Multinomial logistic regression ensembles

Machine learning prediction of the failure of high-flow nasal oxygen therapy in patients with acute respiratory failure

Acute respiratory failure (ARF) is a prevalent and serious condition in intensive care unit (ICU), often associated with high mortality rates. High-flow nasal oxygen (HFNO) therapy has gained popularity for treating ARF in recent years. However, there is a limited understanding of the factors that predict HFNO failure in ARF patients. This study aimed to explore early indicators of HFNO failure in ARF patients, utilizing machine learning (ML) algorithms to more accurately pinpoint individuals at elevated risk of HFNO failure. Utilizing ML algorithms, we developed seven predictive models. Their performance was evaluated using various metrics, including the area under the receiver operating characteristic curve, calibration curve, and precision recall curve. The study enrolled 700 patients, with 490 in the training group and 210 in the validation group. The overall HFNO failure rate was 14.1% among the 700 patients. The ML algorithms demonstrated robust performance in our study. This research underscores the potential of ML techniques in creating clinically relevant models for predicting HFNO outcomes in ARF patients. These models could play a pivotal role in enhancing the risk management of HFNO, leading to more patient-centered and personalized care approaches.

A multidomain bio-inspired feature extraction and selection model for diabetic retinopathy severity classification: an ensemble learning approach

Diabetes retinopathy (DR) is one of the leading causes of blindness globally. Early detection of this condition is essential for preventing patients' loss of eyesight caused by diabetes mellitus being untreated for an extended period. This paper proposes the design of an augmented bioinspired multidomain feature extraction and selection model for diabetic retinopathy severity estimation using an ensemble learning process. The proposed approach initiates by identifying DR severity levels from retinal images that segment the optical disc, macula, blood vessels, exudates, and hemorrhages using an adaptive thresholding process. Once the images are segmented, multidomain features are extracted from the retinal images, including frequency, entropy, cosine, gabor, and wavelet components. These data were fed into a novel Modified Moth Flame Optimization-based feature selection method that assisted in optimal feature selection. Finally, an ensemble model using various ML (machine learning) algorithms, which included Naive Bayes, K-Nearest Neighbours, Support Vector Machine, Multilayer Perceptron, Random Forests, and Logistic Regression were used to identify the various severity complications of DR. The experiments on different openly accessible data sources have shown that the proposed method outperformed conventional methods and achieved an Accuracy of 96.5% in identifying DR severity levels.

Early-life residential exposure to moisture damage is associated with persistent wheezing in a Finnish birth cohort

BACKGROUND AND AIMS

Moisture damage increases the risk for respiratory disorders in childhood. Our aim was to determine whether early age residential exposure to inspector-observed moisture damage or mold is associated with different wheezing phenotypes later in childhood.

METHODS

Building inspections were performed by civil engineers, in a standardized manner, in the children's homes-mostly single family and row houses (N = 344)-in the first year of life. The children were followed up with repeated questionnaires until the age of 6 years and wheezing phenotypes-never/infrequent, transient, intermediate, late onset, and persistent-were defined using latent class analyses. The multinomial logistic regression model was used for statistical analysis.

RESULTS

A total of 63% (n = 218) had infrequent or no wheeze, 23% (n = 80) had transient and 9.6% (n = 21) had a persistent wheeze. Due to the low prevalence, results for intermediate (3.8%, n = 13) and late-onset wheeze (3.5%, n = 12) were not further evaluated. Most consistent associations were observed with the persistent wheeze phenotype with an adjusted odds ratio (95% confidence intervals) 2.04 (0.67-6.18) for minor moisture damage with or without mold spots (present in 23.8% of homes) and 3.68 (1.04-13.05) for major damage or any moisture damage with visible mold in a child's main living areas (present in 13.4% of homes). Early-age moisture damage or mold in the kitchen was associated with transient wheezing.

CONCLUSION

At an early age, residential exposure to moisture damage or mold, can be dose-dependently associated especially with persistent wheezing phenotype later in childhood.

Risk Factors for the Severity of Complications in Minimally Invasive Total Gastrectomy for Gastric Cancer: a Retrospective Cohort Study

PURPOSE

Minimally invasive gastrectomy is a promising surgical method with well-known benefits, including reduced postoperative complications. However, for total gastrectomy of gastric cancers, this approach does not significantly reduce the risk of complications. Therefore, we aimed to evaluate the incidence and risk factors for the severity of complications associated with minimally invasive total gastrectomy for gastric cancer.

MATERIALS AND METHODS

The study included 392 consecutive patients with gastric cancer who underwent either laparoscopic or robotic total gastrectomy between 2011 and 2019. Clinicopathological and operative characteristics were assessed to determine the features related to postoperative complications after minimally invasive total gastrectomy. Binomial and multinomial logistic regression models were used to identify the risk factors for overall complications and mild and severe complications, respectively.

RESULTS

Of 103 (26.3%) patients experiencing complications, 66 (16.8%) and 37 (9.4%) developed mild and severe complications, respectively. On multivariate multinomial regression analysis, independent predictors of severe complications included obesity (OR, 2.56; 95% CI, 1.02-6.43; P=0.046), advanced stage (OR, 2.90; 95% CI, 1.13-7.43; P=0.026), and more intraoperative bleeding (OR, 1.04; 95% CI, 1.02-1.06; P=0.001). Operation time was the only independent risk factor for mild complications (OR, 1.06; 95% CI, 1.001-1.13; P=0.047).

CONCLUSIONS

The risk factors for mild and severe complications were associated with surgery, indicating surgical difficulty. Surgeons should be aware of these potential risks that are related to the severity of complications so as to reduce surgery-related complications after minimally invasive total gastrectomy for gastric cancer.

Detecting Depression Using an Ensemble Logistic Regression Model Based on Multiple Speech Features

Early intervention for depression is very important to ease the disease burden, but current diagnostic methods are still limited. This study investigated automatic depressed speech classification in a sample of 170 native Chinese subjects (85 healthy controls and 85 depressed patients). The classification performances of prosodic, spectral, and glottal speech features were analyzed in recognition of depression. We proposed an ensemble logistic regression model for detecting depression (ELRDD) in speech. The logistic regression, which was superior in recognition of depression, was selected as the base classifier. This ensemble model extracted many speech features from different aspects and ensured diversity of the base classifier. ELRDD provided better classification results than the other compared classifiers. A technique for identifying depression based on ELRDD, ELRDD-E, was here suggested and tested. It offered encouraging outcomes, revealing a high accuracy level of 75.00% for females and 81.82% for males, as well as an advantageous sensitivity/specificity ratio of 79.25%/70.59% for females and 78.13%/85.29% for males.

Assessment of IrisPlex-based multiplex for eye and skin color prediction with application to a Portuguese population

DNA phenotyping research is one of the most emergent areas of forensic genetics. Predictions of externally visible characteristics are possible through analysis of single nucleotide polymorphisms. These tools can provide police with "intelligence" in cases where there are no obvious suspects and unknown biological samples found at the crime scene do not result in any criminal DNA database hits. IrisPlex, an eye color prediction assay, revealed high prediction rates for blue and brown eye color in European populations. However, this is less predictive in some non-European populations, probably due to admixing. When compared to other European countries, Portugal has a relatively admixed population, resulting from a genetic influx derived from its proximity to and historical relations with numerous African territories. The aim of this work was to evaluate the utility of IrisPlex in the Portuguese population. Furthermore, the possibility of supplementing this multiplex with additional markers to also achieve skin color prediction within this population was evaluated. For that, IrisPlex was augmented with additional SNP loci. Eye and skin color prediction was estimated using the multinomial logistic regression and binomial logistic regression models, respectively. The results demonstrated eye color prediction accuracies of the IrisPlex system of 90 and 60% for brown and blue eye color, respectively, and 77% for intermediate eye color, after allele frequency adjustment. With regard to skin color, it was possible to achieve a prediction accuracy of 93%. In the future, phenotypic determination multiplexes must include additional loci to permit skin color prediction as presented in this study as this can be an advantageous tool for forensic investigation.

Temporal trends in the Swedish HIV-1 epidemic: increase in non-B subtypes and recombinant forms over three decades

BACKGROUND

HIV-1 subtype B (HIV-1B) still dominates in resource-rich countries but increased migration contributes to changes in the global subtype distribution. Also, spread of non-B subtypes within such countries occurs. The trend of the subtype distribution from the beginning of the epidemic in the country has earlier not been reported in detail. Thus the primary objective of this study is to describe the temporal trend of the subtype distribution from the beginning of the HIV-1 epidemic in Sweden over three decades.

METHODS

HIV-1 pol sequences from patients (n = 3967) diagnosed in Sweden 1983-2012, corresponding to >40% of patients ever diagnosed, were re-subtyped using several automated bioinformatics tools. The temporal trends of subtypes and recombinants during three decades were described by a multinomial logistic regression model.

RESULTS

All eleven group M HIV-1 subtypes and sub-subtypes (78%), 17 circulating recombinant forms (CRFs) (19%) and 32 unique recombinants forms (URF) (3%) were identified. When all patients were analysed, there was an increase of newly diagnosed HIV-1C (RR, 95%CI: 1.10, 1.06-1.14), recombinants (1.20, 1.17-1.24) and other pure subtypes (1.11, 1.07-1.16) over time compared to HIV-1B. The same pattern was found when all patients infected in Sweden (n = 1165) were analysed. Also, for MSM patients infected in Sweden (n = 921), recombinant forms and other pure subtypes increased.

SIGNIFICANCE

Sweden exhibits one of the most diverse subtype epidemics outside Africa. The increase of non-B subtypes is due to migration and to a spread among heterosexually infected patients and MSM within the country. This viral heterogeneity may become a hotspot for development of more diverse and complex recombinant forms if the epidemics converge.