Improving the Detection of Potential Cases of Familial Hypercholesterolemia: Could Machine Learning Be Part of the Solution?

BACKGROUND: Familial hypercholesterolemia (FH), while highly prevalent, is a significantly underdiagnosed monogenic disorder. Improved detection could reduce the large number of cardiovascular events attributable to poor case finding. We aimed to assess whether machine learning algorithms outperform clinical diagnostic criteria (signs, history, and biomarkers) and the recommended screening criteria in the United Kingdom in identifying individuals with FH-causing variants, presenting a scalable screening criteria for general populations. METHODS AND RESULTS: Analysis included UK Biobank participants with whole exome sequencing, classifying them as having FH when (likely) pathogenic variants were detected in their LDLR, APOB, or PCSK9 genes. Data were stratified into 3 data sets for (1) feature importance analysis; (2) deriving state-of-the-art statistical and machine learning models; (3) evaluating models' predictive performance against clinical diagnostic and screening criteria: Dutch Lipid Clinic Network, Simon Broome, Make Early Diagnosis to Prevent Early Death, and Familial Case Ascertainment Tool. One thousand and three of 454 710 participants were classified as having FH. A Stacking Ensemble model yielded the best predictive performance (sensitivity, 74.93%; precision, 0.61%; accuracy, 72.80%, area under the receiver operating characteristic curve, 79.12%) and outperformed clinical diagnostic criteria and the recommended screening criteria in identifying FH variant carriers within the validation data set (figures for Familial Case Ascertainment Tool, the best baseline model, were 69.55%, 0.44%, 65.43%, and 71.12%, respectively). Our model decreased the number needed to screen compared with the Familial Case Ascertainment Tool (164 versus 227). CONCLUSIONS: Our machine learning-derived model provides a higher pretest probability of identifying individuals with a molecular diagnosis of FH compared with current approaches. This provides a promising, cost-effective scalable tool for implementation into electronic health records to prioritize potential FH cases for genetic confirmation.

Ensemble machine learning methods in screening electronic health records: A scoping review.

Background: Electronic health records provide the opportunity to identify undiagnosed individuals likely to have a given disease using machine learning techniques, and who could then benefit from more medical screening and case finding, reducing the number needed to screen with convenience and healthcare cost savings. Ensemble machine learning models combining multiple prediction estimates into one are often said to provide better predictive performances than non-ensemble models. Yet, to our knowledge, no literature review summarises the use and performances of different types of ensemble machine learning models in the context of medical pre-screening. Method: We aimed to conduct a scoping review of the literature reporting the derivation of ensemble machine learning models for screening of electronic health records. We searched EMBASE and MEDLINE databases across all years applying a formal search strategy using terms related to medical screening, electronic health records and machine learning. Data were collected, analysed, and reported in accordance with the PRISMA scoping review guideline. Results: A total of 3355 articles were retrieved, of which 145 articles met our inclusion criteria and were included in this study. Ensemble machine learning models were increasingly employed across several medical specialties and often outperformed non-ensemble approaches. Ensemble machine learning models with complex combination strategies and heterogeneous classifiers often outperformed other types of ensemble machine learning models but were also less used. Ensemble machine learning models methodologies, processing steps and data sources were often not clearly described. Conclusions: Our work highlights the importance of deriving and comparing the performances of different types of ensemble machine learning models when screening electronic health records and underscores the need for more comprehensive reporting of machine learning methodologies employed in clinical research.

Primary repair versus surgical and transcatheter palliation in infants with tetralogy of Fallot.

OBJECTIVES: Treatment of infants with tetralogy of Fallot (ToF) has evolved in the last two decades with increasing use of primary surgical repair (PrR) and transcatheter right ventricular outflow tract palliation (RVOTd), and fewer systemic-to-pulmonary shunts (SPS). We aim to report contemporary results using these treatment options in a comparative study. METHODS: This a retrospective study using data from the UK National Congenital Heart Disease Audit. All infants (n=1662, median age 181 days) with ToF and no other complex defects undergoing repair or palliation between 2000 and 2013 were considered. Matching algorithms were used to minimise confounding due to lower age and weight in those palliated. RESULTS: Patients underwent PrR (n=1244), SPS (n=311) or RVOTd (n=107). Mortality at 12 years was higher when repair or palliation was performed before the age of 60 days rather than after, most significantly for primary repair (18.7% vs 2.2%, P<0.001), less so for RVOTd (10.8% vs 0%, P=0.06) or SPS (12.4% vs 8.3%, P=0.2). In the matched groups of patients, RVOTd was associated with more right ventricular outflow tract (RVOT) reinterventions (HR=2.3, P=0.05 vs PrR, HR=7.2, P=0.001 vs SPS) and fewer pulmonary valve replacements (PVR) (HR=0.3 vs PrR, P=0.05) at 12 years, with lower mortality after complete repair (HR=0.2 versus PrR, P=0.09). CONCLUSIONS: We found that RVOTd was associated with more RVOT reinterventions, fewer PVR and fewer deaths when compared with PrR in comparable, young infants, especially so in those under 60 days at the time of the first procedure.

Aortic Valve Replacement and the Ross Operation in Children and Young Adults.

BACKGROUND: There are several options available for aortic valve replacement (AVR), with few comparative reports in the literature. The optimal choice for AVR in each age group is not clear. OBJECTIVES: The study sought to report and compare outcomes after AVR in the young using data from a national database. METHODS: AVR procedures were compared after advanced matching, both in pairs and in a 3-way manner, using a Bayesian dynamic survival model. RESULTS: A total of 1,501 patients who underwent AVR in the United Kingdom between 2000 and 2012 were included. Of these, 47.8% had a Ross procedure, 37.8% a mechanical AVR, 10.9% a bioprosthesis AVR, and 3.5% a homograft AVR, with Ross patients being significantly younger when compared to the other groups. Overall survival at 12 years was 94.6%. In children, the Ross procedure had a 12.7% higher event-free probability (death or any reintervention) at 10 years when compared to mechanical AVR (p = 0.05). We also compared all procedures except the homograft in a matched population of young adults, where the bioprosthesis had the lowest event-free probability of 78.8%, followed by comparable results in mechanical AVR and Ross, with 86.3% and 89.6%, respectively. Younger age was associated with mortality and pulmonary reintervention in the Ross group and with aortic reintervention in the mechanical AVR. Of all 3 options, only the patients undergoing the Ross procedure approached the survival of the general population. CONCLUSIONS: AVR in the young achieves good results, with the Ross being overall better suited for this age group, especially in children. Although freedom from aortic valve reintervention is superior after the Ross procedure, the need for homograft reinterventions is an issue to take into account. All methods have advantages and limitations, with reinterventions being an issue in the long term for all, more crucially in smaller children.

Indications and results of systemic to pulmonary shunts: results from a national database.

OBJECTIVES: The systemic-to-pulmonary shunt (SPS) remains an important palliative therapy in many congenital heart defects. Unlike other surgical treatments, the mortality after shunt operations has risen. We used an audit dataset to investigate potential reasons for this change and to report national results. METHODS: A total of 1993 patients classified in 13 diagnoses underwent an SPS procedure between 2000 and 2013. Indication trends by era and also results before repair or next stage are reported. A dynamic hazard model with competing risks and modulated renewal was used to determine predictors of outcomes. RESULTS: The usage of SPS in Tetralogy of Fallot (ToF) has significantly decreased in the last decade, with cases of single ventricle (SV) and pulmonary atresia (PA) with septal communication increasing (P < 0.001 for trends). This is correlated with an increase of early mortality from 5.1% in the first half of the decade to 9.8% in the latter (P = 0.007 for trend). At 1.5 years, 13.9% of patients have died, 17.8% had a shunt reintervention and 68.3% of patients are alive and reintervention-free. Low weight, PA-intact septum, SV and central shunt type are among the factors associated with increased mortality, whereas PA-ventricular septal defect, corrected transposition, isomerism, central shunt and low weight are among those associated with increased reintervention, also having a dynamic effect on the relative risk when compared with ToF patients. Shunt reinterventions are not associated with worse outcomes when adjusted by other covariates, but they do have higher 30-day mortality if occurring earlier than 30 days from the index (P < 0.001). Patients operated in later years were found to have significantly lower survival at a distance from index. CONCLUSIONS: The observed historical rise in mortality for shunt operations relates to complex factors including changing practice for repair of ToF and for univentricular palliation. PA and SV patients are the groups of patients at the highest risk of death. Small size, shunt type and underlying anatomical defect are the main determinants of outcomes. Trends in indication and mortality seem to indicate that more severely ill patients benefit from shunting, but with an increase in mortality.

Contextual interaction of GABAergic circuitry with dynamic synapses.

Visual context shapes human perception, yet our understanding of this phenomenon in terms of synaptic circuitry is still rudimentary. Our in vitro experiments with avian tectum reveal two distinct GABAergic pathways that mediate the spatiotemporal tectal interaction of retinal inputs. One pathway mediates postsynaptic lateral inhibition. The other pathway interacts with the synaptic depression of retinotectal synapses. Simulations of an experimentally constrained model including the two pathways reproduce the observed avian tectum wide-field neuron's sensitivity to small and moving stimuli, while being insensitive to whole-field motion.

Iterative cooperation between parallel pathways for object and background motion.

In a typical visual scene, one or more objects move relative to a larger background, which can itself be in motion as a result of the observer's eyes moving with respect to the outside world. Here we show that accurate estimation of the background motion from an image velocity field can be accomplished through an iterative cooperation between two modules: one that specializes in calculating a weighted average velocity and another one calculating a velocity contrast map. We build on our analysis to provide a model for the tectum-pretectum loop in the nonmammalian midbrain. Our model accounts for some of the known properties of the tectal neurons (sensitivity to relative motion) and pretectal neurons (sensitivity to whole-field motion). It also agrees with our knowledge of the pretectotectal projection (divergent and inhibitory), and with the results of lesion studies in which the pretectal input to the tectum was removed, leading to hyperactivity of the tectal neurons and the animal. Our model also makes a testable prediction regarding the tectopretectal projection, i.e., that the presence of a larger object and a bigger discrepancy between the directions of motion for the object and the background lead to a larger error by the pretectum in estimating the background motion when the tectal input is abolished.

Sparse spatial sampling for the computation of motion in multiple stages.

The avian retino-tecto-rotundal pathway plays a central role in motion analysis and features complex connectivity. Yet, the relation between the pathway's structural arrangement and motion computation has remained elusive. For an important type of tectal wide-field neuron, the stratum griseum centrale type I (SGC-I) neuron, we quantified its structure and found a spatially sparse but extensive sampling of the retinal projection. A computational investigation revealed that these structural properties enhance the neuron's sensitivity to change, a behaviorally important stimulus attribute, while preserving information about the stimulus location in the SGC-I population activity. Furthermore, the SGC-I neurons project with an interdigitating topography to the nucleus rotundus, where the direction of motion is computed. We showed that, for accurate direction-of-motion estimation, the interdigitating projection of tectal wide-field neurons requires a two-stage rotundal algorithm, where the second rotundal stage estimates the direction of motion from the change in the relative stimulus position represented in the first stage.

Motion contrast classification is a linearly nonseparable problem.

Sensitivity to image motion contrast, that is, the relative motion between different parts of the visual field, is a common and computationally important property of many neurons in the visual pathways of vertebrates. Here we illustrate that, as a classification problem, motion contrast detection is linearly nonseparable. In order to do so, we prove a theorem stating a sufficient condition for linear nonseparability. We argue that nonlinear combinations of local measurements of velocity at different locations and times are needed in order to solve the motion contrast problem.

Motion repulsion arises from stimulus statistics when analyzed with a clustering algorithm.

Motion repulsion is the perceived enlargement of the angle between the directions of motion of two transparently moving patterns. An explanation of this illusion has long been sought for in the neural circuitry of the brain. We show that motion repulsion already arises from the statistical properties of the motion transparency problem when analyzed with a clustering algorithm.