Explore the vast range of titles available.

Increasing digitalization in the medical domain gives rise to large amounts of health care data, which has the potential to expand clinical knowledge and transform patient care if leveraged through artificial intelligence (AI). Yet, big data and AI oftentimes cannot unlock their full potential at scale, owing to nonstandardized data formats, lack of technical and semantic data interoperability, and limited cooperation between stakeholders in the health care system. Despite the existence of standardized data formats for the medical domain, such as Fast Healthcare Interoperability Resources (FHIR), their prevalence and usability for AI remain limited. In this paper, we developed a data harmonization pipeline (DHP) for clinical data sets relying on the common FHIR data standard. We validated the performance and usability of our FHIR-DHP with data from the Medical Information Mart for Intensive Care IV database. We present the FHIR-DHP workflow in respect of the transformation of \"raw\" hospital records into a harmonized, AI-friendly data representation. The pipeline consists of the following 5 key preprocessing steps: querying of data from hospital database, FHIR mapping, syntactic validation, transfer of harmonized data into the patient-model database, and export of data in an AI-friendly format for further medical applications. A detailed example of FHIR-DHP execution was presented for clinical diagnoses records. Our approach enables the scalable and needs-driven data modeling of large and heterogenous clinical data sets. The FHIR-DHP is a pivotal step toward increasing cooperation, interoperability, and quality of patient care in the clinical routine and for medical research.

Importance The entry of artificial intelligence into medicine is pending. Several methods have been used for the predictions of structured neuroimaging data, yet nobody compared them in this context. Objective Multi-class prediction is key for building computational aid systems for differential diagnosis. We compared support vector machine, random forest, gradient boosting, and deep feed-forward neural networks for the classification of different neurodegenerative syndromes based on structural magnetic resonance imaging. Design, setting, and participants Atlas-based volumetry was performed on multi-centric T1-weighted MRI data from 940 subjects, i.e., 124 healthy controls and 816 patients with ten different neurodegenerative diseases, leading to a multi-diagnostic multi-class classification task with eleven different classes. Interventions N.A. Main outcomes and measures Cohen’s kappa, accuracy, and F1-score to assess model performance. Results Overall, the neural network produced both the best performance measures and the most robust results. The smaller classes however were better classified by either the ensemble learning methods or the support vector machine, while performance measures for small classes were comparatively low, as expected. Diseases with regionally specific and pronounced atrophy patterns were generally better classified than diseases with widespread and rather weak atrophy. Conclusions and relevance Our study furthermore underlines the necessity of larger data sets but also calls for a careful consideration of different machine learning methods that can handle the type of data and the classification task best.

As energy metabolism in the brain is largely oxidative, the measurement of cerebral metabolic rate of oxygen consumption (CMRO2) is a desirable biomarker for quantifying brain activity and tissue viability. Currently, PET techniques based on oxygen isotopes are the gold standard for obtaining whole brain CMRO2 maps. Among MRI techniques that have been developed as an alternative are dual calibrated fMRI (dcFMRI) methods, which exploit simultaneous measurements of BOLD and ASL signals during a hypercapnic-hyperoxic experiment to modulate brain blood flow and oxygenation. In this study we quantified the repeatability of a dcFMRI approach developed in our lab, evaluating its limits and informing its application in studies aimed at characterising the metabolic state of human brain tissue over time. Our analysis focussed on the estimates of oxygen extraction fraction (OEF), cerebral blood flow (CBF), CBF-related cerebrovascular reactivity (CVR) and CMRO2 based on a forward model that describes analytically the acquired dual echo GRE signal. Indices of within- and between-session repeatability are calculated from two different datasets both at a bulk grey matter and at a voxel-wise resolution and finally compared with similar indices obtained from previous MRI and PET measurements. Within- and between-session values of intra-subject coefficient of variation (CVintra) calculated from bulk grey matter estimates 6.7 ± 6.6% (mean ± std.) and 10.5 ± 9.7% for OEF, 6.9 ± 6% and 5.5 ± 4.7% for CBF, 12 ± 9.7% and 12.3 ± 10% for CMRO2. Coefficient of variation (CV) and intraclass correlation coefficient (ICC) maps showed the spatial distribution of the repeatability metrics, informing on the feasibility limits of the method. In conclusion, results show an overall consistency of the estimated physiological parameters with literature reports and a satisfactory level of repeatability considering the higher spatial sensitivity compared to other MRI methods, with varied performance depending on the specific parameter under analysis, on the spatial resolution considered and on the study design. •Two datasets (N1 = 10 and N2 = 16) from dual calibrated fMRI experiments with respiratory tasks are considered.•Maps of OEF, CBF, CVR and CMRO2 calculated with a recently proposed estimation approach are shown.•Estimated indices of global and voxel-wise repeatability are reported for within- and between- session designs.•Results show a good trade-off between repeatability and spatial resolution when compared to previous studies.

As energy metabolism in the brain is largely oxidative, the measurement of cerebral metabolic rate of oxygen consumption (CMRO ) is a desirable biomarker for quantifying brain activity and tissue viability. Currently, PET techniques based on oxygen isotopes are the gold standard for obtaining whole brain CMRO maps. Among MRI techniques that have been developed as an alternative are dual calibrated fMRI (dcFMRI) methods, which exploit simultaneous measurements of BOLD and ASL signals during a hypercapnic-hyperoxic experiment to modulate brain blood flow and oxygenation. In this study we quantified the repeatability of a dcFMRI approach developed in our lab, evaluating its limits and informing its application in studies aimed at characterising the metabolic state of human brain tissue over time. Our analysis focussed on the estimates of oxygen extraction fraction (OEF), cerebral blood flow (CBF), CBF-related cerebrovascular reactivity (CVR) and CMRO based on a forward model that describes analytically the acquired dual echo GRE signal. Indices of within- and between-session repeatability are calculated from two different datasets both at a bulk grey matter and at a voxel-wise resolution and finally compared with similar indices obtained from previous MRI and PET measurements. Within- and between-session values of intra-subject coefficient of variation (CV ) calculated from bulk grey matter estimates 6.7 ± 6.6% (mean ± std.) and 10.5 ± 9.7% for OEF, 6.9 ± 6% and 5.5 ± 4.7% for CBF, 12 ± 9.7% and 12.3 ± 10% for CMRO . Coefficient of variation (CV) and intraclass correlation coefficient (ICC) maps showed the spatial distribution of the repeatability metrics, informing on the feasibility limits of the method. In conclusion, results show an overall consistency of the estimated physiological parameters with literature reports and a satisfactory level of repeatability considering the higher spatial sensitivity compared to other MRI methods, with varied performance depending on the specific parameter under analysis, on the spatial resolution considered and on the study design.

Machine learning has considerably improved medical image analysis in the past years. Although data-driven approaches are intrinsically adaptive and thus, generic, they often do not perform the same way on data from different imaging modalities. In particular computed tomography (CT) data poses many challenges to medical image segmentation based on convolutional neural networks (CNNs), mostly due to the broad dynamic range of intensities and the varying number of recorded slices of CT volumes. In this paper, we address these issues with a framework that adds domain-specific data preprocessing and augmentation to state-of-the-art CNN architectures. Our major focus is to stabilise the prediction performance over samples as a mandatory requirement for use in automated and semi-automated workflows in the clinical environment. To validate the architecture-independent effects of our approach we compare a neural architecture based on dilated convolutions for parallel multi-scale processing (a modified Mixed-Scale Dense Network: MS-D Net) to traditional scaling operations (a modified U-Net). Finally, we show that an ensemble model combines the strengths across different individual methods. Our framework is simple to implement into existing deep learning pipelines for CT analysis. It performs well on a range of tasks such as liver and kidney segmentation, without significant differences in prediction performance on strongly differing volume sizes and varying slice thickness. Thus our framework is an essential step towards performing robust segmentation of unknown real-world samples.

This study aims to map the acute effects of caffeine ingestion on grey matter oxygen metabolism and haemodynamics with a novel MRI method. Sixteen healthy caffeine consumers (8 males, age=24.7±5.1) were recruited to this randomised, double-blind, placebo-controlled study. Each participant was scanned on two days before and after the delivery of an oral caffeine (250mg) or placebo capsule. Our measurements were obtained with a newly proposed estimation approach applied to data from a dual calibration fMRI experiment that uses hypercapnia and hyperoxia to modulate brain blood flow and oxygenation. Estimates were based on a forward model that describes analytically the contributions of cerebral blood flow (CBF) and of the measured end-tidal partial pressures of CO2 and O2 to the acquired dual-echo GRE signal. The method allows the estimation of grey matter maps of: oxygen extraction fraction (OEF), CBF, CBF-related cerebrovascular reactivity (CVR) and cerebral metabolic rate of oxygen consumption (CMRO2). Other estimates from a multi inversion time ASL acquisition (mTI-ASL), salivary samples of the caffeine concentration and behavioural measurements are also reported. We observed significant differences between caffeine and placebo on average across grey matter, with OEF showing an increase of 15.6% (SEM±4.9%, p<0.05) with caffeine, while CBF and CMRO2 showed differences of −30.4% (SEM±1.6%, p<0.01) and −18.6% (SEM±2.9%, p<0.01) respectively with caffeine administration. The reduction in oxygen metabolism found is somehow unexpected, but consistent with a hypothesis of decreased energetic demand, supported by previous electrophysiological studies reporting reductions in spectral power with EEG. Moreover the maps of the physiological parameters estimated illustrate the spatial distribution of changes across grey matter enabling us to localise the effects of caffeine with voxel-wise resolution. CBF changes were widespread as reported by previous findings, while changes in OEF were found to be more restricted, leading to unprecedented mapping of significant CMRO2 reductions mainly in frontal gyrus, parietal and occipital lobes. In conclusion, we propose the estimation framework based on our novel forward model with a dual calibrated fMRI experiment as a viable MRI method to map the effects of drugs on brain oxygen metabolism and haemodynamics with voxel-wise resolution. •Randomised, double-blind, placebo-controlled drug study with caffeine, (N=16).•Set of physiological parameters acquired with dual calibrated fMRI, experiment and respiratory tasks.•Additional haemodynamic measurements acquired with a mTI, experiment.•Significant overall OEF increase with caffeine, while CBF and CMRO2, decreased.•Calculated maps localise the acute effects of caffeine across grey matter.

Several techniques have been proposed to estimate relative changes in cerebral metabolic rate of oxygen consumption (CMRO2) by exploiting combined BOLD fMRI and cerebral blood flow data in conjunction with hypercapnic or hyperoxic respiratory challenges. More recently, methods based on respiratory challenges that include both hypercapnia and hyperoxia have been developed to assess absolute CMRO2, an important parameter for understanding brain energetics. In this paper, we empirically optimize a previously presented “original calibration model” relating BOLD and blood flow signals specifically for the estimation of oxygen extraction fraction (OEF) and absolute CMRO2. To do so, we have created a set of synthetic BOLD signals using a detailed BOLD signal model to reproduce experiments incorporating hypercapnic and hyperoxic respiratory challenges at 3T. A wide range of physiological conditions was simulated by varying input parameter values (baseline cerebral blood volume (CBV0), baseline cerebral blood flow (CBF0), baseline oxygen extraction fraction (OEF0) and hematocrit (Hct)). From the optimization of the calibration model for estimation of OEF and practical considerations of hypercapnic and hyperoxic respiratory challenges, a new “simplified calibration model” is established which reduces the complexity of the original calibration model by substituting the standard parameters α and β with a single parameter θ. The optimal value of θ is determined (θ=0.06) across a range of experimental respiratory challenges. The simplified calibration model gives estimates of OEF0 and absolute CMRO2 closer to the true values used to simulate the experimental data compared to those estimated using the original model incorporating literature values of α and β. Finally, an error propagation analysis demonstrates the susceptibility of the original and simplified calibration models to measurement errors and potential violations in the underlying assumptions of isometabolism. We conclude that using the simplified calibration model results in a reduced bias in OEF0 estimates across a wide range of potential respiratory challenge experimental designs. •Analysis of a calibration model of the BOLD signal through a simulation study•Optimization of the calibration model based on the accuracy of OEF0 estimates•New simplified calibration model proposed and shown to improve results accuracy•Error propagation analysis performed on both original and simplified model

This study aims to map the acute effects of caffeine ingestion on grey matter oxygen metabolism and haemodynamics with a novel MRI method. Sixteen healthy caffeine consumers (8 males, age=24.7±5.1) were recruited to this randomised, double-blind, placebo-controlled study. Each participant was scanned on two days before and after the delivery of an oral caffeine (250mg) or placebo capsule. Our measurements were obtained with a newly proposed estimation approach applied to data from a dual calibration fMRI experiment that uses hypercapnia and hyperoxia to modulate brain blood flow and oxygenation. Estimates were based on a forward model that describes analytically the contributions of cerebral blood flow (CBF) and of the measured end-tidal partial pressures of CO and O to the acquired dual-echo GRE signal. The method allows the estimation of grey matter maps of: oxygen extraction fraction (OEF), CBF, CBF-related cerebrovascular reactivity (CVR) and cerebral metabolic rate of oxygen consumption (CMRO ). Other estimates from a multi inversion time ASL acquisition (mTI-ASL), salivary samples of the caffeine concentration and behavioural measurements are also reported. We observed significant differences between caffeine and placebo on average across grey matter, with OEF showing an increase of 15.6% (SEM±4.9%, p<0.05) with caffeine, while CBF and CMRO showed differences of -30.4% (SEM±1.6%, p<0.01) and -18.6% (SEM±2.9%, p<0.01) respectively with caffeine administration. The reduction in oxygen metabolism found is somehow unexpected, but consistent with a hypothesis of decreased energetic demand, supported by previous electrophysiological studies reporting reductions in spectral power with EEG. Moreover the maps of the physiological parameters estimated illustrate the spatial distribution of changes across grey matter enabling us to localise the effects of caffeine with voxel-wise resolution. CBF changes were widespread as reported by previous findings, while changes in OEF were found to be more restricted, leading to unprecedented mapping of significant CMRO reductions mainly in frontal gyrus, parietal and occipital lobes. In conclusion, we propose the estimation framework based on our novel forward model with a dual calibrated fMRI experiment as a viable MRI method to map the effects of drugs on brain oxygen metabolism and haemodynamics with voxel-wise resolution.

The quantification of brain activity has been one of the main goals of neuroimaging since the earliest applications. In functional magnetic resonance imaging (fMRI) such an aim has been pursued indirectly by studying changes of the blood oxygenation dependent signal triggered by alterations in blood flow following changes in energy metabolism. Such approach is limited because of the complex relationship between the vascular and neural systems in brain tissue. Therefore methods have been proposed to assess oxygen metabolism, which directly underlies energy supply to brain tissue and therefore brain activity. Investigating existing and novel MRI methods, the thesis aims to improve the assessment of oxygen metabolism for a fully quantitative measurement of this biomarker. A simulation study has been carried out to optimise one of the mathematical (fMRI calibration) models used to relate the measured signal to the underlying physiology. As a result we are able to define a new model, less complex and more accurate for estimation of oxygen extraction fraction. Following this, an estimation approach recently developed in our centre is applied to carbon dioxide and oxygen calibrated fMRI data in an experimental setting firstly for a repeatability study and then for a drug study looking at the acute effects of caffeine on brain metabolism and haemodynamics. The precision of the novel approach shows values consistent with previous methods, but with much higher spatial resolution. Exploiting this, acute caffeine effects are characterized with a voxel-wise level of detail, showing results consistent with literature electrophysiological findings. Finally, an innovative method for estimating oxygen extraction fraction, based on velocity spectral imaging and estimation of transverse relaxation time, is introduced and tested at a proof-of-concept level. The performance and limits are examined through simulation and experimentation, suggesting that it might be a viable alternative to the calibration techniques previously introduced.

Machine Learning has considerably improved medical image analysis in the past years. Although data-driven approaches are intrinsically adaptive and thus, generic, they often do not perform the same way on data from different imaging modalities. In particular Computed tomography (CT) data poses many challenges to medical image segmentation based on convolutional neural networks (CNNs), mostly due to the broad dynamic range of intensities and the varying number of recorded slices of CT volumes. In this paper, we address these issues with a framework that combines domain-specific data preprocessing and augmentation with state-of-the-art CNN architectures. The focus is not limited to optimise the score, but also to stabilise the prediction performance since this is a mandatory requirement for use in automated and semi-automated workflows in the clinical environment. The framework is validated with an architecture comparison to show CNN architecture-independent effects of our framework functionality. We compare a modified U-Net and a modified Mixed-Scale Dense Network (MS-D Net) to compare dilated convolutions for parallel multi-scale processing to the U-Net approach based on traditional scaling operations. Finally, we propose an ensemble model combining the strengths of different individual methods. The framework performs well on a range of tasks such as liver and kidney segmentation, without significant differences in prediction performance on strongly differing volume sizes and varying slice thickness. Thus our framework is an essential step towards performing robust segmentation of unknown real-world samples.