Computational Surgery and Dual Training

This paper introduces the fundamental concepts of computational surgery—Garbey et al. [Computational surgery and dual training, Springer, XVI, 315pp (Hardcover, ISBN: 978-1-4419-1122-3, 2009), 2010]—and proposes a road map for progress in this new multidisciplinary field of applied investigation. Recognizing this introduction will serve as common ground for discussion for both communities, surgeons and computational scientists, the scope of the presentation is broad rather than deep. Indeed, the field of computational surgery is sufficiently young that even the definition of computational surgery is still in the making. In this introduction, we propose multiple areas of investigation where the intersection of surgery and computational sciences is clearly in practice at the present time, though surprisingly unrecognized to date. We present examples of these intersections and demonstrate the usefulness and novelty of computational surgery as a new field of research. While some of the elements we present may be considered as basic for a specialized investigator, the simplicity of the presentation is intended as a proof of principle that basic concepts in computational sciences are of core value in solving many existing problems in clinical surgery; we also hope this initial evaluation will highlight potential obstacles and challenges. As the digital revolution transforms the working environment of the surgeon, close collaboration between surgeons and computational scientists is not only unavoidable but also essential to harness the capabilities of both fields to optimize the surgical care. We believe that this new collaboration will allow the community not only to develop the predictive models for the outcomes of surgery but also to enhance the process of surgery—from procedural planning, to execution of procedures and technology interfaces, to assessment of the healing process—investigations that will potentially provide great impact on patient care that far beyond the operating room.

Plato’s CAVE™ (Computer Augmented Virtual Environment) is a presurgical planning, multidimensional “situation clinical platform” designed, developed, and introduced to clinical practice by the Department of Radiation Oncology at Houston Methodist Hospital, located in Houston’s Texas Medical Center. At approximately 500 square feet, Plato’s CAVE was specifically designed to permit a team of physicians to review all available diagnostic images of the patient. The initial clinical focus was on interventions within the domain of surgical oncology/radiation oncology including radiation therapy, reconstructive surgery, and organ transplantation. This advanced clinical visualization process, supported by a novel and creative assemblage of FDA-approved, commercially available diagnostic imaging components, is available for all relevant patient care services within The Methodist Hospital System.

Technological advances in radiation oncology have led to clinical implementation of novel treatment modality. Stereotactic body radiotherapy (SBRT)/stereotactic ablative body radiotherapy (SABR) is an emerging treatment paradigm as a result of image-guidance technology and more sophisticated computational treatment planning system. SBRT/SABR, an example of computational radiosurgery, is a continuum of advances in computational surgery. The role of SBRT/SABR is most important in the management of lung cancer for early primary lung cancer and in oligometastatic lung disease. SBRT/SABR combines the challenges of patient/tumor/normal tissues motion with that of meeting the stringent dosimetric requirements of stereotactic radiosurgery (SRS). Target delineation, image guidance, patient immobilization, computer-assisted treatment planning, and delivery are essential in the safe and successful practice of SBRT/SABR. Radio-biologic rationale, technical and clinical aspects of SBRT/SABR in the treatment of both primary and metastatic lung cancer as well as the future challenges will be addressed.

We developed a portable software that allows users to conveniently determine the Cobb angle, the rigidity, and shift of the spinal cord based on the radiographs of scoliosis patients. The software is developed on the MATLAB platform, small in size and easy to be installed and used without formal training or special skills. The radiographs can be uploaded from a data file of a radiographic equipment. The operator uses a pointing device to draw two lines that go through the upper and lower endplates of the selected vertebrae. The software then automatically calculates the Cobb angle and the spinal cord shift. From the bending posture radiographs the system calculates the spinal rigidity. A comparative study on 20 cases using our software, a commercially available software called eFilm and by hand was conducted. The results showed that although the average values obtained from all three methods were not statistically different, our software yielded more reliable results and the measurement time was shorter. Our software therefore is a useful tool to assist the treatment of scoliosis in clinics.

Breast conserving therapy (BCT) is a less radical surgery consisting of the removal of the tumor (partial mastectomy) including a negative margin followed by radiotherapy. It provides the same incidence of local recurrence—reappearance of the cancer in the vicinity of a previously removed cancer—than a complete mastectomy (complete removal of the breast), with the advantage of offering faster recovery and better cosmetic outcome for patients. Nevertheless, many patients remain with some major cosmetic defects such as concave deformities, distortion of the nipple aerolar complex, and asymmetric changes.There are currently no procedures, other than surgical experience and judgment, allowing prediction on the impact of partial mastectomy on the contour and the deformity of the treated breast.The present work defines the basic principles of a virtual surgery toolbox that will allow to predict BCT intervention outcome.

Metabolomics is an exponentially growing field of “omics” research concerned with the high throughout comparison, identification and quantification of large numbers of metabolites in biological system. This emergent science of metabolomics brings increasing promise to identify biomarker diseases that integrate biochemical changes in disease and predict human reaction to treatment. In this context, the 2D High Resolution Magic Angle Spinning (HR-MAS) Nuclear Magnetic Resonance (NMR) spectroscopy has emerged as an ideal platform for studying metabolites of biopsies. In this study, we particularly focus on the 2D Heteronuclear Single Quantum Coherence (HSQC) NMR spectrum analysis. The metabolomic analysis requires comparison of metabolite profiles obtained from multiple replicates of samples exposed to different experimental conditions. What adds difficulty to automating this analysis process is that each peak of a given metabolite (a set of peaks with specified locations) can be shifted slightly from one sample to the next. In this study, we propose a new framework to detect and align simultaneously peaks representing different metabolites within a biopsy for metabonomic analysis. The method was validated on synthetic and real HSQC spectra.

Computer-aided simulations have a profound impact on day-to-day operations in medical centers. They represent one of the building blocks towards personalized medicine, which allows the treatment of patients and diseases on an individual basis. The computation requirements of these simulations can be, however, tremendous, posing unique challenges to software and hardware infrastructure. In this chapter we discuss recent computer hardware developments and evaluate them based on two representative application to understand how the new hardware can be used to reduce the execution time of these applications.

The abilities to identify a target through imaging and to navigate through the blood vessels and interrogate the vessel wall are core technical competencies in cardiovascular medicine. The imaging modalities used in the cardiovascular space include ultrasound, intravascular imaging, real-time X-ray (fluoroscopy), magnetic resonance imaging (MRI), computed tomography (CT), and positron emission tomography (PET) scanning. When merged, these imaging components provide a 3-dimensional (3D) functional view of a vessel and the pathway for an interventional system to access and to target a lesion. In addition to this fused imaging, other components are integral to intravascular navigation, including spatial awareness of the imaging system that builds on the fused images and the ability to steer a catheter based on that knowledge. Hybrid procedural rooms where fluoroscopy, cross-sectional imaging, and post-processing systems are concomitantly available allow for integration of real-time anatomy with real-time images and previously acquired images. Once this fused imaging array is linked to a robotically driven catheter that can maintain stability, follow center line flow and articulate in a 3D space, one has achieved a hybrid integrated therapeutic system which is capable of complex interventions. At present, cardiac, neurovascular, and endovascular interventions can be markedly enhanced by such integrated hybrid imaging. The potential to extend the fundamentals of the hybrid systems into other medical fields is significant, as is the potential to build more rapid computational capabilities to take the systems to the next level and further minimize the human operator/device interface.

2D and 3D-based innovative methods for surgical planning and simulation systems in orthopedic surgery have emerged enabling the interactive or semiautomatic identification of the clinical landmarks (CL) on the patient individual virtual bone anatomy. They allow the determination of the optimal implant sizes and positioning according to the computed CL, the visualization of the virtual bone resections, and the simulation of the overall intervention prior to surgery. Such innovative methods allow designing personalized resection guides, which substitute the traditional jigs and avoid any other alignment instrument and even navigation support. The virtual palpation of CL, highly dependent upon examiner’s expertise, was proved to be time consuming and suffered from considerable inter-observer variability. In this contribute, we disclose a fully automatic algorithmic framework that processes the femur surface, integrating surface curvature analysis, quadric fitting, recursive clustering, and clinical knowledge, aiming at computing the main parameters femur CL, namely the femoral shaft (FDA), transepicondylar (TEA), anterior–posterior (WL), posterior condylar (PCL) axes. At highest surface resolutions, the relative median error in the direction of the FDA, AFA, PCL, WL, and TEA was less than 0.50°, 1.20°, 1.0°, 1.30°, 1.50°, respectively. As expected, at the lowest surface resolution, the repeatability decreased to 1.20°, 2.70°, 3.30°, 3.0°, 4.70°, respectively. The computed directions of the FDA, PCL, WL, and TA were in agreement (0.60°, 1.55°, 1.90°, 2.40°) with the corresponding reference parameters manually identified in the original CT images by medical experts and with literature. We summarize that: (a) the AFA can be robustly computed by a geometrical analysis of the posterior profiles of the two condyles, and it can be considered a useful alternative to the TEA; (b) higher surface resolutions lead to higher repeatability of all computed quantities; (c) the TEA is less repeatable of the other axes. In conclusion, the method does not require any manual initialization, it can be automatically applied to the left and right surfaces, it is independent of the coordinate system of the original CT datasets, it is independent of the scale of the surface, and the algorithms show high stability and reduced computational load.

Lung cancer is the most common cause of cancer-related deaths in the USA. We review the epidemiology of lung cancer, the multimodality treatment, and the surgical treatment of this disease. In the section covering the surgical treatment of lung cancer, we will discuss open lobectomy, video-assisted lobectomy, and the new approach of robot-assisted lobectomy for treatment of lung cancer.

In robot-assisted medical procedures, the control of interactions between tools and tissues has recently raised much interest. In this chapter, we give an original perspective to this subject. In Sect. 11.1, we discuss the pros and cons of force feedback for medical applications. We also highlight some original works and point out open problems that would deserve more interest. Then, in Sects. 11.2 and 11.3, we focus on two of our recent contributions to the field of force feedback teleoperation for medical procedures: the automatic detection of haptic events and the compensation of physiological motions.

Flexible endoscopic systems are widely used for intraluminal exams and for minimally invasive surgical interventions, but the manual use of these systems is problematic. Providing telemanipulation modes by robotizing flexible endoscopes could be an interesting solution. However, it appears that the accurate control of these systems is not possible through standard control scheme because of large backlash in the cable driving mechanism. In this paper we discuss several possible solutions to this control problem. Especially we show that it is possible to improve accuracy and bandwidth by using an exteroceptive sensor coupled with software backlash compensation. Laboratory results as well as tests on animal models show the possibilities of these approaches.

The existence of residual motion with passive cardiac stabilizers hinders the development of minimally invasive beating heart procedures such as Coronary Artery Bypass Grafting (CABG). This issue can be overcome thanks to active stabilization. It consists in actuating the stabilizer in order to compensate in real time for residual heart motion. This approach has been investigated with the design of two different devices: the Cardiolock and GyroLock systems. Both are presented here and assessed based on experimental results obtained in vivo, proving the efficiency of the active stabilization approach for the heart motion compensation.

In this chapter, we present a fully automatic and accurate segmentation framework for 2D cardiac tagged MR images, a semiautomatic method for 3D segmentation from CT data, and the results of blood flow simulation using these highly detailed models. The 2D segmentation system consists of a semiautomatic segmentation framework to obtain the training contours, and a learning-based framework that is trained by the semiautomatic results, and achieves fully automatic and accurate segmentation.

We then present a method to simulate and visualize blood flow through the human heart, using the reconstructed 4D motion of the endocardial surface of the left ventricle as boundary conditions. The reconstruction captures the motion of the full 3D surfaces of the complex features, such as the papillary muscles and the ventricular trabeculae. We use visualizations of the flow field to view the interactions between the blood and the trabeculae in far more detail than has been achieved previously, which promises to give a better understanding of cardiac flow. Finally, we use our simulation results to compare the blood flow within one healthy heart and two diseased hearts.

Aortic dissections are a lethal disease affecting thousands of people in the USA each year. This chapter illustrates the application of personalized computational fluid dynamics (CFD) in understanding the hemodynamics of DeBakey type III/Stanford B aortic dissections (dissections confined to the descending aorta), pre- and post-surgical interventions, and simulating hemodynamic changes as a pretreatment planning tool. In this regard, CFD simulations using patient-derived data may be useful for gaining a conceptual understanding of the hemodynamic factors for a particular aortic dissection before intervention and how these factors change with treatment or disease progression. CFD simulations have a potential role in evaluating a number of scenarios and configurations, guiding therapy, and providing a basis for outcome prediction.

Atherosclerosis develops from oxidized low-density lipoprotein (LDL) molecules. When oxidized LDL evolves in plaque formation within an artery wall, a series of reactions occur to repair the damage to the artery wall caused by oxidized LDL. The body’s immune system responds to damage to the artery wall caused by oxidized LDL by sending specialized white blood cells-macrophages (Mphs) to absorb the oxidized-LDL and form specialized foam cells. Macrophages accumulate inside arterial intima. Also smooth muscle cells accumulate in the atherosclerotic arterial intima, where they proliferate and secrete extracellular matrix to generate a fibrous cap.

In this study, a model of plaque formation on the pig left anterior descending (LAD) coronary artery is simulated numerically using a specific animal data obtained from IVUS and histological recordings. The 3D bloodflow is described by the Navier–Stokes equations, together with the continuity equation. Mass transfer within the blood lumen and through the arterial wall is coupled with the blood flow and is modeled by a convection-diffusion equation. The LDL transports in lumen of the vessel and through the vessel tissue (which has a mass consumption term) are coupled by Kedem–Katchalsky equations. The inflammatory process is modeled using three additional reaction-diffusion partial differential equations. A full three-dimensional model was created which includes blood flow and LDL concentration, as well as plaque formation. Matching of IVUS and histological animal data is performed using a 3D histological image reconstruction and 3D deformation of elastic body. Computed concentration of macrophages indicates that there is a newly formed matter in the intima, especially in the LAD 15 mm region from bifurcation. Understanding and prediction of the evolution of atherosclerotic plaques either into vulnerable or stable plaques are major tasks for the medical community.

Vascular adaptation following local injury occurs through a combination of intimal hyperplasia and wall (inward/outward) remodeling. Over the past two decades, researchers have applied a wide variety of approaches to investigate neointimal hyperplasia and vascular remodeling in an effort to identify novel therapeutic strategies. However, despite incremental progress over these decades, specific cause/effect links between hemodynamic factors, inflammatory biochemical mediators, cellular effectors, and vascular occlusive phenotype remain lacking.

We propose in this paper a first cellular automata model to implement the feedback mechanism between environment condition described by continuous dynamic and tissue plasticity described at the cellular level with the cellular automata. We propose in particular a careful construction of the probabilistic rules of the model from in vitro experiments results that can be validated against in vivo data.

The transport of matter is the fundamental biomechanical process in living organisms. It occurs on all time and length scales, from picoseconds to days and from molecular to organ levels. The role of computer modeling is to help elucidating the basic mechanisms in the transport phenomena, investigated experimentally under laboratory and clinical conditions.

In this report we briefly present computational approaches to model transport on small—nanoscale, within nanoconfinement, and on macroscale—considering transport of Low-Density Lipoprotein (LDL) within blood in arterial vessel and within blood vessel tissue. The results illustrate important surface effects on diffusion of molecules when dimensions of diffusion domain are comparable to those of the transporting molecules. On the other hand, the transport model of LDL in the vessel lumen and through tissue and the model of plaque initiation can help in development of drugs and procedures in treating atherosclerosis.

This chapter presents an example of an application of a mathematical model: the goal is here to help clinicians evaluate the aggressiveness of some metastases to the lung. For this matter, an adequate spatial model is described and two algorithms (one using a reduced model approach and the other one a sensitivity technique) are shown. They allow us to find reasonable values of the parameters of this model for a given patient with a sequence of medical images. The quality of the prognosis obtained through the calibrated model is then illustrated with several clinical cases.

Gliomas are highly invasive primary brain tumors, notorious for their recurrence after treatment, and are considered uniformly fatal. Confounding progress is the fact that there is a diffuse extent of tumor cell invasion well beyond what is visible on routine clinical imaging such as MRI. By incorporating diffusion tensor imaging (DTI) which shows the directional orientation of fiber tracts in the brain, we compare patient-specific model simulations to observed tumor growth for two patients, visually, volumetrically and spatially to quantify the effect of anisotropic diffusion on the ability to predict the actual shape and diffuse invasion of tumor as observed on MRI. The ultimate goal is the development of the best patient-specific tool for predicting brain tumor growth and invasion in individual patients, which can aid in treatment planning.

The dual classification of cancer as localized or metastatic disease is one of the key points in the elaboration of the best therapy for each patient. Nevertheless, many studies reveal that part of these localized diseases is already metastatic. The presence of undetectable or micro-metastases explains the necessity of adjuvant chemotherapies after resection of the primary tumor even for some T1N0M0 cancer. There is probably a continuum between these two stages.We expose here how a mathematical model of growing metastases could reflect this continuum of the disease and how such a model could help the oncologists in the choice of the treatment. This phenomenological model is based on a structured transport equations with nonlocal boundary condition describing the evolution of the density of metastasis. Thanks to this model, we forge a new numerical index, that we call Metastatic Index, able to reveal either the micro-metastatic state of the patient, or the visible metastatic one. Numerical illustrations show how this new index can be used.

New surgical and localization techniques allow for precise and personalized evaluation and treatment of intractable epilepsies. These techniques include the use of subdural and depth electrodes for localization, and the potential use for cell-targeted stimulation using optogenetics as part of treatment. Computer modeling of seizures, also individualized to the patient, will be important in order to make full use of the potential of these new techniques. This is because epilepsy is a complex dynamical disease involving multiple scales across both time and space. These complex dynamics make prediction extremely difficult. Cause and effect are not cleanly separable, as multiple embedded causal loops allow for many scales of unintended consequence. We demonstrate here a small model of sensory neocortex which can be used to look at the effects of microablations or microstimulation. We show that ablations in this network can either prevent spread or prevent occurrence of the seizure. In this example, focal electrical stimulation was not able to terminate a seizure but selective stimulation of inhibitory cells, a future possibility through use of optogenetics, was efficacious.

In this work parts of the research done at HLRS, to derive an anisotropic, linear elastic and inhomogeneous material model for cancellous bone from micro-focus computer tomographic data via direct mechanical simulations, are described.First a short introduction to the background of biomechanical simulations of bone-implant-systems is given. After that the direct mechanics approach to the calculation of continuum material data for micro structured materials is introduced. Since this method reveals some major drawbacks when utilizing the material data generated with it in continuum mechanical Finite Element Simulations the method is then extended to the calculation of what we call the “discrete, effective stiffness” of a micro structured volume element. To demonstrate the application of the procedure and clarify that not only micro structural effects can be captured by it, its application to the homogeneous unit volume as well as a real micro-focus computer tomographic dataset is shown in the results section.

Technological advances have enabled new paradigms for skill training using virtual reality and robotics. We present three recent research advances in the field of virtual reality and human–robot interaction (HRI) for training. First, skill assessment in these systems is discussed, with an emphasis on the derivation of meaningful and objective quantitative performance metrics from motion data acquired through sensors on the robotic devices. We show how such quantitative measures derived for the robotic stroke rehabilitation domain correlate strongly with clinical measures of motor impairment. For virtual reality-based task training, we present task analysis and motion-based performance metrics for a manual control task. Lastly, we describe specific challenges in the surgical domain, with a focus on the development of tasks for skills assessment in surgical robotics.

This chapter presents the implementation of a progressive haptic guidance scheme for training in a dynamic motor skill acquisition task similar to some dynamic surgical tasks. The training is administered in a haptic and visual virtual environment. The results of the task training protocol concurrently compare the performance and workload of the proposed haptic guidance scheme to a similar visual guidance scheme and to virtual practice with no guidance. The human-user training protocol lasted 11 sessions over a 2-month period. The computerized version of the NASA task load index was administered to all participants during each session, thereby providing subjective workload data across the entire protocol. The analysis of the experimental results demonstrates that only early in the protocol, the progressive haptic guidance group outperforms all other groups. The workload analysis suggests that participants using the proposed haptic scheme have a significantly lower mental load and report less frustration than the others. These findings can be transferred to other virtual training environments used for surgical task training.