Novel artificial intelligence machine learning approaches to precisely predict survival and site-specific recurrence in cervical cancer: A multi-institutional study

BackgroundMachine learning (ML) has been gradually integrated into oncologic research but seldom applied to predict cervical cancer (CC), and no model has been reported to predict survival and site-specific recurrence simultaneously. Thus, we aimed to develop ML models to predict survival and site-specific recurrence in CC and to guide individual surveillance.MethodsWe retrospectively collected data on CC patients from 2006 to 2017 in four hospitals. The survival or recurrence predictive value of the variables was analyzed using multivariate Cox, principal component, and K-means clustering analyses. The predictive performances of eight ML models were compared with logistic or Cox models. A novel web-based predictive calculator was developed based on the ML algorithms.ResultsThis study included 5112 women for analysis (268 deaths, 343 recurrences): (1) For site-specific recurrence, larger tumor size was associated with local recurrence, while positive lymph nodes were associated with distant recurrence. (2) The ML models exhibited better prognostic predictive performance than traditional models. (3) The ML models were superior to traditional models when multiple variables were used. (4) A novel predictive web-based calculator was developed and externally validated to predict survival and site-specific recurrence.ConclusionML models might be a better analytic approach in CC prognostic prediction than traditional models as they can predict survival and site-specific recurrence simultaneously, especially when using multiple variables. Moreover, our novel web-based calculator may provide clinicians with useful information and help them make individual postoperative follow-up plans and further treatment strategies.     

Soft tissue structures resisting anterior instability in a computational glenohumeral joint model

The glenohumeral joint is the most dislocated joint in the body due to the lack of bony constraints and the dependence on soft tissue for stability. The roles that various structures provide to joint function are important for understanding injury treatment and orthopaedic device design purposes. The goal of this study was to develop a computational model of the glenohumeral joint whereby joint behaviour was dictated by articular contact, ligamentous constraints, muscle loading and external perturbations. The bone structure of the computational model consisted of assembled computer tomographic images of the scapula, humerus and clavicle. The soft tissue elements were composed of forces and tension-only springs that represented muscles and ligaments. Validation of this model was achieved by comparing computational predictions to the results of a cadaveric experiment in which the relative contribution of muscles and ligaments to anterior joint stability was examined. The computational model predicted an anterior subluxation force that was similar to the cadaveric experimental results in humeral external rotation. The individual structure results showed the subscapularis to be critical to stabilisation in both neutral and external rotations, the biceps stabilised the joint in neutral but not in external rotation, and the inferior glenohumeral ligament resisted anterior displacement only in external rotation. The model's predictions were similar to the conclusions of the cadaveric experiment and the literature. Knowledge gained from this type of model could assist in further understanding the contribution of soft tissue stabilisers to joint function, pre-operative planning or the design of orthopaedic implants.     

CFD analysis of the Ahmed Glaucoma Valve and design of an alternative device

Computational fluid dynamics (CFD) modelling based on a commercial package, FLUENT, has been used in the present study. The primary aim of this study is to develop a novel implant by employing CFD techniques. Firstly, CFD analyses on the best design commercially available, which is the Ahmed Glaucoma Valve (AGV^®), are accomplished. In the light of the results, the new design focus is selected as the valve. The new design is analysed using GAMBIT and FLUENT software. CFD analyses of the new design and the AGV^® are compared and the strengths of the new design are revealed. The results are also compared with the experimental studies AGV^® in the literature. It is deduced that the proposed model shows a nonlinear pressure drop response, which is quite similar to that of AGV^®. The optimum combination would be a flow rate of 2.5 μl/min and a pressure drop of 1054.58 Pa for the proposed model.     

A method for 3D detection of symmetry line in asymmetric postures

The methods for symmetry line detection presented in the literature are typically suited to analyse symmetric upright postures, both standing and seated. The proposed method focuses on the symmetry line detection in subjects assuming asymmetric postures in which this line falls far outside the sagittal plane. The proposed approach evaluates the symmetry line by means of an autoregressive process in order to determine the set of planes suited to slice the back coherently with its geometric spatial configuration. The method is analysed assuming the cutaneous marking as reference and it is compared with a previous one, also developed by these authors. Results are analysed and critically discussed.     

A computational model of in vitro angiogenesis based on extracellular matrix fibre orientation

Recent interest in the process of vascularisation within the biomedical community has motivated numerous new research efforts focusing on the process of angiogenesis. Although the role of chemical factors during angiogenesis has been well documented, the role of mechanical factors, such as the interaction between angiogenic vessels and the extracellular matrix, remains poorly understood. In vitro methods for studying angiogenesis exist; however, measurements available using such techniques often suffer from limited spatial and temporal resolutions. For this reason, computational models have been extensively employed to investigate various aspects of angiogenesis. This paper outlines the formulation and validation of a simple and robust computational model developed to accurately simulate angiogenesis based on length, branching and orientation morphometrics collected from vascularised tissue constructs. Microvessels were represented as a series of connected line segments. The morphology of the vessels was determined by a linear combination of the collagen fibre orientation, the vessel density gradient and a random walk component. Excellent agreement was observed between computational and experimental morphometric data over time. Computational predictions of microvessel orientation within an anisotropic matrix correlated well with experimental data. The accuracy of this modelling approach makes it a valuable platform for investigating the role of mechanical interactions during angiogenesis.     

Effect of spatial inlet velocity profiles on the vortex formation pattern in a dilated left ventricle

Despite the advancement of cardiac imaging technologies, these have traditionally been limited to global geometrical measurements. Computational fluid dynamics (CFD) has emerged as a reliable tool that provides flow ?eld information and other variables essential for the assessment of the cardiac function. Extensive studies have shown that vortex formation and propagation during the filling phase acts as a promising indicator for the diagnosis of the cardiac health condition. Proper setting of the boundary conditions is crucial in a CFD study as they are important determinants, that affect the simulation results. In this article, the effect of different transmitral velocity profiles (parabolic and uniform profile) on the vortex formation patterns during diastole was studied in a ventricle with dilated cardiomyopathy (DCM). The resulting vortex evolution pattern using the uniform inlet velocity profile agreed with that reported in the literature, which revealed an increase in thrombus risk in a ventricle with DCM. However the application of a parabolic velocity profile at the inlet yields a deviated vortical flow pattern and overestimates the propagation velocity of the vortex ring towards the apex of the ventricle. This study highlighted that uniform inlet velocity profile should be applied in the study of the filling dynamics in a left ventricle because it produces results closer to that observed experimentally.     

Patient-specific computational haemodynamics: generation of structured and conformal hexahedral meshes from triangulated surfaces of vascular bifurcations

Measuring the blood flow is still limited by current imaging technologies and is generally overcome using computational fluid dynamics (CFD) which, because of the complex geometry of blood vessels, has widely relied on tetrahedral meshes. Hexahedral meshes offer more accurate results with lower-density meshes and faster computation as compared to tetrahedral meshes, but their use is limited by the far more complex mesh generation. We present a robust methodology for conformal and structured hexahedral mesh generation – applicable to complex arterial geometries as bifurcating vessels – starting from triangulated surfaces. Cutting planes are used to slice the lumen surface and to construct longitudinal Bezier splines. Afterwards, an isoparametric transformation is used to map a parametrically defined quadrilateral surface mesh into the vessel volume, resulting in stacks of sections which can then be used for sweeping. Being robust and open source based, this methodology may improve the current standard in patient-specific mesh generation and enhance the reliability of CFD to patient-specific haemodynamics.     

A finite element model of stress-mediated vascular adaptation: application to abdominal aortic aneurysms

Despite rapid expansion of our knowledge of vascular adaptation, developing patient-specific models of diseased arteries is still an open problem. In this study, we extend existing finite element models of stress-mediated growth and remodelling of arteries to incorporate a medical image-based geometry of a healthy aorta and, then, simulate abdominal aortic aneurysm. Degradation of elastin initiates a local dilatation of the aorta while stress-mediated turnover of collagen and smooth muscle compensates the loss of elastin. Stress distributions and expansion rates during the aneurysm growth are studied for multiple spatial distribution functions of elastin degradation and kinetic parameters. Temporal variations of the degradation function are also investigated with either direct time-dependent degradation or stretch-induced degradation as possible biochemical and biomechanical mechanisms for elastin degradation. The results show that this computational model has the capability to capture the complexities of aneurysm progression due to variations of geometry, extent of damage and stress-mediated turnover as a step towards patient-specific modelling.     

An automatic 2D–3D image matching method for reproducing spatial knee joint positions using single or dual fluoroscopic images

Fluoroscopic image technique, using either a single image or dual images, has been widely applied to measure in vivo human knee joint kinematics. However, few studies have compared the advantages of using single and dual fluoroscopic images. Furthermore, due to the size limitation of the image intensifiers, it is possible that only a portion of the knee joint could be captured by the fluoroscopy during dynamic knee joint motion. In this paper, we presented a systematic evaluation of an automatic 2D–3D image matching method in reproducing spatial knee joint positions using either single or dual fluoroscopic image techniques. The data indicated that for the femur and tibia, their spatial positions could be determined with an accuracy and precision less than 0.2 mm in translation and less than 0.4° in orientation when dual fluoroscopic images were used. Using single fluoroscopic images, the method could produce satisfactory accuracy in joint positions in the imaging plane (in average up to 0.5 mm in translation and 1.3° in rotation), but large variations along the out-plane direction (in average up to 4.0 mm in translation and 2.2° in rotation). The precision of using single fluoroscopic images to determine the actual knee positions was worse than its accuracy obtained. The data also indicated that when using dual fluoroscopic image technique, if the knee joint outlines in one image were incomplete by 80%, the algorithm could still reproduce the joint positions with high precisions.