Sensitive Balance Research Articles

Semi-supervised learning from coarse histopathology labels

ABSTRACT Ultrasound imaging is commonly used to guide sampling the prostate tissue in transrectal biopsies, followed by detection of cancer through histopathological analysis and coarse labelling of sampled tissue. Ideally, the procedure should be improved by developing machine learning solutions that can identify the presence of cancer in ultrasound images to guide the biopsy procedure. Training a fully supervised learning model using coarse histopathology labels suffers from weakly annotated data which introduce label noise for each image pixel. To address this challenge, we propose a semi-supervised framework for learning with noisy labels. We leverage a two-component mixture model to cluster the training data into clean and noisy label samples based on their loss values. Then, during the semi-supervised training phase, we utilise the well-known MixMatch algorithm which incorporates consistency regularisation, entropy minimisation, and the Mixup regularisation as well as the cross-entropy loss function for noisy and clean sets, respectively. We evaluate the proposed framework with prostate ultrasound data obtained from 71 subjects, while sampling 264 biopsy cores. We achieve balanced accuracy, sensitivity, and specificity of 78.6%, 80.0%, and 77.1%, respectively. In a detailed comparison study, we demonstrate that our proposed framework outperforms the fully supervised method with state-of-the-art robust loss functions.

Relationship between physical activity level and plantar sensitivity, balance in postmenopausal women

This research was planned to investigate the relationship between physical activity level and plantar sensitivity, static and dynamic balance in postmenopausal women. Forty women between the ages of 50 and 65, who had not had a menstrual period for at least 12 months, and who had undergone menopause due to hormonal, surgical or natural means were involved in the research. International Physical Activity Questionnaires Short Form (IPAQ-SF) was used to assess the level of physical activity. Light touch pressure threshold on soles of the individuals and two-point discrimination sense were evaluated bilaterally from 6 different regions of the sole of the foot. Single leg and tandem stance tests were used for static balance assessment. Dynamic balance was evaluated with Berg Balance Scale (BBS) and Timed Up and Go Test (TUG). As a result of the statistical analysis, a negative correlation was found between the total physical activity level of the participants and the light touch pressure of the right big toe. There was a significant negative correlation between total physical activity level and left big toe, and the 1st and 5th metatarsal heads of the left foot. A significant negative correlation between total physical activity level and TUG test was detected. However, no significant difference between total physical activity and BBS, static balance tests was found. As a result of this research, it was concluded that the level of physical activity affects the light touch sense of big toes of both feet, and the 1st and 5th metatarsal heads of left foot. During this study, it was seen that if total physical activity level decreases, then the score of TUG can increase. Physical activity level and related factors should be examined with a larger number of samples.

Association of Postnatal Growth Changes and Neurodevelopmental Outcomes in Preterm Neonates of

To examine associations between weight and head circumference (HC) changes and neurodevelopment in preterm infants. This retrospective cohort study of Canadian Neonatal Network and Canadian Neonatal Follow-Up Network sites included preterm infants born 2010-2018. Logistic regression and model diagnostics evaluated relationships between changes in z score and velocity of weight and HC from birth to discharge from a tertiary neonatal intensive care unit, discharge to 18-24months corrected age (CA), and birth to 18-24months CA and significant cognitive/motor impairment at 18-24months CA classified using a Bayley Scales of Infant and Toddler Development-Third Edition cognitive or motor composite score <70. In total, 4530 infants (53.0% male) with a mean (SD) gestational age of 26.3 (1.4) weeks and birth weight of 920 (227) g were included. Weight and HC changes were associated with lower odds of significant cognitive/motor impairment including an OR of 0.87 (95% CI: 0.83, 0.91; P< .001) for a 1-g/d increase in weight from discharge to 18-24months CA and 0.81 (95% CI: 0.75, 0.88; P< .001) for a 1-unit increase in HC z score from birth to 18-24months CA. Associations were not statistically significant in morbidity-free neonates. Weight and HC gains poorly discriminated between infants with and without significant cognitive/motor impairment (areas under the receiver operating characteristic curve of <0.64). No growth measure had a clinically useful balance of sensitivity and specificity. Weight and HC changes were associated with significant cognitive/motor impairment but had poor discriminatory capability. Neonatal morbidities may make a larger contribution than postnatal growth to neurodevelopment.

Clinical evaluation of malignancy diagnosis of rare thyroid carcinomas by an artificial intelligent automatic diagnosis system.

To evaluate the application value of a generally trained artificial intelligence (AI) automatic diagnosis system in the malignancy diagnosis of rare thyroid carcinomas, such as follicular thyroid carcinoma, medullary thyroid carcinoma, primary thyroid lymphoma and anaplastic thyroid carcinoma and compare the diagnostic performance with radiologists of different experience levels. We retrospectively studied 342 patients with 378 thyroid nodules that included 196 rare malignant nodules by using postoperative pathology as the gold standard, and compared the diagnostic performances of three radiologists (one junior, one mid-level, one senior) and that of AI automatic diagnosis system. The accuracy of the AI system in malignancy diagnosis was 0.825, which was significantly higher than that of all three radiologists and higher than the best radiologist in this study by a margin of 0.097 with P-value of 2.252 × 10-16. The mid-level radiologist and senior radiologist had higher sensitivity (0.857 and 0.959) than that of the AI system (0.847) at the cost of having much lower specificity (0.533, 0.478 versus 0.802). The junior radiologist showed relatively balanced sensitivity and specificity (0.816 and 0.549) but both were lower than that of the AI system. The generally trained AI automatic diagnosis system showed high accuracy in the differential diagnosis of begin nodules and rare malignancy nodules. It may assist radiologists for screening of rare malignancy nodules that even senior radiologists are not acquainted with.

'Optimal' cutoff selection in studies of depression screening tool accuracy using the PHQ-9, EPDS, or HADS-D: A meta-research study.

Optimal cutoff thresholds are selected to separate 'positive' from 'negative' screening results. We evaluated how depression screening tool studies select optimal cutoffs. We included studies from previously conducted meta-analyses of Patient Health Questionnaire-9, Edinburgh Postnatal Depression Scale, or Hospital Anxiety and Depression Scale-Depression accuracy. Outcomes included whether an optimal cutoff was selected, method used, recommendations made, and reporting guideline and protocol citation. Of 212 included studies, 172 (81%) attempted to identify an optimal cutoff, and 147 of these 172 (85%) reported one or more methods. Methods were heterogeneous with Youden's J (N=35, 23%) most common. Only 23 of 147 (16%) studies described a rationale for their method. Rationales focused on balancing sensitivity and specificity without describing why desirable. 131 of 172 studies (76%) identified an optimal cutoff other than the standard; most did not make use recommendations (N=56; 43%) or recommended using a non-standard cutoff (N=53; 40%). Only 4 studies cited a reporting guideline, and 4 described a protocol with optimal cutoff selection methods, but none used the protocol method in the published study. Research is needed to guide how selection of cutoffs for depression screening tools can be standardized and reflect clinical considerations.

Model-based Cross-correlation Search for Gravitational Waves from the Low-mass X-Ray Binary Scorpius X-1 in LIGO O3 Data

We present the results of a model-based search for continuous gravitational waves from the low-mass X-ray binary Scorpius X-1 using LIGO detector data from the third observing run of Advanced LIGO and Advanced Virgo. This is a semicoherent search that uses details of the signal model to coherently combine data separated by less than a specified coherence time, which can be adjusted to balance sensitivity with computing cost. The search covered a range of gravitational-wave frequencies from 25 to 1600 Hz, as well as ranges in orbital speed, frequency, and phase determined from observational constraints. No significant detection candidates were found, and upper limits were set as a function of frequency. The most stringent limits, between 100 and 200 Hz, correspond to an amplitude h 0 of about 10−25 when marginalized isotropically over the unknown inclination angle of the neutron star’s rotation axis, or less than 4 × 10−26 assuming the optimal orientation. The sensitivity of this search is now probing amplitudes predicted by models of torque balance equilibrium. For the usual conservative model assuming accretion at the surface of the neutron star, our isotropically marginalized upper limits are close to the predicted amplitude from about 70 to 100 Hz; the limits assuming that the neutron star spin is aligned with the most likely orbital angular momentum are below the conservative torque balance predictions from 40 to 200 Hz. Assuming a broader range of accretion models, our direct limits on gravitational-wave amplitude delve into the relevant parameter space over a wide range of frequencies, to 500 Hz or more.

Performance of the GAD-7 in adults with dissociative seizures

PurposeLittle is known about the accuracy of the GAD-7, a self-report anxiety measure, in detecting generalised anxiety disorder (GAD) in people with dissociative seizures (DS). We evaluated the reliability, validity and uniformity of the GAD-7 using a diagnosis of GAD on the Mini-International Neuropsychiatric Interview as a reference. MethodsWe assessed 368 adults with DS at the pre-randomisation phase of the CODES trial. Factor analysis for categorical data assessed GAD-7 uniformity. Diagnostic accuracy was assessed by estimating the area under the curve (AUC). We evaluated discriminant validity, reviewed data on convergent validity and calculated internal consistency. We explored correlations between GAD-7 scores and monthly DS frequency, frequency of severe seizures and measures of behavioural and emotional avoidance. ResultsInternal consistency of the GAD-7 was high (α = 0.92). Factor analysis elicited one main factor and general measurement invariance. Diagnostic accuracy was fair (AUC = 0.72) but the best balance of sensitivity and specificity occurred at a cut-off of ≥12 and still had a specificity rate of only 68%. Discriminant and convergent validity were good. GAD-7 scores correlated positively with DS frequency, severe seizure frequency, behavioural and emotional avoidance (all p < 0.001). ConclusionFindings regarding internal consistency and factor structure parallel previous psychometric evaluations of the GAD-7. Correlations between GAD-7 scores and DS occurrence/severity and avoidance are evidence of the concept validity of GAD-7 and provide further support for a fear-avoidance treatment model for DS. However, the utility of the GAD-7 as a diagnostic instrument for generalised anxiety disorder is limited in patients with DS.

A long-term mass-balance reconstruction (1974–2021) and a decadal in situ mass-balance record (2011–2021) of Rikha Samba Glacier, central Himalaya

AbstractDespite their importance for regional water resource planning and as indicators of climate change, records of in situ glacier mass balance remain short and spatially sparse in the Himalaya. Here, we present an updated series of in situ mass-balance measurements from Rikha Samba Glacier, Nepal, between 2011 and 2021. The updated in situ mass balance is −0.39 ± 0.32 m w.e. for this period. We use an energy-mass balance model to extend the annual mass-balance series back to 1974. The model is forced using daily meteorological variables from ERA5-Land reanalysis data that is linearly bias-corrected using observations from an automatic weather station situated near the glacier terminus. The modeled mass balance is consistent with the in situ mass-balance series measured 2011–2021 and with previous glaciological and geodetic estimates. The model results indicate a mass balance of −0.56 ± 0.27 m w.e. a−1over the reconstruction period of 1974–2021, which is comparable to the mass losses experienced by other Himalayan glaciers during this time. An assessment of the sensitivity of the glacier mass balance to meteorological forcing suggests that a change in temperature of ±1 K has a stronger effect on the calculated mass balance compared to a ±20% change in either precipitation, or relative humidity, or solar radiation.

Sequential gene expression analysis of cervical malignant transformation identifies RFC4 as a novel diagnostic and prognostic biomarker

BackgroundCervical squamous cell carcinoma (SCC) is known to arise through increasingly higher-grade squamous intraepithelial lesions (SILs) or cervical intraepithelial neoplasias (CINs). This study aimed to describe sequential molecular changes and identify biomarkers in cervical malignant transformation.MethodsMultidimensional data from five publicly available microarray and TCGA-CESC datasets were analyzed. Immunohistochemistry was carried out on 354 cervical tissues (42 normal, 62 CIN1, 26 CIN2, 47 CIN3, and 177 SCC) to determine the potential diagnostic and prognostic value of identified biomarkers.ResultsWe demonstrated that normal epithelium and SILs presented higher molecular homogeneity than SCC. Genes in the region (e.g., 3q, 12q13) with copy number alteration or HPV integration were more likely to lose or gain expression. The IL-17 signaling pathway was enriched throughout disease progression with downregulation of IL17C and decreased Th17 cells at late stage. Furthermore, we identified AURKA, TOP2A, RFC4, and CEP55 as potential causative genes gradually upregulated during the normal-SILs-SCC transition. For detecting high-grade SIL (HSIL), TOP2A and RFC4 showed balanced sensitivity (both 88.2%) and specificity (87.1 and 90.1%), with high AUC (0.88 and 0.89). They had equivalent diagnostic performance alone to the combination of p16INK4a and Ki-67. Meanwhile, increased expression of RFC4 significantly and independently predicted favorable outcomes in multi-institutional cohorts of SCC patients.ConclusionsOur comprehensive study of gene expression profiling has identified dysregulated genes and biological processes during cervical carcinogenesis. RFC4 is proposed as a novel surrogate biomarker for determining HSIL and HSIL+, and an independent prognostic biomarker for SCC.

Combined brain network topological metrics with machine learning algorithms to identify essential tremor.

Essential tremor (ET) is a common movement syndrome, and the pathogenesis mechanisms, especially the brain network topological changes in ET are still unclear. The combination of graph theory (GT) analysis with machine learning (ML) algorithms provides a promising way to identify ET from healthy controls (HCs) at the individual level, and further help to reveal the topological pathogenesis in ET. Resting-state functional magnetic resonance imaging (fMRI) data were obtained from 101 ET and 105 HCs. The topological properties were analyzed by using GT analysis, and the topological metrics under every single threshold and the area under the curve (AUC) of all thresholds were used as features. Then a Mann-Whitney U-test and least absolute shrinkage and selection operator (LASSO) were conducted to feature dimensionality reduction. Four ML algorithms were adopted to identify ET from HCs. The mean accuracy, mean balanced accuracy, mean sensitivity, mean specificity, and mean AUC were used to evaluate the classification performance. In addition, correlation analysis was carried out between selected topological features and clinical tremor characteristics. All classifiers achieved good classification performance. The mean accuracy of Support vector machine (SVM), logistic regression (LR), random forest (RF), and naïve bayes (NB) was 84.65, 85.03, 84.85, and 76.31%, respectively. LR classifier achieved the best classification performance with 85.03% mean accuracy, 83.97% sensitivity, and an AUC of 0.924. Correlation analysis results showed that 2 topological features negatively and 1 positively correlated with tremor severity. These results demonstrated that combining topological metrics with ML algorithms could not only achieve high classification accuracy for discrimination ET from HCs but also help us to reveal the potential topological pathogenesis of ET.

Thumbs up or thumbs down: neural processing of social feedback and links to social motivation in adolescent girls.

Adolescence is a period of rapid biological and psychological development, characterized by increasing emotional reactivity and risk-taking, especially in peer contexts. Theories of adolescent neural development suggest that the balance in sensitivity across neural threat, reward and regulatory systems contributes to these changes. Building on previous research, this study used a novel social feedback task to explore activation and functional connectivity in the context of social threat and reward in a sample of mid-adolescent girls (n = 86, Mage = 16.32). When receiving negative peer feedback, adolescents showed elevated activation in, and amygdala connectivity with, social processing regions [e.g. medial prefrontal cortex (mPFC) and temporoparietal junction (TPJ)]. When receiving positive feedback, adolescents showed elevated activation in social and reward (e.g. mPFC and ventromedial prefrontal cortex) processing regions and less striatum-cerebellum connectivity. To understand the psychological implications of neural activation and co-activation, we examined associations between neural processing of threat and reward and self-reported social goals. Avoidance goals predicted elevated amygdala and striatum connectivity with social processing regions [e.g. medial temporal gyrus (MTG)], whereas approach goals predicted deactivation in social processing regions (e.g. MTG/TPJ and precuneus), highlighting the importance of considering individual differences in sensitivity to social threat and reward in adolescence.

Validity of the Single-Item Screen–Cannabis (SIS-C) for Cannabis Use Disorder Screening in Routine Care

Cannabis use is prevalent and increasing, and frequent use intensifies the risk of cannabis use disorder (CUD). CUD is underrecognized in medical settings, but a validated single-item cannabis screen could increase recognition. To evaluate the Single-Item Screen-Cannabis (SIS-C), administered and documented in routine primary care, compared with a confidential reference standard measure of CUD. This diagnostic study included a sample of adult patients who completed routine cannabis screening between January 28 and September 12, 2019, and were randomly selected for a confidential survey about cannabis use. Random sampling was stratified by frequency of past-year use and race and ethnicity. The study was conducted at an integrated health system in Washington state, where adult cannabis use is legal. Data were analyzed from May 2021 to March 2022. The SIS-C asks about frequency of past-year cannabis use with responses (none, less than monthly, monthly, weekly, daily or almost daily) documented in patients' medical records. The Diagnostic and Statistical Manual, Fifth Edition (DSM-5) Composite International Diagnostic Interview-Substance Abuse Module (CIDI-SAM) for past-year CUD was completed on a confidential survey and considered the reference standard. The SIS-C was compared with 2 or more criteria on the CIDI-SAM, consistent with CUD. All analyses were weighted, accounting for survey design and nonresponse, to obtain estimates representative of the health system primary care population. Of 5000 sampled adult patients, 1688 responded to the cannabis survey (34% response rate). Patients were predominantly middle-aged (weighted mean [SD] age, 50.7 [18.1]), female or women (weighted proportion [SE], 55.9% [4.1]), non-Hispanic (weighted proportion [SE], 96.7% [1.0]), and White (weighted proportion [SE], 74.2% [3.7]). Approximately 6.6% of patients met criteria for past-year CUD. The SIS-C had an area under receiver operating characteristic curve of 0.89 (95% CI, 0.78-0.96) for identifying CUD. A threshold of less than monthly cannabis use balanced sensitivity (0.88) and specificity (0.83) for detecting CUD. In populations with a 6% prevalence of CUD, predictive values of a positive screen ranged from 17% to 34%, while predictive values of a negative screen ranged from 97% to 100%. In this diagnostic study, the SIS-C had excellent performance characteristics in routine care as a screen for CUD. While high negative predictive values suggest that the SIS-C accurately identifies patients without CUD, low positive predictive values indicate a need for further diagnostic assessment following positive results when screening for CUD in primary care.

Evaluation of indicators of goniometry of the ankle joint in the complex rehabilitation of athletes with peroneal tendinopathy

The study reflects the results of comprehensive rehabilitation of professional athletes with fibular tendinopathy. The effectiveness was evaluated based on the indicators of goniometry of the ankle joint at different stages of the rehabilitation process.Objective. To evaluate the effectiveness of the developed comprehensive program of physical rehabilitation with tendinopathy of the fibular tendons.Materials and methods. The study involved 60 professional male athletes with the diagnosis of “fibular tendinopathy” established by ultrasound examination. All patients were randomly divided into two equal groups of 30 people. Group I patients, in addition to the main ones, were assigned specially designed exercises for balance, balance and proprioceptive sensitivity on the Biodex Balance System (BBS) biofeedback system (Biodex Medical Systems, Inc. New York, USA), as well as balancing disksResults. The percentage of reduction in the total amplitude of movement in the ankle joint, that is, its deficiency, statistically significantly differed between the groups. Thus, at the time of examination before the start of treatment, the total amplitude deficit in group I patients was 13.2°, and in group II – 14.0°. At week 6 after the start of treatment, this indicator improved in group I patients and amounted to 3.8°, and in group II – 12.2°.

An artificial intelligence ultrasound system's ability to distinguish benign from malignant follicular-patterned lesions.

To evaluate the application value of a generally trained artificial intelligence (AI) automatic diagnosis system in the malignancy diagnosis of follicular-patterned thyroid lesions (FPTL), including follicular thyroid carcinoma (FTC), adenomatoid hyperplasia nodule (AHN) and follicular thyroid adenoma (FTA) and compare the diagnostic performance with radiologists of different experience levels. We retrospectively reviewed 607 patients with 699 thyroid nodules that included 168 malignant nodules by using postoperative pathology as the gold standard, and compared the diagnostic performances of three radiologists (one junior, two senior) and that of AI automatic diagnosis system in malignancy diagnosis of FPTL in terms of sensitivity, specificity and accuracy, respectively. Pairwise t-test was used to evaluate the statistically significant difference. The accuracy of the AI system in malignancy diagnosis was 0.71, which was higher than the best radiologist in this study by a margin of 0.09 with a p-value of 2.08×10-5. Two radiologists had higher sensitivity (0.84 and 0.78) than that of the AI system (0.69) at the cost of having much lower specificity (0.35, 0.57 versus 0.71). One senior radiologist showed balanced sensitivity and specificity (0.62 and 0.54) but both were lower than that of the AI system. The generally trained AI automatic diagnosis system can potentially assist radiologists for distinguishing FTC from other FPTL cases that share poorly distinguishable ultrasonographical features.

Diagnostic Performance of Noninvasive Tests for Advanced Hepatic Fibrosis in Young Age Population

Most noninvasive tests (NITs) for hepatic fibrosis are designed for middle-aged patients with chronic liver disease. We compared the diagnostic performance of major NITs (aspartate aminotransferase-to-platelet ratio index [APRI], Fibrosis-4 index, and nonalcoholic fatty liver disease fibrosis score) for a community-based cohort. This cross-sectional study analyzed 8775 participants who underwent magnetic resonance elastography at community health check-up centers. Advanced hepatic fibrosis (≥F3)was defined by magnetic resonance elastography thresholds of 3.6 kPa. Thediagnostic performance of 3 NITs was evaluated according to the etiology of liver disease, sex,metabolic syndrome, obesity, and increased aminotransferase levels in 4 age groups. The APRI generally showed the best area under the receiver operating characteristic curve in patients aged 45 years or younger, and it was statistically significant in patients with chronic viral hepatitis and alcoholic fatty liver disease (P < .043). The best APRI cut-off value for detecting advanced hepatic fibrosis was 0.4, with a sensitivity and specificity of 75.8% and 73.5%, respectively, in the community-based cohort. The APRI showed balanced sensitivity and specificity across all age groups, whereas the other metrics showed low sensitivity in those aged <45 and low specificity in those >65 years. The APRI showed better sensitivity and negative predictive value than the Fibrosis-4 index and the nonalcoholic fatty liver disease fibrosis score in community-based populations with mixed etiology, and, thus, can be performed as the primary test in young adults (age, ≤45 y).

Identifying Prerequisite Courses in Undergraduate Biology Using Machine Learning

Many undergraduate students who matriculated in Science, Technology, Engineering and Mathematics (STEM) degree programs drop out or switch their major. Previous studies indicate that performance of students in prerequisite courses is important for attrition of students in STEM. This study analyzed demographic information, ACT/SAT score, and performance of students in freshman year courses to develop machine learning models predicting their success in earning a bachelor’s degree in biology. The predictive model based on Random Forest (RF) and Extreme Gradient Boosting (XGBoost) showed a better performance in terms of AUC (Area Under the Curve) with more balanced sensitivity and specificity than Logistic Regression (LR), K-Nearest Neighbor (KNN), and Neural Network (NN) models. An explainable machine learning approach called break-down was employed to identify important freshman year courses that could have a larger impact on student success at the biology degree program and student levels. More important courses identified at the program level can help program coordinators to prioritize their effort in addressing student attrition while more important courses identified at the student level can help academic advisors to provide more personalized, data-driven guidance to students.

Bayesian Coherence Analysis for Microcircuit Structure Learning

Functional microcircuits model the coordinated activity of neurons and play an important role in physiological computation and behaviors. Most existing methods to learn microcircuit structures are correlation-based and often generate dense microcircuits that cannot distinguish between direct and indirect association. We treat microcircuit structure learning as a Markov blanket discovery problem and propose Bayesian Coherence Analysis (BCA) which utilizes a Bayesian network architecture called Bayesian network with inverse-tree structure to efficiently and effectively detect Markov blankets for high-dimensional neural activity data. BCA achieved balanced sensitivity and specificity on simulated data. For the real-world anterior lateral motor cortex study, BCA identified microcircuit subtypes that predicted trial types with an accuracy of 0.92. BCA is a powerful method for microcircuit structure learning.

Performance of a multi-biomarker panel for prediction of cardiovascular event in patients with chronic kidney disease

BackgroundPatients with chronic kidney disease (CKD) undergoing coronary catheterization are at increased risk of cardiovascular events (CVE). Measuring biomarkers before the procedure may guide clinicians in identifying patients at higher risk of future cardiovascular events. MethodsIn this sub-study the Catheter Sampled Blood Archive in Cardiovascular Diseases (CASABLANCA), 927 patients underwent coronary catheterization and were followed up for two years. Using machine learning algorithm and targeted proteomics from samples of patients with CKD, 4 biomarkers (kidney injury molecule-1, N-terminal pro B-type natriuretic peptide, osteopontin, and tissue inhibitor of metalloproteinase-1) were integrated into a prognostic algorithm to predict CVE. Results from the panel are expressed in a graded fashion (CVE higher risk and lower risk) using a data-driven cutoff optimized for balanced sensitivity and specificity. ResultsDuring the 2-year follow-up, 74 CVE were ascertained. 51 (rate: 51/378 = 13.5%) events occurred in stage 1–2 CKD and 23 (rate: 23/68 = 33.8%) events occurred in stage 3–5 CKD. The C-statistic for predicting 2-years cardiovascular events in all 446 patients was 0.77 (0.72, 0.82). The model was well-calibrated (Hosmer-Lemeshow test p-value >0.40). Considering patients at CVE lower-risk within each CKD staging group as a reference, the hazard ratio (95% confidence interval) of cardiovascular events was 2.82 (1.53, 5.22) for CKD stage 1–2/CVE higher-risk, and 8.32 (1.12, 61.76) for CKD stage 3–5/CVE higher-risk. ConclusionMeasuring biomarker panel prior to coronary catheterization may be useful to individualize CVE risk assessment among patients with CKD.

Uncertainty estimation for margin detection in cancer surgery using mass spectrometry.

Rapid evaporative ionization mass spectrometry (REIMS) is an emerging technology for clinical margin detection. Deployment of REIMS depends on construction of reliable deep learning models that can categorize tissue according to its metabolomic signature. Challenges associated with developing these models include the presence of noise during data acquisition and the variance in tissue signatures between patients. In this study, we propose integration of uncertainty estimation in deep models to factor predictive confidence into margin detection in cancer surgery. iKnife is used to collect 693 spectra of cancer and healthy samples acquired from 91 patients during basal cell carcinoma resection. A Bayesian neural network and two baseline models are trained on these data to perform classification as well as uncertainty estimation. The samples with high estimated uncertainty are then removed, and new models are trained using the clean data. The performance of proposed and baseline models, with different ratios of filtered data, is then compared. The data filtering does not improve the performance of the baseline models as they cannot provide reliable estimations of uncertainty. In comparison, the proposed model demonstrates a statistically significant improvement in average balanced accuracy (75.2%), sensitivity (74.1%) and AUC (82.1%) after removing uncertain training samples. We also demonstrate that if highly uncertain samples are predicted and removed from the test data, sensitivity further improves to 88.2%. This is the first study that applies uncertainty estimation to inform model training and deployment for tissue recognition in cancer surgery. Uncertainty estimation is leveraged in two ways: by factoring a measure of input noise in training the models and by including predictive confidence in reporting the outputs. We empirically show that considering uncertainty for model development can help improve the overall accuracy of a margin detection system using REIMS.

Machine learning for classification of hypertension subtypes using multi-omics: A multi-centre, retrospective, data-driven study.

SummaryBackgroundArterial hypertension is a major cardiovascular risk factor. Identification of secondary hypertension in its various forms is key to preventing and targeting treatment of cardiovascular complications. Simplified diagnostic tests are urgently required to distinguish primary and secondary hypertension to address the current underdiagnosis of the latter.MethodsThis study uses Machine Learning (ML) to classify subtypes of endocrine hypertension (EHT) in a large cohort of hypertensive patients using multidimensional omics analysis of plasma and urine samples. We measured 409 multi-omics (MOmics) features including plasma miRNAs (PmiRNA: 173), plasma catechol O-methylated metabolites (PMetas: 4), plasma steroids (PSteroids: 16), urinary steroid metabolites (USteroids: 27), and plasma small metabolites (PSmallMB: 189) in primary hypertension (PHT) patients, EHT patients with either primary aldosteronism (PA), pheochromocytoma/functional paraganglioma (PPGL) or Cushing syndrome (CS) and normotensive volunteers (NV). Biomarker discovery involved selection of disease combination, outlier handling, feature reduction, 8 ML classifiers, class balancing and consideration of different age- and sex-based scenarios. Classifications were evaluated using balanced accuracy, sensitivity, specificity, AUC, F1, and Kappa score.FindingsComplete clinical and biological datasets were generated from 307 subjects (PA=113, PPGL=88, CS=41 and PHT=112). The random forest classifier provided ∼92% balanced accuracy (∼11% improvement on the best mono-omics classifier), with 96% specificity and 0.95 AUC to distinguish one of the four conditions in multi-class ALL-ALL comparisons (PPGL vs PA vs CS vs PHT) on an unseen test set, using 57 MOmics features. For discrimination of EHT (PA + PPGL + CS) vs PHT, the simple logistic classifier achieved 0.96 AUC with 90% sensitivity, and ∼86% specificity, using 37 MOmics features. One PmiRNA (hsa-miR-15a-5p) and two PSmallMB (C9 and PC ae C38:1) features were found to be most discriminating for all disease combinations. Overall, the MOmics-based classifiers were able to provide better classification performance in comparison to mono-omics classifiers.InterpretationWe have developed a ML pipeline to distinguish different EHT subtypes from PHT using multi-omics data. This innovative approach to stratification is an advancement towards the development of a diagnostic tool for EHT patients, significantly increasing testing throughput and accelerating administration of appropriate treatment.FundingEuropean Union's Horizon 2020 Research and Innovation Programme under Grant Agreement No. 633983, Clinical Research Priority Program of the University of Zurich for the CRPP HYRENE (to Z.E. and F.B.), and Deutsche Forschungsgemeinschaft (CRC/Transregio 205/1).