based Fusion on Pre-Operative Computed Tomography and Intra-Operative Fluoroscopy Using Transesophageal Echocardiography

Abstract: A method and system for model-based fusion of pre-operative image data and intra-operative fluoroscopic images is disclosed. A fluoroscopic image and an ultrasound image are received. The ultrasound image is mapped to a 3D coordinate system of a fluoroscopic image acquisition device used to acquire the fluoroscopic image. Contours of an anatomical structure are detected in the ultrasound image, and a transformation is calculated between the ultrasound image and a pre-operative CT image based on the contours and a patient-specific physiological model extracted from the pre-operative CT image. A final mapping is determined between the CT image and the fluoroscopic image based on the transformation between the ultrasound image and physiological model and the mapping of the ultrasound image to the 3D coordinate system of the fluoroscopic image acquisition device. The CT image or the physiological model can then be projected into the fluoroscopic image.

Lattice Boltzmann method for fast patient-specific simulation of liver tumor ablation from CT images.

Abstract: Radio-frequency ablation (RFA), the most widely used minimally invasive ablative therapy of liver cancer, is challenged by a lack of patient-specific planning In particular, the presence of blood vessels and time-varying thermal diffusivity makes the prediction of the extent of the ablated tissue difficult This may result in incomplete treatments and increased risk of recurrence We propose a new model of the physical mechanisms involved in RFA of abdominal tumors based on Lattice Boltzmann Method to predict the extent of ablation given the probe location and the biological parameters Our method relies on patient images, from which level set representations of liver geometry, tumor shape and vessels are extracted Then a computational model of heat diffusion, cellular necrosis and blood flow through vessels and liver is solved to estimate the extent of ablated tissue After quantitative verifications against an analytical solution, we apply our framework to 5 patients datasets which include pre- and post-operative CT images, yielding promising correlation between predicted and actual ablation extent (mean point to mesh errors of 87 mm) Implemented on graphics processing units, our method may enable RFA planning in clinical settings as it leads to near real-time computation: 1 minute of ablation is simulated in 114 minutes, which is almost 60 × faster than standard finite element method

Method and system for pericardium based model fusion of pre-operative and intra-operative image data for cardiac interventions

Abstract: A method and system for model based fusion pre-operative image data, such as computed tomography (CT), and intra-operative C-arm CT is disclosed. A first pericardium model is segmented in the pre-operative image data and a second pericardium model is segmented in a C-arm CT volume. A deformation field is estimated between the first pericardium model and the second pericardium model. A model of a target cardiac structure, such as a heart chamber model or an aorta model, extracted from the pre-operative image data is fused with the C-arm CT volume based on the estimated deformation field between the first pericardium model and the second pericardium model. An intelligent weighted average may be used improve the model based fusion results using models of the target cardiac structure extracted from pre-operative image data of patients other than a current patient.

From medical images to fast computational models of heart electromechanics: an integrated framework towards clinical use

Abstract: With the recent advances in computational power, realistic modeling of heart function within a clinical environment has come into reach Yet, current modeling frameworks either lack overall completeness or computational performance, and their integration with clinical imaging and data is still tedious In this paper, we propose an integrated framework to model heart electromechanics from clinical and imaging data, which is fast enough to be embedded in a clinical setting More precisely, we introduce data-driven techniques for cardiac anatomy estimation and couple them with an efficient GPU (graphics processing unit) implementation of the orthotropic Holzapfel-Ogden model of myocardium tissue, a GPU implementation of a mono-domain electrophysiology model based on the Lattice-Boltzmann method, and a novel method to correctly capture motion during isovolumetric phases Benchmark experiments conducted on patient data showed that the computation of a whole heart cycle including electrophysiology and biomechanics with mesh resolutions of around 70k elements takes on average 1min 10s on a standard desktop machine (Intel Xeon 24GHz, NVIDIA GeForce GTX 580) We were able to compute electrophysiology up to 405× faster and biomechanics up to 152× faster than with prior CPU-based approaches, which breaks ground towards model-based therapy planning

Graphics processing unit accelerated one-dimensional blood flow computation in the human arterial tree

Abstract: SUMMARY One-dimensional blood flow models have been used extensively for computing pressure and flow waveforms in the human arterial circulation. We propose an improved numerical implementation based on a graphics processing unit (GPU) for the acceleration of the execution time of one-dimensional model. A novel parallel hybrid CPU–GPU algorithm with compact copy operations (PHCGCC) and a parallel GPU only (PGO) algorithm are developed, which are compared against previously introduced PHCG versions, a single-threaded CPU only algorithm and a multi-threaded CPU only algorithm. Different second-order numerical schemes (Lax–Wendroff and Taylor series) are evaluated for the numerical solution of one-dimensional model, and the computational setups include physiologically motivated non-periodic (Windkessel) and periodic boundary conditions (BC) (structured tree) and elastic and viscoelastic wall laws. Both the PHCGCC and the PGO implementations improved the execution time significantly. The speed-up values over the single-threaded CPU only implementation range from 5.26 to 8.10 × , whereas the speed-up values over the multi-threaded CPU only implementation range from 1.84 to 4.02 × . The PHCGCC algorithm performs best for an elastic wall law with non-periodic BC and for viscoelastic wall laws, whereas the PGO algorithm performs best for an elastic wall law with periodic BC. Copyright © 2013 John Wiley & Sons, Ltd.

Advanced intervention planning for Transcatheter Aortic Valve Implantations (TAVI) from CT using volumetric models

Abstract: Aortic valve stenosis is a serious heart disease affecting a large group of elderly people. Recently minimal invasive procedures, such as the Transcatheter Aortic Valve Implantation (TAVI), are beginning to substitute conventional surgical techniques. Current methods [1] can extract basic biomarkers for TAVI such as optimal C-arm angulations, area and diameter measurements. However as the most prevalent TAVI complications (stroke and paravalvular leakages) are correlated with calcium and leaflet interactions within the valve a more advanced solution is needed. We propose a fully integrated system to extract automatically the patient specific model of the aortic valve including the volumetric model of the aortic valve leaflets and calcium from high resolution single phase CT. Based on the volumetric model advanced clinical parameters can be derived and used for e.g. patient selection, paravalvular leakage prediction and patient stroke risk assessment. We employ robust machine learning algorithms to estimate the valve model parameters. A multi-class classification method is introduced to label regions of calcium, leaflet and blood pool within the aortic valve and extract volumetric models of the aortic valve leaflets. Extensive quantitative and qualitative experiments on 198 volumetric data sets demonstrate an accurate DICE similarity score, i.e. 0.7 for the aortic valve leaflets and 0.86 for calcium tissue. Within 6 seconds a complete patient-specific model of the aortic valve can be estimated.

Providing real-time marker detection for a stent in medical imaging

Abstract: Real-time marker detection in medical imaging of a stent may be provided. A plurality of frames of image data may be obtained. A plurality of candidate markers for the stent may be determined in the plurality of frames of image data. One or more markers from the plurality of candidate markers may be detected. The detecting may be based on automatic initialization using a subset of frames of image data from the plurality of frames of image data. The detecting may be performed in real-time with the obtaining.

Model-Based Estimation of 4D Relative Pressure Map from 4D Flow MR Images

Abstract: We propose a new framework for 4D relative pressure map computations from 4D flow MRI that uses enhanced geometric models for the blood vessels and flow-aware surface and volumetric tags. The enhanced geometric modeling provides better accuracy compared to a simple voxelized mask, while tagging of inlets and outlets allows imposing physiologically meaningful boundary conditions, contributing to more accurate pressure computations. An integrated software suite for semi-automatic processing of 4D flow MR images, preparation and computation of the flow parameters is presented. This enables a fast and intuitive workflow, with accurate final results, ready in minutes.

Data-driven reduction of a cardiac myofilament model

Abstract: This manuscript presents a novel, data-driven approach to reduce a detailed cellular model of cardiac myofilament (MF) for efficient and accurate cellular simulations towards cell-to-organ computation. Based on 700 different sarcomere dynamics calculated using Rice model, we show through manifold learning that sarcomere force (SF) dynamics lays surprisingly in a linear manifold despite the non-linear equations of the MF model. Then, we learn a multivariate adaptive regression spline (MARS) model to predict SF from the Rice model parameters and sarcomere length dynamics. Evaluation on 300 testing data showed a prediction error of less than 0.4 nN/mm2 in terms of maximum force amplitude and 0.87 ms in terms of time to force peak, which is comparable to the differences observed with experimental data. Moreover, MARS provided insights on the driving parameters of the model, mainly MF geometry and cell mechanical passive properties. Thus, our approach may not only constitute a fast and accurate alternative to the original Rice model but also provide insights on parameter sensitivity.

Estimation of patient-specific material properties of the mitral valve using 4D Transesophageal Echocardiography

Abstract: 4D Transesophageal Echocardiography (TEE) is a newly developed tool to visualize the morphology and dynamics of the mitral valve for diagnosis and treatment planning Quantitative patient-specific modeling of the mitral valve is demanded since it allows for the reliable predictive simulation of medical intervention State-of-the-art image-based and biomechanical models with generic material parameters have limited predictive power as they are only partially fitted to patient-data As a step closer to a fully personalized model, an estimation algorithm is presented in this paper The method combines image-derived mitral valve dynamics with a biomechanical model to estimate regional patient-specific material parameters in-vivo In particular, the extended Kalman filter (EKF) is adopted in a way that it becomes flexible to integrate any biomechanical model and more parameters of interest in the estimation The algorithm was verified on synthetic data with known Young's modulus and shear modulus, yielding less than 5% error The algorithm was also evaluated on 4D TEE images of five patients Estimated Young's moduli agree with the clinical observation that material parameters vary regionally and among the population The estimated material parameters can be used in patient-specific modeling as well as the detection and evaluation of diseased areas

System and methods for integrated and predictive analysis of molecular, imaging, and clinical data for patient-specific management of diseases

Abstract: A system operating in a plurality of modes to provide an integrated analysis of molecular data, imaging data, and clinical data associated with a patient includes a multi-scale model, a molecular model, and a linking component. The multi-scale model is configured to generate one or more estimated multi-scale parameters based on the clinical data and the imaging data when the system operates in a first mode, and generate a model of organ functionality based on one or more inferred multi-scale parameters when the system operates in a second mode. The molecular model is configured to generate one or more first molecular findings based on a molecular network analysis of the molecular data, wherein the molecular model is constrained by the estimated parameters when the system operates in the first mode. The linking component, which is operably coupled to the multi-scale model and the molecular model, is configured to transfer the estimated multi-scale parameters from the multi-scale model to the molecular model when the system operates in the first mode, and generate, using a machine learning process, the inferred multi-scale parameters based on the molecular findings when the system operates in the second mode.