A perinatal stress calculator for the neonatal intensive care unit: an unobtrusive approach

OBJECTIVE

Early experience of pain and stress in the neonatal intensive care unit is known to have an effect on the neurodevelopment of the infant. However, an automated method to quantify the procedural pain or perinatal stress in premature patients does not exist.

APPROACH

In the current study, EEG and ECG data were collected for more than 3 hours from 136 patients in order to quantify stress exposure. Specifically, features extracted from the EEG and heart-rate variability in both quiet and non-quiet sleep segments were used to develop a subspace linear-discriminant analysis stress classifier.

MAIN RESULTS

The main novelty of the study lies in the absence of intrusive methods or pain elicitation protocols to develop the stress classifier. Three main findings can be reported. First, we developed different stress classifiers for the different age groups and stress intensities, obtaining an area under the curve in the range [0.78-0.93] for non-quiet sleep and [0.77-0.96] for quiet sleep. Second, a dysmature EEG was found in patients under stress. Third, an enhanced cortical connectivity and increased brain-heart communication was correlated with a higher stress load, while the autonomic activity did not seem to be associated to stress exposure.

SIGNIFICANCE

The results shed a light on the pain and stress processing in preterm neonates, suggesting that software tools to investigate dysmature EEG might be helpful to assess stress load in premature patients. These results could be the foundation to assess the impact of stress on infants' development and to tune preventive care.

Monitoring of Brain Hemodynamics Coupling in Neonates using Updated Tensor Decompositions

In this paper we explore the use of updated tensor decompositions for the monitoring of brain hemodynamics in neonates. For this study, we used concomitant measurements of heart rate, mean arterial blood pressure, arterial oxygen saturation, EEG, and brain oxygenation - measured using near-infrared spectroscopy. These measurements were obtained from 22 neonates undergoing an INSURE procedure (INtubation, SURfactant and Extubation) and sedation using propofol. To develop the monitoring framework using tensors, we used radial basis kernel function (RBF) to construct a similarity matrix for consecutive segments of the signals. These matrices were concatenated forming a tensor. Updating canonical polyadic decomposition was used to evaluate the impact of propofol in the coupling between the different signals. Results indicate, as previously reported, a drop in the interaction between signals due to propofol administration. This shows that tensor decompositions can be useful in order to monitor the coupling between different physiological signals.

The implementation of an apnea-based perinatal stress calculator

Early life stress in the neonatal intensive care unit (NICU) predisposes premature infants to adverse health outcomes. Although those patients experience frequent apneas and sleep-wake disturbances during their hospital stay, clinicians still rely on clinical scales to assess pain and stress burden. This study addresses the relationship between stress and apneic spells in NICU patients to implement an automatic stress detector. EEG, ECG and SpO₂ were recorded from 40 patients for at least 3 hours and the stress burden was assessed using the Leuven Pain Scale. Different logistic regression models were designed to detect the presence or the absence of stress based on the signals reactivity to each apneic spell. The classification shows that stress can be detected with an area under the curve of 0.94 and a misclassification error of 19.23%. These results were obtained via SpO₂ dips and EEG regularity. These findings suggest that stress deepens the physiological reaction to apneas, which could ultimately impact the neurological and behavioral development.

Non-invasive assessment of disease progression and neuroprotective effects of dietary coconut oil supplementation in the ALS SOD1G93A mouse model: A 1H-magnetic resonance spectroscopic study

Amyotrophic Lateral Sclerosis (ALS) is an incurable neurodegenerative disease primarily characterized by progressive degeneration of motor neurons in the motor cortex, brainstem and spinal cord. Due to relatively fast progression of ALS, early diagnosis is essential for possible therapeutic intervention and disease management. To identify potential diagnostic markers, we investigated age-dependent effects of disease onset and progression on regional neurochemistry in the SOD1^G93A ALS mouse model using localized in vivo magnetic resonance spectroscopy (MRS). We focused mainly on the brainstem region since brainstem motor nuclei are the primarily affected regions in SOD1^G93A mice and ALS patients. In addition, metabolite profiles of the motor cortex were also assessed. In the brainstem, a gradual decrease in creatine levels were detected starting from the pre-symptomatic age of 70 days postpartum. During the early symptomatic phase (day 90), a significant increase in the levels of the inhibitory neurotransmitter γ- aminobutyric acid (GABA) was measured. At later time points, alterations in the form of decreased NAA, glutamate, glutamine and increased myo-inositol were observed. Also, decreased glutamate, NAA and increased taurine levels were seen at late stages in the motor cortex. A proof-of-concept (PoC) study was conducted to assess the effects of coconut oil supplementation in SOD^G93A mice. The PoC revealed that the coconut oil supplementation together with the regular diet delayed disease symptoms, enhanced motor performance, and prolonged survival in the SOD1^G93A mouse model. Furthermore, MRS data showed stable metabolic profile at day 120 in the coconut oil diet group compared to the group receiving a standard diet without coconut oil supplementation. In addition, a positive correlation between survival and the neuronal marker NAA was found. To the best of our knowledge, this is the first study that reports metabolic changes in the brainstem using in vivo MRS and effects of coconut oil supplementation as a prophylactic treatment in SOD1^G93A mice.

Automatic quiet sleep detection based on multifractality in preterm neonates: Effects of maturation

This study investigates the multifractal formalism framework for quiet sleep detection in preterm babies. EEG recordings from 25 healthy preterm infants were used in order to evaluate the performance of multifractal measures for the detection of quiet sleep. Results indicate that multifractal analysis based on wavelet leaders is able to identify quiet sleep epochs, but the classifier performances seem to be highly affected by the infant's age. In particular, from the developed classifiers, the lowest area under the curve (AUC) has been obtained for EEG recordings at very young age (≤ 31 weeks post-menstrual age), and the maximum at full-term age (≥ 37 weeks post-menstrual age). The improvement in classification performances can be due to a change in the multifractality properties of neonatal EEG during the maturation of the infant, which makes the EEG sleep stages more distinguishable.

Improved neonatal seizure detection using adaptive learning

In neonatal intensive care units performing continuous EEG monitoring, there is an unmet need for around-the-clock interpretation of EEG, especially for recognizing seizures. In recent years, a few automated seizure detection algorithms have been proposed. However, these are suboptimal in detecting brief-duration seizures (<; 30s), which frequently occur in neonates with severe neurological problems. Recently, a multi-stage neonatal seizure detector, composed of a heuristic and a data-driven classifier was proposed by our group and showed improved detection of brief seizures. In the present work, we propose to add a third stage to the detector in order to use feedback of the Clinical Neurophysiologist and adaptively retune a threshold of the second stage to improve the performance of detection of brief seizures. As a result, the false alarm rate (FAR) of the brief seizure detections decreased by 50% and the positive predictive value (PPV) increased by 18%. At the same time, for all detections, the FAR decreased by 35% and PPV increased by 5% while the good detection rate remained unchanged.

Comparison of unsupervised classification methods for brain tumor segmentation using multi-parametric MRI

Tumor segmentation is a particularly challenging task in high-grade gliomas (HGGs), as they are among the most heterogeneous tumors in oncology. An accurate delineation of the lesion and its main subcomponents contributes to optimal treatment planning, prognosis and follow-up. Conventional MRI (cMRI) is the imaging modality of choice for manual segmentation, and is also considered in the vast majority of automated segmentation studies. Advanced MRI modalities such as perfusion-weighted imaging (PWI), diffusion-weighted imaging (DWI) and magnetic resonance spectroscopic imaging (MRSI) have already shown their added value in tumor tissue characterization, hence there have been recent suggestions of combining different MRI modalities into a multi-parametric MRI (MP-MRI) approach for brain tumor segmentation. In this paper, we compare the performance of several unsupervised classification methods for HGG segmentation based on MP-MRI data including cMRI, DWI, MRSI and PWI. Two independent MP-MRI datasets with a different acquisition protocol were available from different hospitals. We demonstrate that a hierarchical non-negative matrix factorization variant which was previously introduced for MP-MRI tumor segmentation gives the best performance in terms of mean Dice-scores for the pathologic tissue classes on both datasets.

•

Unsupervised classification algorithms are applied for brain tumor segmentation on multi-parametric MRI datasets.

•

Reported mean Dice-scores are in the range of state-of-the-art segmentation algorithms.

•

Hierarchical NMF obtained the best segmentation results in terms of mean Dice-scores for most of the tissue classes.

Does ignoring clustering in multicenter data influence the performance of prediction models? A simulation study

Clinical risk prediction models are increasingly being developed and validated on multicenter datasets. In this article, we present a comprehensive framework for the evaluation of the predictive performance of prediction models at the center level and the population level, considering population-averaged predictions, center-specific predictions, and predictions assuming an average random center effect. We demonstrated in a simulation study that calibration slopes do not only deviate from one because of over- or underfitting of patterns in the development dataset, but also as a result of the choice of the model (standard versus mixed effects logistic regression), the type of predictions (marginal versus conditional versus assuming an average random effect), and the level of model validation (center versus population). In particular, when data is heavily clustered (ICC 20%), center-specific predictions offer the best predictive performance at the population level and the center level. We recommend that models should reflect the data structure, while the level of model validation should reflect the research question.

Nonnegative Canonical Polyadic Decomposition for Tissue-Type Differentiation in Gliomas

Magnetic resonance spectroscopic imaging (MRSI) reveals chemical information that characterizes different tissue types in brain tumors. Blind source separation techniques are used to extract the tissue-specific profiles and their corresponding distribution from the MRSI data. We focus on automatic detection of the tumor, necrotic and normal brain tissue types by constructing a 3D MRSI tensor from in vivo 2D-MRSI data of individual glioma patients. Nonnegative canonical polyadic decomposition (NCPD) is applied to the MRSI tensor to differentiate various tissue types. An in vivo study shows that NCPD has better performance in identifying tumor and necrotic tissue type in glioma patients compared to previous matrix-based decompositions, such as nonnegative matrix factorization and hierarchical nonnegative matrix factorization.

The suppression curve as a quantitative approach for measuring brain maturation in preterm infants

OBJECTIVES

We apply the suppression curve (SC) as an automated approach to describe the maturational change in EEG discontinuity in preterm infants. This method allows to define normative values of interburst intervals (IBIs) at different postmenstrual ages (PMA).

METHODS

Ninety-two multichannel EEG recordings from 25 preterm infants (born ⩽32weeks) with normal developmental outcome at 9months, were first analysed using the Line Length method, an established method for burst detection. Subsequently, the SC was defined as the 'level of EEG discontinuity'. The mean and the standard deviation of the SC, as well as the IBIs from each recording were calculated and correlated with PMA.

RESULTS

Over the course of development, there is a decrease in EEG discontinuity with a strong linear correlation between the mean SC and PMA till 34weeks. From 30weeks PMA, differences between discontinuous and continuous EEG become smaller, which is reflected by the decrease of the standard deviation of the SC. IBIs are found to have a significant correlation with PMA.

CONCLUSIONS

Automated detection of individual maturational changes in EEG discontinuity is possible with the SC. These changes include more continuous tracing, less amplitude differences and shorter suppression periods, reflecting development of the vigilance states.

SIGNIFICANCE

The suppression curve facilitates automated assessment of EEG maturation. Clinical applicability is straight forward since values for IBIs according to PMA are generated automatically.

Early development of synchrony in cortical activations in the human

•

We study the early development of cortical activations synchrony index (ASI).

•

Cortical activations become increasingly synchronized during the last trimester.

•

Interhemispheric synchrony increases more than intrahemispheric synchrony.

•

Our EEG metric ASI can be directly translated to experimental animal studies.

•

ASI holds promise as an early functional biomarker of brain networks.

Early intermittent cortical activity is thought to play a crucial role in the growth of neuronal network development, and large scale brain networks are known to provide the basis for higher brain functions. Yet, the early development of the large scale synchrony in cortical activations is unknown. Here, we tested the hypothesis that the early intermittent cortical activations seen in the human scalp EEG show a clear developmental course during the last trimester of pregnancy, the period of intensive growth of cortico-cortical connections. We recorded scalp EEG from altogether 22 premature infants at post-menstrual age between 30 and 44 weeks, and the early cortical synchrony was quantified using recently introduced activation synchrony index (ASI). The developmental correlations of ASI were computed for individual EEG signals as well as anatomically and mathematically defined spatial subgroups. We report two main findings. First, we observed a robust and statistically significant increase in ASI in all cortical areas. Second, there were significant spatial gradients in the synchrony in fronto-occipital and left-to-right directions. These findings provide evidence that early cortical activity is increasingly synchronized across the neocortex. The ASI-based metrics introduced in our work allow direct translational comparison to in vivo animal models, as well as hold promise for implementation as a functional developmental biomarker in future research on human neonates.

Tensor based tumor tissue type differentiation using magnetic resonance spectroscopic imaging

Magnetic resonance spectroscopic imaging (MRSI) has the potential to characterise different tissue types in brain tumors. Blind source separation techniques are used to extract the specific tissue profiles and their corresponding distribution from the MRSI data. A 3-dimensional MRSI tensor is constructed from in vivo 2D-MRSI data of individual tumor patients. Non-negative canonical polyadic decomposition (NCPD) with common factor in mode-1 and mode-2 and l(1) regularization on mode-3 is applied on the MRSI tensor to differentiate various tissue types. Initial in vivo study shows that NCPD has better performance in identifying tumor and necrotic tissue type in high grade glioma patients compared to previous matrix-based decompositions, such as non-negative matrix factorization and hierarchical non-negative matrix factorization.

Classifying the auditory P300 using mobile EEG recordings without calibration phase

One of the major drawbacks in mobile EEG Brain Computer Interfaces (BCI) is the need for subject specific training data to train a classifier. By removing the need for supervised classification and calibration phase, new users could start immediate interaction with a BCI. We propose a solution to exploit the structural difference by means of canonical polyadic decomposition (CPD) for three-class auditory oddball data without the need for subject-specific information. We achieve this by adding average event-related-potential (ERP) templates to the CPD model. This constitutes a novel similarity measure between single-trial pairs and known-templates, which results in a fast and interpretable classifier. These results have similar accuracy to those of the supervised and cross-validated stepwise LDA approach but without the need for having subject-dependent data. Therefore the described CPD method has a significant practical advantage over the traditional and widely used approach.

Improvement of an automated neonatal seizure detector using a post-processing technique

Visual recognition of neonatal seizures during continuous EEG monitoring in neonatal intensive care units (NICUs) is labor-intensive, has low inter-rater agreement and requires special expertise that is not available around the clock. Development of an accurate automated seizure detection system with a low false alarm rate will support clinical decision making and alleviate significantly the workload. However, this is an ongoing difficult challenge for engineers as the neonatal EEG signal is non-stationary and often includes complex patterns of seizures and artifacts. In this study, we show an improvement of our previously developed neonatal seizure detector (developed using heuristic if-then rules). In order to improve the detection accuracy, mean phase coherence as a new feature is used to characterize artifacts and also support vector machine is applied to perform the post-processing step to remove false detections. As a result, the false alarm rate drops 42% (from 2.6 h(-1) to 1.5 h(-1)), whereas the good detection rate reduces only by 4%.

A simulation study of sample size demonstrated the importance of the number of events per variable to develop prediction models in clustered data

OBJECTIVES

This study aims to investigate the influence of the amount of clustering [intraclass correlation (ICC) = 0%, 5%, or 20%], the number of events per variable (EPV) or candidate predictor (EPV = 5, 10, 20, or 50), and backward variable selection on the performance of prediction models.

STUDY DESIGN AND SETTING

Researchers frequently combine data from several centers to develop clinical prediction models. In our simulation study, we developed models from clustered training data using multilevel logistic regression and validated them in external data.

RESULTS

The amount of clustering was not meaningfully associated with the models' predictive performance. The median calibration slope of models built in samples with EPV = 5 and strong clustering (ICC = 20%) was 0.71. With EPV = 5 and ICC = 0%, it was 0.72. A higher EPV related to an increased performance: the calibration slope was 0.85 at EPV = 10 and ICC = 20% and 0.96 at EPV = 50 and ICC = 20%. Variable selection sometimes led to a substantial relative bias in the estimated predictor effects (up to 118% at EPV = 5), but this had little influence on the model's performance in our simulations.

CONCLUSION

We recommend at least 10 EPV to fit prediction models in clustered data using logistic regression. Up to 50 EPV may be needed when variable selection is performed.

Motor Unit Tracking Using High Density Surface Electromyography (HDsEMG) . Automated Correction of Electrode Displacement Errors

INTRODUCTION

This article is part of the Focus Theme of Methods of Information in Medicine on "Biosignal Interpretation: Ad-vanced Methods for Neural Signals and Images".

OBJECTIVES

The study discusses a technique to automatically correct for effects of electrode grid displacement across serial surface EMG measurements with high-density electrode arrays (HDsEMG). The goal is to match motor unit signatures from subsequent measurements and by this, achieve automated motor unit tracking.

METHODS

Test recordings of voluntary muscle contractions using HDsEMG were performed on three healthy individuals. Electrode grid displacements were mimicked in repeated recordings while measuring the exact position of the grid. A concept of accounting for translational and rotational displacements by making the projection of the recorded motor unit action potentials is first introduced. Then, this concept was tested for the performed measurements attempting the automated matching of the similar motor unit action potentials across different trials.

RESULTS

The ability to perform automated correction (projection) of the isolated motor unit action potentials was first shown using large angular displacements. Then, for accidental (small) displacements of the recording grid, the ability to automatically track motor units across different measurement trials was shown. It was possible to track 10 -15% of identified motor units.

CONCLUSIONS

This proof of concept study demonstrates an automated correction allowing the identification of an increased number of same motor unit action potentials across different measurements. By this, great potential is demonstrated for assisting motor unit tracking studies, indicating that otherwise electrode displacements cannot always be precisely described.

Different quantification algorithms may lead to different results: a comparison using proton MRS lipid signals

Proton magnetic resonance spectroscopy (MRS) is a sensitive method for investigating the biochemical compounds in a tissue. The interpretation of the data relies on the quantification algorithms applied to MR spectra. Each of these algorithms has certain underlying assumptions and may allow one to incorporate prior knowledge, which could influence the quality of the fit. The most commonly considered types of prior knowledge include the line-shape model (Lorentzian, Gaussian, Voigt), knowledge of the resonating frequencies, modeling of the baseline, constraints on the damping factors and phase, etc. In this article, we study whether the statistical outcome of a biological investigation can be influenced by the quantification method used. We chose to study lipid signals because of their emerging role in the investigation of metabolic disorders. Lipid spectra, in particular, are characterized by peaks that are in most cases not Lorentzian, because measurements are often performed in difficult body locations, e.g. in visceral fats close to peristaltic movements in humans or very small areas close to different tissues in animals. This leads to spectra with several peak distortions. Linear combination of Model spectra (LCModel), Advanced Method for Accurate Robust and Efficient Spectral fitting (AMARES), quantitation based on QUantum ESTimation (QUEST), Automated Quantification of Short Echo-time MRS (AQSES)-Lineshape and Integration were applied to simulated spectra, and area under the curve (AUC) values, which are proportional to the quantity of the resonating molecules in the tissue, were compared with true values. A comparison between techniques was also carried out on lipid signals from obese and lean Zucker rats, for which the polyunsaturation value expressed in white adipose tissue should be statistically different, as confirmed by high-resolution NMR measurements (considered the gold standard) on the same animals. LCModel, AQSES-Lineshape, QUEST and Integration gave the best results in at least one of the considered groups of simulated or in vivo lipid signals. These outcomes highlight the fact that quantification methods can influence the final result and its statistical significance.

Early respiratory dysfunction as a biomarker for epileptic encephalopathy

OBJECTIVES

West syndrome is an epileptic encephalopathy starting in infancy with almost continuous interictal epileptic activity, so-called hypsarrhythmia pattern, and therefore is an interesting model for investigating the effect of interictal epileptic activity on autonomic function. It is known that autonomic dysfunction contributes to morbidity and mortality in epilepsy. Our aim is to investigate the effect of interictal epileptic activity in West syndrome on respiratory control.

MATERIALS AND METHODS

Interictal single-lead ECG signals were extracted from 24-h video-EEG recordings in 10 children suffering from West syndrome and 14 control subjects. RR interval time series were calculated, and respiration was derived from the ECG signal. ECG-derived respiration (EDR) signals were computed and time and frequency domain parameters were extracted to characterize the respiration pattern.

RESULTS

In time domain, the standard deviation of the EDR signal is significantly lower in patients with West syndrome compared to control subjects. This finding is an indication of a less variable respiratory rate. In frequency domain, we analyzed the mean power spectrum for the EDR. In patients with West syndrome, there is more activity at the lower frequencies considered to be a risk factor for apneas. Second, there is an attenuated peak at the higher frequency band where normal respiratory rate is to be found, indicating an abnormal breathing pattern.

CONCLUSIONS

Our results show that there is a clear dysfunction in autonomic respiratory control in patients with West syndrome, in between the typical ictal epileptic spasms, compared to control subjects. Respiration is more fixed and contains a higher risk of apneas.

Preprocessing by means of subspace projections for continuous Cerebral Autoregulation assessment using NIRS

Cerebral Autoregulation (CA) refers to the capability of the brain to maintain a more or less stable cerebral blood flow (CBF), despite the changes in blood perfusion. Monitoring this mechanism is of vital importance, especially in neonates, in order to prevent damage due to ischemia or hemorrhage. In clinical practice near-infrared spectroscopy (NIRS) measurements are used as a surrogate measurement for CBF. However, NIRS signals are highly dependent on the variations in arterial oxygen saturation (SaO2). Therefore, only segments with relatively constant SaO2 are used for CA assessment; which limits the possibilities of the use of NIRS for online monitoring. In this paper we propose the use of subspace projections to subtract the influence of SaO2 from NIRS measurements. Since this approach will be used in an online monitoring system, this preprocessing is carried out in a window-by-window framework. However, the use of subspace projections in consecutive segments produces discontinuities; we propose a methodology to reduce these effects. Obtained results indicate that the proposed method reduces the effect of discontinuities between consecutive segments. In addition, this methodology is able to subtract the influence of SaO2 from NIRS measurements. This approach facilitates the introduction of NIRS for online CA assessment.

Predicting successful vaginal birth after Cesarean section using a model based on Cesarean scar features examined by transvaginal sonography

OBJECTIVE

To develop a model to predict the success of a trial of vaginal birth after Cesarean section (VBAC) based on sonographic measurements of Cesarean section (CS) scar features, demographic variables and previous obstetric history.

METHODS

We used transvaginal sonography (TVS) to examine the CS scar of 320 consecutive pregnant women. TVS was carried out at 11-13, 19-21 and 34-36 weeks' gestation and prospective measurements of the scar were recorded at each visit according to a defined protocol. A logistic regression model to predict success of VBAC was developed for those patients with a visible scar on ultrasound and only one previous CS. The model was evaluated using bootstrap validation.

RESULTS

There were 131 women with one previous CS and a visible scar, of whom 10 underwent CS prior to labor and were excluded from analysis. Successful VBAC was achieved in 74/121 (61%) of the remaining cases. The prediction model developed was based on patient age, previous history of VBAC, residual myometrial thickness (RMT) and the change in RMT from the first to the second trimester (ΔRMT). The internally validated area under the receiver-operating characteristics curve was 0.62 when measurements of RMT and ΔRMT were excluded, but 0.94 when scar information was incorporated into the model.

CONCLUSION

Ultrasound measurements of CS scar, namely RMT and the change in RMT from the first to the second trimester of pregnancy, when incorporated into a mathematical model, can predict accurately a successful trial of labor in patients with one previous CS.

Unilocular adnexal cysts with papillary projections but no other solid components: is there a diagnostic method that can classify them reliably as benign or malignant before surgery?

OBJECTIVES

To develop a logistic regression model for discrimination between benign and malignant unilocular solid cysts with papillary projections but no other solid components, and to compare its diagnostic performance with that of subjective evaluation of ultrasound findings (subjective assessment), CA  125 and the risk of malignancy index (RMI).

METHODS

Among the 3511 adnexal masses in the International Ovarian Tumor Analysis (IOTA) database there were 252 (7%) unilocular solid cysts with papillary projections but no other solid components ('unilocular cysts with papillations'). All had been examined with transvaginal ultrasound using the IOTA standardized research protocol. The ultrasound examiner had also classified each mass as certainly or probably benign, unclassifiable, or certainly or probably malignant. A logistic regression model to discriminate between benignity and malignancy was developed for all unilocular cysts with papillations (175 tumors in the training set and 77 in the test set) and for unilocular cysts with papillations for which the ultrasound examiner was not certain about benignity/malignancy (113 tumors in the training set and 53 in the test set). The gold standard was the histological diagnosis of the surgically removed adnexal mass.

RESULTS

A model containing six variables was developed for all unilocular cysts with papillations. The model had an area under the receiver-operating characteristics curve (AUC) on the test set of 0.83 (95% CI, 0.74-0.93). The optimal risk cut-off, as defined on the training set (0.35), resulted in sensitivity 69% (20/29), specificity 79% (38/48), positive likelihood ratio (LR +) 3.31 and negative likelihood ratio (LR-) 0.39 on the test set. The corresponding values for subjective assessment when using the ultrasound examiner's dichotomous classification of the mass as benign or malignant were 97% (28/29), 79% (38/48), 4.63 and 0.04. A model containing four variables was developed for unilocular cysts with papillations for which the ultrasound examiner was not certain about benignity/malignancy. The model had an AUC of 0.74 (95% CI, 0.60-0.88) on the test set. The optimal risk cut-off of the model, as defined on the training set (0.30), resulted in sensitivity 57% (12/21), specificity 78% (25/32), LR + 2.61 and LR- 0.55 on the test set. The corresponding values for subjective assessment were 95% (20/21), 78% (25/32), 4.35 and 0.06. CA  125 and RMI had virtually no diagnostic ability.

CONCLUSIONS

Even though logistic regression models to predict malignancy in unilocular cysts with papillations can be developed, they have at most moderate performance and are not superior to subjective assessment for discrimination between benignity and malignancy.

Does the presence of a Caesarean section scar affect implantation site and early pregnancy outcome in women attending an early pregnancy assessment unit?

STUDY QUESTION

Are there any differences in the location and distance to the internal cervical ostium of the implantation site of the intrauterine gestation sacs, early pregnancy symptoms and pregnancy outcome at 12 weeks gestation between women with and without a previous Caesarean section (CS)?

SUMMARY ANSWER

The presence of a CS scar affects the site of implantation, and the distance between implantation site and the scar is related to the risk of spontaneous abortion. WHAT IS KNOWN ALREADY?: Little is known about the impact of a CS scar on implantation other than the risk of Caesarean scar pregnancy (CSP). Furthermore, there is a paucity of information on how the proximity of implantation to the scar impacts on pregnancy outcome in the first trimester.

STUDY DESIGN, SIZE, AND DURATION

A prospective cohort study conducted over 15 months in the early pregnancy unit of a London Teaching Hospital. Three hundred and eighty women underwent a transvaginal scan at 6-11 weeks of gestation. A total of 170 women had undergone ≥1 CS, and 210 women had no history of CS.

PARTICIPANTS/MATERIALS, SETTING, METHODS

The 380 women were recruited as consecutive non-selected cases. The relationship between the implanted sac and the CS scar was assessed by quantifiable measures and by subjective impression. Logistic regression analysis was used to determine the influence of the presence of a CS scar on pregnancy outcome. The final outcome of the study was the viability of the pregnancy at 12 weeks.

MAIN RESULTS AND THE ROLE OF CHANCE

Implantation was most frequently posterior (53%) in the CS group and fundal in the non-CS group (42%). Gestation sac implantation was 8.7 mm lower in the CS group (95% confidence interval (CI) 6.7-10.7, P < 0.0001). Presenting complaints differed in women with and without a previous CS (P = 0.0009). More frequent vaginal bleeding [73 versus 55%, difference -18, 95% CI (-27 to -8%] yet no clearly increased spontaneous abortion rates were noted in the CS group compared with the non-CS group (adjusted odds ratio = 1.1, 95% CI 0.6-1.9, P = 0.74). Subjective impression showed that in eight cases the implantation site crossed the scar, seven of which resulted in spontaneous abortion, while the remaining case survived to term complicated by placenta praevia and post-partum haemorrhage. The subjective impression of the examiner was supported by the measurements of distance between implantation site and CS scar.

LIMITATIONS, REASONS FOR CAUTION

A weakness of the study is the lack of a reference technique to verify the location of implantation.

WIDER IMPLICATIONS OF THE FINDINGS

This study adds further support to the hypothesis that the presence of a CS on the uterus impacts on the implantation site of a future pregnancy. The possibility that the CS scar has an impact on the risk of spontaneous abortion should be further studied. Caution must be exercised when implantation occurs near to, and crosses, a CS scar as this is not always associated with the diagnosis of CSP. A potential limitation of the study is that we did not examine scar dimensions and morphology.

Risk of malignancy in unilocular cysts: a study of 1148 adnexal masses classified as unilocular cysts at transvaginal ultrasound and review of the literature

OBJECTIVES

The aim of this study was to estimate the rate of malignancy in adnexal lesions described as unilocular cysts at transvaginal ultrasound examination and to investigate if there are differences in clinical and ultrasound characteristics between benign and malignant unilocular cysts.

METHODS

A total of 3511 patients with an adnexal mass underwent transvaginal ultrasound examination between 1999 and 2007. Sonologists used the International Ovarian Tumor Analysis terms and definitions to describe their ultrasound findings. Only masses operated on within 120  days after the ultrasound examination were included in the analysis and the histopathological diagnosis of the mass was used as the gold standard.

RESULTS

Of the 3511 masses, 1148 (33%) were classified as unilocular cysts on ultrasound. Of these, 11 (0.96% (95% CI, 0.48-1.71)) were malignant. The malignancy rate was lower in premenopausal than in postmenopausal women: 0.54% (5/931; 95% CI, 0.17-1.25) vs. 2.76% (6/217; 95% CI, 1.02-5.92); P = 0.009. More patients with malignant unilocular cysts had a personal history of breast cancer (18% vs. 2%; P = 0.02) or ovarian cancer (18% vs 0.6%; P = 0.003). Hemorrhagic cyst contents on ultrasound were more common in malignant than in benign unilocular cysts (18% vs. 2%; P = 0.03). In seven of the 11 malignancies judged to be unilocular cysts at scan, papillary projections or other solid components were seen at macroscopic inspection of the surgical specimen.

CONCLUSIONS

The malignancy rate in surgically removed adnexal lesions judged to be unilocular cysts at transvaginal scan is c 1%. Postmenopausal status, personal history of breast or ovarian cancer and hemorrhagic cyst contents on ultrasound increase the risk of malignancy. To avoid misclassifying adnexal lesions as unilocular cysts at scan, it is important to scrutinize unilocular cysts for the presence of solid components.

Clinically oriented three-step strategy for assessment of adnexal pathology

OBJECTIVE

To determine the diagnostic performance of ultrasound-based simple rules, risk of malignancy index (RMI), two logistic regression models (LR1 and LR2) and real-time subjective assessment by experienced ultrasound examiners following the exclusion of masses likely to be judged as easy and 'instant' to diagnose by an ultrasound examiner, and to develop a new strategy for the assessment of adnexal pathology based on this.

METHODS

3511 patients with at least one persistent adnexal mass preoperatively underwent transvaginal ultrasonography to assess tumor morphology and vascularity. They were included in two consecutive prospective studies by the International Ovarian Tumor Analysis (IOTA) group: Phase 1 (1999-2005), development of the simple rules and logistic regression models LR1 and LR2, and Phase 2, a validation study (2005-2007).

RESULTS

Almost half of the cases (43%) were identified as 'instant' to diagnose on the basis of descriptors applied to the database. To assess diagnostic performance in the more difficult 'non-instant' masses, we used only Phase 2 data (n = 1036). The sensitivity of LR2 was 88%, of RMI it was 41% and of subjective assessment it was 87%. The specificity of LR2 was 67%, of RMI it was 90% and of subjective assessment it was 86%. The simple rules yielded a conclusive result in almost 2/3 of the masses, where they resulted in sensitivity and specificity similar to those of real-time subjective assessment by experienced ultrasound examiners: sensitivity 89 vs 89% (P = 0.76), specificity 91 vs 91% (P = 0.65). When a three-step strategy was applied with easy 'instant' diagnoses as Step 1, simple rules where conclusive as Step 2 and subjective assessment by an experienced ultrasound examiner in the remaining masses as Step 3, we obtained a sensitivity of 92% and specificity of 92% compared with sensitivity 90% (P = 0.03) and specificity 93% (P = 0.44) when using real-time subjective assessment by experts in all tumors.

CONCLUSION

A diagnostic strategy using simple descriptors and ultrasound rules when applied to the variables contained in the IOTA database obtains results that are at least as good as those obtained by subjective assessment of a mass by an expert.

Detection of intracavitary uterine pathology using offline analysis of three-dimensional ultrasound volumes: interobserver agreement and diagnostic accuracy

OBJECTIVE

To estimate the diagnostic accuracy and interobserver agreement in predicting intracavitary uterine pathology at offline analysis of three-dimensional (3D) ultrasound volumes of the uterus.

METHODS

3D volumes (unenhanced ultrasound and gel infusion sonography with and without power Doppler, i.e. four volumes per patient) of 75 women presenting with abnormal uterine bleeding at a 'bleeding clinic' were assessed offline by six examiners. The sonologists were asked to provide a tentative diagnosis. A histological diagnosis was obtained by hysteroscopy with biopsy or operative hysteroscopy. Proliferative, secretory or atrophic endometrium was classified as 'normal' histology; endometrial polyps, intracavitary myomas, endometrial hyperplasia and endometrial cancer were classified as 'abnormal' histology. The diagnostic accuracy of the six sonologists with regard to normal/abnormal histology and interobserver agreement were estimated.

RESULTS

Intracavitary pathology was diagnosed at histology in 39% of patients. Agreement between the ultrasound diagnosis and the histological diagnosis (normal vs abnormal) ranged from 67 to 83% for the six sonologists. In 45% of cases all six examiners agreed with regard to the presence/absence of intracavitary pathology. The percentage agreement between any two examiners ranged from 65 to 91% (Cohen's κ, 0.31-0.81). The Schouten κ for all six examiners was 0.51 (95% CI, 0.40-0.62), while the highest Schouten κ for any three examiners was 0.69.

CONCLUSION

When analyzing stored 3D ultrasound volumes, agreement between sonologists with regard to classifying the endometrium/uterine cavity as normal or abnormal as well as the diagnostic accuracy varied substantially. Possible actions to improve interobserver agreement and diagnostic accuracy include optimization of image quality and the use of a consistent technique for analyzing the 3D volumes.

Improving spatiotemporal characterization of cognitive processes with data-driven EEG-fMRI analysis

To fully understand the cognitive processes occurring in the human brain, high resolution in both spatial and temporal information is needed. Most neuroimaging approaches, however, only possess high accuracy in one of these two domains. Therefore, the multimodal analysis of brain activity is becoming more and more popular among the research community. One of these approaches concerns the integration of simultaneously acquired electroencephalographic (EEG) and functional magnetic resonance imaging (fMRI) data. This combination poses a series of challenges, ranging from recovering data quality to the fusion of two types of data of a completely different nature. In this work, several of these challenges will be addressed, and an overview of different integration approaches is provided.

Decoupling the influence of systemic variables in the peripheral and cerebral haemodynamics during ECMO procedure by means of oblique and orthogonal subspace projections

Extra-Corporeal Membrane Oxygenation (ECMO) is a life support system for infants and children with cardio-respiratory failure. During ECMO it is possible to have unstable cerebral haemodynamics, due to strong oscillations in the systemic variables, among other factors, which may lead to brain damage in the patients. Therefore, monitoring the coupling between cerebral haemodynamics and systemic signals might alert us of possible imminent brain damage. In this study we explore the use of orthogonal and oblique subspace projections in the decoupling of these variables, by assessing the ratio between the projections of the haemodynamic variables, onto the subspace spanned by the systemic variables, and the original signals. The coupling of these two systems may differ as different protection mechanisms protect the peripheral system and the brain. Subspace projection was able to decompose the heamodynamic variables as a sum of components related to each systemic variable, separately. As expected, stronger coupling was found between the peripheral haemodynamic and the systemic variables.

Clinical implications of intra- and interobserver reproducibility of transvaginal sonographic measurement of gestational sac and crown-rump length at 6-9 weeks' gestation

OBJECTIVES

To assess intra- and interobserver agreement of routinely performed measurements-crown-rump length (CRL) and mean gestational sac diameter (MSD)-for assessing the likelihood of miscarriage in the first trimester of pregnancy using transvaginal sonography.

METHODS

A cross-sectional study of CRL and gestational sac measurements in first-trimester pregnancies was conducted in a fetal medicine referral center with a predominantly Caucasian population. Gestational age ranged from 6 to 9 weeks. All patients underwent a transvaginal ultrasound examination using a high-resolution ultrasound machine. Two measurements of CRL and measurements of three diameters of the gestational sac were obtained by two observers. Agreement within and between observers for CRL and between observers for MSD was analyzed using 95% prediction intervals, Bland-Altman plots with 95% limits of agreement and the intraclass correlation coefficient (ICC).

RESULTS

In total 54 patients were included in the study, with measurements obtained by both observers in 44 of these. Intra- and interobserver ICCs were high for CRL measurements, with values of 0.992 and 0.993 for intraobserver agreement and 0.993 for interobserver agreement. For the MSD, the interobserver ICC was 0.952. Limits of agreement were ± 8.91 and ± 11.37% for intraobserver agreement of CRL and ± 14.64% for interobserver agreement of CRL. For MSD, the interobserver limits of agreement were ± 18.78%. For an MSD measurement of 20 mm by the first observer, the prediction interval for the second observer was 16.8-24.5 mm. For a CRL measurement of 6 mm, the prediction interval for the second observer was 5.4-6.7 mm.

CONCLUSION

For dating purposes, there is reasonable reproducibility of CRL measurements using transvaginal ultrasonography at 6-9 weeks' gestation. When diagnosing miscarriage based on measurements of CRL care must be taken for values close to any decision boundary. The higher interobserver variability that we observed for MSD has implications for the diagnosis of miscarriage based on this measurement in the absence of a visible embryo or yolk sac.

Adnexal masses difficult to classify as benign or malignant using subjective assessment of gray-scale and Doppler ultrasound findings: logistic regression models do not help

OBJECTIVE

To develop a logistic regression model that can discriminate between benign and malignant adnexal masses perceived to be difficult to classify by subjective evaluation of gray-scale and Doppler ultrasound findings (subjective assessment) and to compare its diagnostic performance with that of subjective assessment, serum CA 125 and the risk of malignancy index (RMI).

METHODS

We used data from the 3511 patients with an adnexal mass included in the International Ovarian Tumor Analysis (IOTA) studies. All patients had been examined using transvaginal gray-scale and Doppler ultrasound following a standardized research protocol carried out by an experienced ultrasound examiner using a high-end ultrasound system. In addition to prospectively collecting information on > 40 clinical and ultrasound variables, the ultrasound examiner classified each mass as certainly or probably benign, unclassifiable, or certainly or probably malignant. A logistic regression model to discriminate between benignity and malignancy was developed for the unclassifiable masses (n = 244, i.e. 7% of all tumors) using a training set (160 tumors, 45 malignancies) and then tested on a test set (84 tumors, 28 malignancies). The gold standard was the histological diagnosis of the surgically removed adnexal mass. The area under the receiver-operating characteristics curve (AUC), sensitivity, specificity, positive likelihood ratio (LR+) and negative likelihood ratio (LR-) were used to describe diagnostic performance and were compared between subjective assessment, CA 125, the RMI and the logistic regression model created.

RESULTS

One variable was retained in the logistic regression model: the largest diameter (in mm) of the largest solid component of the tumor (odds ratio (OR) = 1.04; 95% CI, 1.02-1.06). The model had an AUC of 0.68 (95% CI, 0.59-0.78) on the training set and an AUC of 0.65 (95% CI, 0.53-0.78) on the test set. On the test set, a cut-off of 25% probability of malignancy (corresponding to the largest diameter of the largest solid component of 23 mm) resulted in a sensitivity of 64% (18/28), a specificity of 55% (31/56), an LR+ of 1.44 and an LR- of 0.65. The corresponding values for subjective assessment were 68% (19/28), 59% (33/56), 1.65 and 0.55. On the test set of patients with available CA 125 results, the LR+ and LR- of the logistic regression model (cut-off = 25% probability of malignancy) were 1.29 and 0.73, of subjective assessment were 1.45 and 0.63, of CA 125 (cut-off = 35 U/mL) were 1.24 and 0.84 and of RMI (cut-off = 200) were 1.21 and 0.92.

CONCLUSIONS

About 7% of adnexal masses that are considered appropriate for surgical removal cannot be classified as benign or malignant by experienced ultrasound examiners using subjective assessment. Logistic regression models to estimate the risk of malignancy, CA 125 measurements and the RMI are not helpful in these masses.

Effect of gel-instillation sonography on Doppler ultrasound findings in endometrial polyps

OBJECTIVE

Saline infusion sonohysterography has been reported to suppress the color signal within the endometrium at color or power Doppler evaluation. The aim of this study was to evaluate if gel-instillation sonography (GIS) affects the power Doppler signal in patients with endometrial polyps.

METHODS

Ultrasound volumes of the uterus, obtained by three-dimensional ultrasound imaging of 25 women with histologically confirmed endometrial polyps, were assessed offline by six gynecologists with a special interest in gynecological ultrasound. Each woman contributed four volumes: one gray-scale volume and one power Doppler volume before GIS, and one gray-scale volume and one power Doppler volume at GIS. Power Doppler features before and after gel infusion were compared.

RESULTS

At unenhanced ultrasound a pedicle artery was seen in 27-46% of cases, whereas, after gel infusion the examiners reported a pedicle artery in 30-46% of cases (Exact McNemar's test P-values ranged from 0.50 to 1.00). The level of agreement between unenhanced ultrasound and GIS ranged from 59 to 91% (Cohen's kappa values ranged from 0.17 to 0.79). There was no tendency for a pedicle artery to be identified less often at GIS than before gel instillation.

CONCLUSION

Gel infusion does not affect the power Doppler signal in patients with endometrial polyps.

Lobular and non-lobular breast cancers differ regarding axillary lymph node metastasis: a cross-sectional study on 4,292 consecutive patients

Invasive lobular carcinoma (ILC) accounts for 8-14% of all breast cancers and carries distinct prognostic and biologic implications. The goal of our study was to investigate the impact of lobular histology on axillary lymph node (ALN) involvement. This is a cross-sectional study of 4,292 consecutive patients surgically treated for breast carcinoma at the University Hospitals Leuven. Logistic regression analysis was used to relate ILC to lymph node involvement while controlling for the following clinicopathologic features: tumor size, multifocal disease, tumor grade, lobular subtype and the combined steroid, and Her-2 status. Odds ratios (ORs) and 95% confidence intervals (CIS) were computed. A subgroup analysis was performed for patients that underwent a sentinel lymph node (SLN) procedure. The observed incidence of ILC was 13%. ILCs were larger, were more often grade II, multifocal, steroid receptor positive and Her-2 negative, and tended to be present in older patients. Incidence of ALN involvement was 42.0% for ILCs versus 38.3% for other tumors (OR 1.17, 95% CI 0.97-1.40). For the SLN subgroup, ILCs were less often ALN positive than non-ILCs (20.5% versus 28.3%, OR 0.66, 95% CI: 0.41-1.00). In the multivariable analysis, the lobular subtype was identified as less likely to have ALN involvement (adjusted OR 0.66, 95% CI 0.53-0.82). The analysis for the SLN subgroup showed comparable results (adjusted OR 0.49, 95% CI 0.30-0.78). This study has demonstrated that the lobular subtype is an independent predictor of lymph node involvement with ILC having a lower incidence of involved lymph nodes. The mildly higher incidence of ALN metastasis in lobular cancers in univariable analysis is not due to the lobular subtype, but due to confounding factors that interact with lymph node involvement.

A model and scoring system to predict outcome of intrauterine pregnancies of uncertain viability

OBJECTIVES

To define the incidence and outcome of intrauterine pregnancy of uncertain viability (PUV) and to develop and assess the performance of a model and a scoring system to predict ongoing viability.

METHODS

Of 1881 consecutive women undergoing transvaginal ultrasonography, a cohort of 493 women with an empty gestational sac < 20 mm in mean diameter, gestational sac < 25 mm in mean diameter and containing yolk sac only or an embryonic pole < 6 mm in maximum length and without visible heart activity were followed until the end of the first trimester. Women with multiple pregnancies or who underwent termination of pregnancy were excluded. Outcome measures were pregnancy viability at initial 7-14-day follow-up and first-trimester viability at 11-14 weeks. The data were split randomly into two sets (two-thirds and one-third, respectively) in order to first develop and then test a mathematical model and a 'simple' model in the prediction of viability at each outcome point, based on maternal demographics, ultrasound features and symptoms. The performance of each system was assessed by receiver-operating characteristics (ROC) curve analysis and calibration plots on a test dataset.

RESULTS

The incidence of PUV in this population was 29.2% (549/1881). Of the 493 pregnancies with initial (7-14 days) follow-up available, 307 (62.3%) were viable at this time and of the 444 pregnancies with follow-up at the end of the first trimester, 225 (50.7%) were still viable. Initial (7-14-day) viability was predicted by the model with an area under the ROC curve (AUC) of 0.837 (95% CI, 0.791-0.884) in the training dataset and 0.821 (95% CI, 0.756-0.885) in the test dataset. First-trimester (11-14-week) viability was predicted by the model with an AUC of 0.788 (95% CI, 0.734-0.842) in the training dataset and 0.774 (95% CI, 0.701-0.848) in the test dataset. The scoring system performed slightly worse than did the model, but had the advantage of being easily applicable.

CONCLUSIONS

When early pregnancy viability cannot be established immediately with ultrasound, use of either a logistic regression model or a scoring system allows an individualized prediction of first-trimester outcome.

Non-negative blind source separation techniques for tumor tissue typing using HR-MAS signals

Given High Resolution Magic Angle Spinning (HR-MAS) signals from several glioblastoma tumor subjects, the goal is to differentiate between tumor tissue types by separating the different sources that contribute to the profile of each spectrum. Blind source separation techniques are applied for obtaining characteristic profiles for necrosis, high cellular tumor and border tumor tissue, and providing the contribution (abundance) of each tumor tissue to the profile of the spectra. The problem is formulated as a non-negative source separation problem. We illustrate the effectiveness of the proposed methods and we analyze to which extent the dimension of the input space could influence the performance by comparing the results on the full magnitude signals and on dimensionally reduced spaces.