Ground Truth Model Research Articles

Prediction and Verification of Parker Solar Probe Solar Wind Sources at 13.3 R⊙

AbstractDrawing connections between heliospheric spacecraft and solar wind sources is a vital step in understanding the evolution of the solar corona into the solar wind and contextualizing in situ timeseries. Furthermore, making advanced predictions of this linkage for ongoing heliospheric missions, such as Parker Solar Probe (Parker), is necessary for achieving useful coordinated remote observations and maximizing scientific return. The general procedure for estimating such connectivity is straightforward (i.e., magnetic field line tracing in a coronal model) but validating the resulting estimates is difficult due to the lack of an independent ground truth and limited model constraints. In its most recent orbits, Parker has reached perihelia of 13.3R⊙ and moreover travels extremely fast prograde relative to the solar surface, covering over 120° longitude in 3 days. Here we present footpoint predictions and subsequent validation efforts for Parker Encounter 10, the first of the 13.3R⊙ orbits, which occurred in November 2021. We show that the longitudinal dependence of in situ plasma data from these novel orbits provides a powerful method of footpoint validation. With reference to other encounters, we also illustrate that the conditions under which source mapping is most accurate for near‐ecliptic spacecraft (such as Parker) occur when solar activity is low, but also require that the heliospheric current sheet is strongly warped by mid‐latitude or equatorial coronal holes. Lastly, we comment on the large‐scale coronal structure implied by the Encounter 10 mapping, highlighting an empirical equatorial cut of the Alfvèn surface consisting of localized protrusions above unipolar magnetic separatrices.

McSNAC: A software to approximate first-order signaling networks from mass cytometry data.

Mass cytometry (CyTOF) gives unprecedented opportunity to simultaneously measure up to 40 proteins in single cells, with a theoretical potential to reach 100 proteins. This high-dimensional single-cell information can be very useful in dissecting mechanisms of cellular activity. In particular, measuring abundances of signaling proteins like phospho-proteins can provide detailed information on the dynamics of single-cell signaling processes. However, computational analysis is required to reconstruct such networks with a mechanistic model. We propose our Mass cytometry Signaling Network Analysis Code (McSNAC), a new software capable of reconstructing signaling networks and estimating their kinetic parameters from CyTOF data. McSNAC approximates signaling networks as a network of first-order reactions between proteins. This assumption often breaks down as signaling reactions can involve binding and unbinding, enzymatic reactions, and other nonlinear constructions. Furthermore, McSNAC may be limited to approximating indirect interactions between protein species, as cytometry experiments are only able to assay a small fraction of protein species involved in signaling. We carry out a series of in silico experiments here to show (1) McSNAC is capable of accurately estimating the ground-truth model in a scalable manner when given data originating from a first-order system; (2) McSNAC is capable of qualitatively predicting outcomes to perturbations of species abundances in simple second-order reaction models and in a complex in silico nonlinear signaling network in which some proteins are unmeasured. These findings demonstrate that McSNAC can be a valuable screening tool for generating models of signaling networks from time-stamped CyTOF data.

Reinforcement Learning for Efficient Design Space Exploration With Variable Fidelity Analysis Models

Abstract Reinforcement learning algorithms can autonomously learn to search a design space for high-performance solutions. However, modern engineering often entails the use of computationally intensive simulation, which can lead to slower design timelines with highly iterative approaches such as reinforcement learning. This work provides a reinforcement learning framework that leverages models of varying fidelity to enable an effective solution search while reducing overall computational needs. Specifically, it utilizes models of varying fidelity while training the agent, iteratively progressing from low- to high fidelity. To demonstrate the effectiveness of the proposed framework, we apply it to two multimodal multi-objective constrained mixed integer nonlinear design problems involving the components of a ground and aerial vehicle. Specifically, for each problem, we utilize a high-fidelity and a low-fidelity deep neural network surrogate model, trained on performance data generated from underlying ground truth models. A tradeoff between solution quality and the proportion of low-fidelity surrogate model usage is observed. Specifically, high-quality solutions are achieved with substantial reductions in computational expense, showcasing the effectiveness of the framework for design problems where the use of just a high-fidelity model is infeasible. This solution quality-computational efficiency tradeoff is contextualized by visualizing the exploration behavior of the design agents.

Machine learning for detecting substance use behaviors from wearable biosensor data streams: A survey and outlook for 5G/6G connectivity

Wearable biosensors allow continuous monitoring of physiological and behavioral signals associated with substance use disorders. Nevertheless, generating clinically meaningful insights requires novel machine learning techniques capable of handling high-velocity, noisy data streams. This paper surveys existing algorithms proposed for detecting substance use from wearables, highlighting key trends such as deep learning approaches and personalization. Ongoing challenges related to ground truth labeling, model interpretation, and computational efficiency are also discussed. The advent of 5G/6G networks is poised to transform this field by facilitating coordinated sensing across multiple wearables and enabling advanced on-device analytics. This work consolidates progress in wearable analytics for substance use monitoring and identifies open problems to steer future research. An experimental evaluation using wearable data from 20 participants demonstrates that a personalized CNN model achieved the best performance in detecting cannabis use events, boasting 89% precision, 83% recall, and a 0.94 AUROC score, outperforming classical machine learning approaches such as random forests.

TopasOpt: An open-source library for optimization with Topas Monte Carlo.

To describe and test TopasOpt: a free, open-source and extensible library for performing mathematical optimization of Monte Carlo simulations in Topas. TopasOpt enables any Topas model to be transformed into an optimization problem, and any parameter within the model to be treated as an optimization variable. Three case studies are presented. The starting model consists of a 10 MeV electron beam striking a tungsten target. The resulting bremsstrahlung X-ray spectrum is collimated by a primary and secondary collimator before being scored in a water tank. In the first case study (electron phase space optimization), five parameters describing the electron beam were treated as optimization variables and assigned a random starting value. An objective function was defined based on differences of depth-dose and profiles in water between the original (ground truth) model and a given model generated by TopasOpt. The problem was solved using Bayesian Optimization and the Nelder-Mead method. One hundred iterations were run in each case. In the second case study, (collimator geometry optimization), this process was repeated, but three geometric parameters defining the secondary collimator were treated as optimization variables and assigned random starting values, and forty iterations were run. In the third case study, the optimization was repeated with different number of primary particles to study the effect of noise on convergence. For case 1 (phase space optimization), both optimization algorithms successfully minimized the objective function, with absolute mean differences in profile dose of 0.4% (Bayesian) and 0.3% (Nelder-Mead) and 0.2% in depth-dose for both algorithms. The beam energy was recovered to within 1%, however some parameters had relative errors of up to 171% - a result consistent with the known X-ray dose is insensitivity to many electron beam parameters. For case 2 (geometry optimization), absolute mean differences in profile dose were 0.6% (Bayesian) and 0.9% (Nelder-Mead), and 0.5% and 0.9% in depth-dose. The maximum percentage error in any parameter was 9% with Bayesian Optimization and 28% with Nelder-Mead. Finally, the Bayesian Optimization algorithm was demonstrated to be robust to moderate levels of noise; when the standard deviation of the objective function was 16% of the mean, the maximum error in any parameter value was 16%, and the absolute mean difference in dose was 0.9% (profile) and 0.8% (depth-dose). An open-source library for optimization with Topas Monte Carlo has been developed, tested, and released. This tool will improve accuracy and efficiency in any situation in which the optimal value of a parameter in a Monte Carlo simulation is unknown. Applications for this tool include (1) The design of new components (2) Reverse engineering of models based on limited experimental or published data, and (3) Tuning of Monte Carlo "hyper parameters" such as variance reduction, physics settings, or scoring parameters.

Evaluation of Photogrammetry Tools following Progress Detection of Rebar towards Sustainable Construction Processes

In the attainment of digitization and sustainable solutions under Industry 4.0, effective and economical technology like photogrammetry is gaining popularity in every field among professionals and researchers alike. In the market, various photogrammetry tools are available. These tools employ different techniques and it is hard to identify the best among them. This study is an attempt to develop a methodology for the assessment of photogrammetry tools. Overall, 37 photogrammetry tools were found via literature review and open sources, out of which 12 tools were shortlisted. The evaluation process consisted of three steps, i.e., metadata and visual inspection, comparison with the ground truth model, and comparison with the averaged-merged point cloud model. In addition, a validation test was also performed on the final sorted photogrammetry tools. This study followed a sustainable construction progress monitoring theme for rebar and covered the maximum number of photogrammetry tools for comparison by considering the most authentic evaluation and validation techniques, which make it exclusive.

Plane Fitting in 3D Reconstruction to Preserve Smooth Homogeneous Surfaces

Photogrammetric techniques for weakly-textured surfaces without sufficient information about the R (red), G (green) and B (blue) primary colors of light are challenging. Considering that most urban or indoor object surfaces follow simple geometric shapes, a novel method for reconstructing smooth homogeneous planar surfaces based on MVS (Multi-View Stereo) is proposed. The idea behind it is to extract enough features for the image description, and to refine the dense points generated by the depth values of pixels with plane fitting, to favor the alignment of the surface to the detected planes. The SIFT (Scale Invariant Feature Transform) and AKAZE (Accelerated-KAZE) feature extraction algorithms are combined to ensure robustness and help retrieve connections in small samples. The smoothness of the enclosed watertight Poisson surface can be enhanced by enforcing the 3D points to be projected onto the absolute planes detected by a RANSAC (Random Sample Consensus)-based approach. Experimental evaluations of both cloud-to-mesh comparisons in the per-vertex distances with the ground truth models and visual comparisons with a popular mesh filtering based post-processing method indicate that the proposed method can considerably retain the integrity and smoothness of the reconstruction results. Combined with other primitive fittings, the reconstruction extent of homogeneous surfaces can be further extended, serving as primitive models for 3D building reconstruction, and providing guidance for future works in photogrammetry and 3D surface reconstruction.

Topology-based UAV path planning for multi-view stereo 3D reconstruction of complex structures

This paper introduces a new UAV path planning method for creating high-quality 3D reconstruction models of large and complex structures. The core of the new method is incorporating the topology information of the surveyed 3D structure to decompose the multi-view stereo path planning into a collection of overlapped view optimization problems that can be processed in parallel. Different from the existing state-of-the-arts that recursively select the vantage camera views, the new method iteratively resamples all nearby cameras (i.e., positions/orientations) together and achieves a substantial reduction in computation cost while improving reconstruction quality. The new approach also provides a higher-level automation function that facilitates field implementations by eliminating the need for redundant camera initialization as in existing studies. Validations are provided by measuring the variance between the reconstructions to the ground truth models. Results from three synthetic case studies and one real-world application are presented to demonstrate the improved performance. The new method is expected to be instrumental in expanding the adoption of UAV-based multi-view stereo 3D reconstruction of large and complex structures.

A neural-embedded discrete choice model: Learning taste representation with strengthened interpretability

Discrete choice models (DCMs) require a priori knowledge of the utility functions, especially how tastes vary across individuals. Utility misspecification may lead to biased estimates, inaccurate interpretations and limited predictability. In this paper, we utilize a neural network to learn taste representation. Our formulation consists of two modules: a neural network (TasteNet) that learns taste parameters (e.g., time coefficient) as flexible functions of individual characteristics; and a multinomial logit (MNL) model with utility functions defined with expert knowledge. Taste parameters learned by the neural network are fed into the choice model and link the two modules.Our approach extends the L-MNL model (Sifringer et al., 2020) by allowing the neural network to learn the interactions between individual characteristics and alternative attributes. Moreover, we formalize and strengthen the interpretability condition — requiring realistic estimates of behavior indicators (e.g., value-of-time, elasticity) at the disaggregated level, which is crucial for a model to be suitable for scenario analysis and policy decisions. Through a unique network architecture and parameter transformation, we incorporate prior knowledge and guide the neural network to output realistic behavior indicators at the disaggregated level.We show that TasteNet-MNL reaches the ground-truth model’s predictability and recovers the nonlinear taste functions on synthetic data. Its estimated value-of-time and choice elasticities at the individual level are close to the ground truth. In contrast, exemplary logit models with misspecified systematic utility lead to biased parameter estimates and lower prediction accuracy. On a publicly available Swissmetro dataset, TasteNet-MNL outperforms benchmarking MNLs and Mixed Logit model’s predictability. It learns a broader spectrum of taste variations within the population and suggests a higher average value-of-time. Our source code is available for research and application.

3D localization of vena contracta using Doppler ICE imaging in tricuspid valve interventions

PurposeTricuspid valve (TV) interventions face the challenge of imaging the anatomy and tools because of the ‘TEE-unfriendly’ nature of the TV. In edge-to-edge TV repair, a core step is to position the clip perpendicular to the coaptation gap. In this study, we provide a semi-automated method to localize the VC from Doppler intracardiac echo (ICE) imaging in a tracked 3D space, thus providing a pre-mapped location of the coaptation gap to assist device positioning.MethodsA magnetically tracked ICE probe with Doppler imaging capabilities is employed in this study for imaging three patient-specific TVs placed in a pulsatile heart phantom. For each of the valves, the ICE probe is positioned to image the maximum regurgitant flow for five cardiac cycles. An algorithm then extracts the regurgitation imaging and computes the exact location of the vena contracta on the image.ResultsAcross the three pathological, patient-specific valves, the average distance error between the detected VC and the ground truth model is ({1.22 pm 2.00})mm. For each of the valves, one case represented the outlier where the algorithm misidentified the vena contracta to be near the annulus. In such cases, it is recommended to retake the five-second imaging data.ConclusionThis study presented a method for ultrasound-based localization of vena contracta in 3D space. Mapping such anatomical landmarks has the potential to assist with device positioning and to simplify tricuspid valve interventions by providing more contextual information to the interventionalists, thus enhancing their spatial awareness. Additionally, ICE can be used to provide live US and Doppler imaging of the complex TV anatomy throughout the procedure.

Performance comparison analysis of 3D reconstruction modeling software in construction site visualization and mapping

Unmanned aerial vehicle (UAV) technology has overcome the limitations of conventional construction management methods using advanced and automated visualization and 3D reconstruction modeling techniques. Although the mapping techniques and reconstruction modeling software can generate real-time and high-resolution descriptive textural, physical, and spatial data, they may fail to develop an accurate and complete 3D model of the construction site. To generate a quality 3D reconstruction model, the construction manager must optimize the trade-offs among three major software-selection factors: functionalities, technical capabilities, and the system hardware specifications. These factors directly affect the robust 3D reconstruction model of the construction site and objects. Accordingly, the purpose of this research was to apply nine well-established 3D reconstruction modeling software tools (DroneDeploy, COLMAP, 3DF+Zephyr, Autodesk Recap, LiMapper, PhotoModeler, 3D Survey, AgiSoft Photoscan, and Pix4D Mapper) and compare their performances and reliabilities in generating complete 3D models. The research was conducted in an eco-home building at the University of Technology, Malaysia. A series of regression analyses were conducted to compare the performances of the selected 3D reconstruction modeling software in alignment and registration, distance computing, geometric measurement, and plugin execution. Regression analysis determined that among the software programs, LiMapper had the strongest positive linear correlation with the ground truth model. Furthermore, the correlation analysis showed a statistically significant p-value for all software, except for 3D Survey. In addition, the research found that Autodesk Recap generated the most-robust and highest-quality dense point clouds. DroneDeploy can create an accurate point cloud and triangulation without using many points as required by COLMAP and LiMapper. It was concluded that most of the software is robustly, positively, and linearly correlated with the corresponding ground truth model. In the future, other factors involving software selection should be studied, such as vendor-related, user-related, and automation factors.

Evaluation of segmentation accuracy and its impact on patient-specific CFD analysis

Medical image segmentation, especially for biological soft tissues, is an issue of great interest. The aim of this study is to evaluate the segmentation performance of a commercial and an open-source software, to segment aortic root and coronary arteries. 3D printing stereolithography technology was used to generate ground truth models, which were then re-acquired by means of a micro-CT scanner. Measurements from the printed and reconstructed models with both the software were compared, in order to evaluate the level of agreement. In the second phase of this study, Computational Fluid Dynamics (CFD) simulations were conducted, to compare the outputs between the models segmented with the two software. The goal was to understand how differences in the segmentation process propagate in CFD results. Results showed that both software guarantee satisfactory segmentation performance, with average geometrical differences between reconstructed and physical models in the order of a few percentage points. However, when we consider thin details, as a sharp stenotic region, the commercial validated software seems to be more accurate in replicating the real anatomy. We also realized how apparently negligible geometrical differences, varying the employed software, can turn into enormous variations of hemodynamic parameters, such as velocity and wall shear stress, which place in the centre the delicate role the segmentation process holds. This evidence is crucial in the biomedical field and especially in a coronary arteries study, where CFD simulations can be exploited as a starting point for surgery considerations.

Feasibility Study of Drone-Based 3-D Measurement of Defects in Concrete Structures

Recognition of defects in concrete structures, identification of cracks, concrete spalling, or other geometrical defects are important tools for structural damage detection. Defects in structures can include cracks, but also missing parts, due to the wear caused by weather or aging phenomena. These last types of defects in structures can be identified using red-green-blue (RGB) cameras, but the level of damage could be difficult to evaluate with 2-D images. In this sense, the application of 3-D reconstruction techniques can be helpful to determine the 3-D dimensions of spalling, swelling of concrete or the presence of visible steel parts of reinforced concrete. The use of drones for this type of measurement is very attractive for reducing the costs and time of measurement campaigns. However, the lack of accurate trajectory information and the vibrations affect the accuracy of 3-D measurements. In this article, a metrological characterization of measurement systems for the evaluation and recognition of defects in concrete structures is presented, starting from the acquisition of 3-D point clouds by low-cost time-of-flight (ToF) sensors, placed on drones. To evaluate the uncertainty of these systems, a mock-up with realistic defects was developed and characterized using a reference 3-D scanner. The 3-D reconstructions obtained via the selected sensors were used to evaluate the discrepancies of the 3-D shape compared to a ground truth model and the uncertainty of the selected scanners. The results show that, for all the defects tested, the standard deviation of the discrepancies between the defect reconstructed using the drone and the ground truth is below 2.5 mm.

Reconstruction of stomach geometry using magnetic source localization.

Routine diagnosis of gastric motility disorders represents a significant problem to current clinical practice. The non-invasive electrogastrogram (EGG) and magnetogastrogram (MGG) enable the assessment of gastric slow wave (SW) dysrhythmias that are associated with motility disorders. However, both modalities lack standardized methods for reliably detecting patterns of SW activity. Subject-specific anatomical information relating to the geometry of the stomach and its position within the torso have the potential to aid the development of relations between SWs and far-fields. In this study, we demonstrated the feasibility of using magnetic source localization to reconstruct the geometry of an anatomically realistic 3D stomach model. The magnetic fields produced by a small (6.35 × 6.35 mm) N35 neodymium magnet sequentially positioned at 64 positions were recorded by an array of 27 magnetometers. Finally, the magnetic dipole approximation and a particle swarm optimizer were used to estimate the position and orientation of the permanent magnet. Median position and orientation errors of 3.8 mm and 7.3° were achieved. The estimated positions were used to construct a surface mesh, and the Hausdorff Distance and Average Hausdorff Distance dissimilarity metrics for the reconstructed and ground-truth models were 11.6 mm and 2.4 mm, respectively. The results indicate that source localization using the magnetic dipole model can successfully reconstruct the geometry of the stomach.

Reclassifying stroke lesion anatomy

Cognitive and behavioural outcomes in stroke reflect the interaction between two complex anatomically-distributed patterns: the functional organization of the brain and the structural distribution of ischaemic injury. Conventional outcome models—for individual prediction or population-level inference—commonly ignore this complexity, discarding anatomical variation beyond simple characteristics such as lesion volume. This sets a hard limit on the maximum fidelity such models can achieve. High-dimensional methods can overcome this problem, but only at prohibitively large data scales. Drawing on one of the largest published collections of anatomically-registered imaging of acute stroke—N = 1333—here we use non-linear dimensionality reduction to derive a succinct latent representation of the anatomical patterns of ischaemic injury, agglomerated into 21 distinct intuitive categories. We compare the maximal predictive performance it enables against both simpler low-dimensional and more complex high-dimensional representations, employing multiple empirically-informed ground truth models of distributed structure–outcome relationships. We show our representation sets a substantially higher ceiling on predictive fidelity than conventional low-dimensional approaches, but lower than that achievable within a high-dimensional framework. Where descriptive simplicity is a necessity, such as within clinical care or research trials of modest size, the representation we propose arguably offers a favourable compromise of compactness and fidelity.

Ventral intermediate nucleus structural connectivity-derived segmentation: anatomical reliability and variability

The Ventral intermediate nucleus (Vim) of thalamus is the most targeted structure for the treatment of drug-refractory tremors. Since methodological differences across existing studies are remarkable and no gold-standard pipeline is available, in this study, we tested different parcellation pipelines for tractography-derived putative Vim identification.Thalamic parcellation was performed on a high quality, multi-shell dataset and a downsampled, clinical-like dataset using two different diffusion signal modeling techniques and two different voxel classification criteria, thus implementing a total of four parcellation pipelines. The most reliable pipeline in terms of inter-subject variability has been picked and parcels putatively corresponding to motor thalamic nuclei have been selected by calculating similarity with a histology-based mask of Vim. Then, spatial relations with optimal stimulation points for the treatment of essential tremor have been quantified. Finally, effect of data quality and parcellation pipelines on a volumetric index of connectivity clusters has been assessed.We found that the pipeline characterized by higher-order signal modeling and threshold-based voxel classification criteria was the most reliable in terms of inter-subject variability regardless data quality. The maps putatively corresponding to Vim were those derived by precentral and dentate nucleus-thalamic connectivity. However, tractography-derived functional targets showed remarkable differences in shape and sizes when compared to a ground truth model based on histochemical staining on seriate sections of human brain. Thalamic voxels connected to contralateral dentate nucleus resulted to be the closest to literature-derived stimulation points for essential tremor but at the same time showing the most remarkable inter-subject variability. Finally, the volume of connectivity parcels resulted to be significantly influenced by data quality and parcellation pipelines. Hence, caution is warranted when performing thalamic connectivity-based segmentation for stereotactic targeting.

Comparison of approaches to transcriptomic analysis in multi-sampled tumors.

Intratumoral heterogeneity is a well-documented feature of human cancers and is associated with outcome and treatment resistance. However, a heterogeneous tumor transcriptome contributes an unknown level of variability to analyses of differentially expressed genes (DEGs) that may contribute to phenotypes of interest, including treatment response. Although current clinical practice and the vast majority of research studies use a single sample from each patient, decreasing costs of sequencing technologies and computing power have made repeated-measures analyses increasingly economical. Repeatedly sampling the same tumor increases the statistical power of DEG analysis, which is indispensable toward downstream analysis and also increases one's understanding of within-tumor variance, which may affect conclusions. Here, we compared five different methods for analyzing gene expression profiles derived from repeated sampling of human prostate tumors in two separate cohorts of patients. We also benchmarked the sensitivity of generalized linear models to linear mixed models for identifying DEGs contributing to relevant prostate cancer pathways based on a ground-truth model.

Very high resolution aboveground carbon mapping in subtropical thicket

Aboveground carbon (AGC) maps will assist with the monitoring and verification of carbon stored through restoration of subtropical thicket. Field methods are required to capture AGC ground truth data at the plot scale, but they are highly impractical for large area mapping of carbon stocks. Against this background, a remote sensing method to estimate AGC from very high spatial resolution multi-spectral imagery was developed. AGC ground truth was acquired for 85 plots in a small study area in the Baviaanskloof, South Africa. Using the ground truth, univariate and multivariate snapshot models were developed to predict AGC from features in a 2017 WorldView-3 (0.34-m resolution 8-band) satellite image. Informative features were selected from a large set of spectral, textural, and vegetation index features using stepwise forward selection. Using this approach, the multivariate model produced a coefficient of determination (R2) of 0.886 and root mean square error of 2.862 t C ha − 1. This study demonstrates the efficacy of regression approaches for estimating AGC from multi-spectral imagery and provides a foundation for the spatial and temporal extension of AGC remote sensing in the thicket biome.

Abstract 396: Beyond thresholds in precision oncology: Use of probability of benefit as function of continuous biomarker levels leads to better patient care

Abstract The current threshold-based precision medicine paradigm stems in part from the ‘all-or-nothing' nature of regulatory labeling: a lung cancer patient can access pembrolizumab on-label only if tumor mutational burden (TMB) is at least 10 mutations/megabase, and crizotinib only if at least 15% of their tumor cells have ALK rearrangements (Pembrolizumab and Crizotinib package inserts). However, the drug-centric question - defining the patients most likely to respond to the drug – does not necessarily optimize patient care: in this case, choosing the best drug for a patient with nonzero levels of TMB and ALK. Furthermore, the current paradigm requires the drug sponsor to choose a threshold between a low value, to provide access to as many potentially benefitting patients as possible, and a high value, to maximize probability of success of the pivotal trial. We hypothesize that modeling probability of benefit (PoB) from each drug as a function of continuous biomarker levels (CBLs), and then giving the patient the treatment with the larger PoB not only eliminates the sponsor threshold selection dilemma, it provides better patient care. To test this hypothesis, we developed a simulation study. Given two drug/biomarker pairs, DX/BX and DY/BY, a ‘training' population of 100 virtual patients responsive to DX, DY, or neither DX nor DY was generated using a ‘ground truth' model of PoB from DX or DY as a function of patient CBLs of BX and BY, which were drawn from a prespecified random distribution. Logistic regression models were trained to the patients' benefit outcomes to quantify PoBs as functions of CBLs. The threshold paradigm was simulated by choosing an ‘altruistic' optimal biomarker threshold (OBT) for each drug which minimizes the net error rate (NER = False Positive + False Negative rate). Then a ‘test' population of 1000 patients was generated using the same ground truth model used to generate the training population. We then applied the trained PoB model to give each virtual test patient the drug with the highest PoB based on CBLs and applied the threshold method to give each virtual test patient a drug only if the corresponding CBL > OBT. Finally, we compared the NER between PoB and OBT methods. NERs were estimated for three methods: OBT yielded a NER of 20.4%. Maximum univariate PoB, where PoB(DX) is a function of CBL BX alone and PoB(DY) is a function of CBL BY alone, yielded a NER of 14.8%. Maximum bivariate PoB, where PoB(DX) and PoB(DY) are functions of both CBLs BX and BY, yielded a NER of 10.7%. Our analysis predicts that of 100 patients treated according to the OBT method, 20 would either not benefit from the drug they were given or did not get a drug that would have benefited them. In contrast only 10-15 patients would be incorrectly treated using the PoB method, representing a 25-50% lower error rate than OBT. We believe the PoB method warrants consideration in the future of precision oncology. Citation Format: Cameron McBride, Dean Bottino. Beyond thresholds in precision oncology: Use of probability of benefit as function of continuous biomarker levels leads to better patient care [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 396.

Investigating the reference domain influence in personalised models of cardiac mechanics : Effect of unloaded geometry on cardiac biomechanics.

A major concern in personalised models of heart mechanics is the unknown zero-pressure domain, a prerequisite for accurately predicting cardiac biomechanics. As the reference configuration cannot be captured by clinical data, studies often employ in-vivo frames which are unlikely to correspond to unloaded geometries. Alternatively, zero-pressure domain is approximated through inverse methodologies, which, however, entail assumptions pertaining to boundary conditions and material parameters. Both approaches are likely to introduce biases in estimated biomechanical properties; nevertheless, quantification of these effects is unattainable without ground-truth data. In this work, we assess the unloaded state influence on model-derived biomechanics, by employing an in-silico modelling framework relying on experimental data on porcine hearts. In-vivo images are used for model personalisation, while in-situ experiments provide a reliable approximation of the reference domain, creating a unique opportunity for a validation study. Personalised whole-cycle cardiac models are developed which employ different reference domains (image-derived, inversely estimated) and are compared against ground-truth model outcomes. Simulations are conducted with varying boundary conditions, to investigate the effect of data-derived constraints on model accuracy. Attention is given to modelling the influence of the ribcage on the epicardium, due to its close proximity to the heart in the porcine anatomy. Our results find merit in both approaches for dealing with the unknown reference domain, but also demonstrate differences in estimated biomechanical quantities such as material parameters, strains and stresses. Notably, they highlight the importance of a boundary condition accounting for the constraining influence of the ribcage, in forward and inverse biomechanical models.