A machine learning approach to identify groups of patients with hematological malignant disorders.

BACKGROUND AND OBJECTIVE: Vaccination against SARS-CoV-2 in immunocompromised patients with hematologic malignancies (HM) is crucial to reduce the severity of COVID-19. Despite vaccination efforts, over a third of HM patients remain unresponsive, increasing their risk of severe breakthrough infections. This study aims to leverage machine learning's adaptability to COVID-19 dynamics, efficiently selecting patient-specific features to enhance predictions and improve healthcare strategies. Highlighting the complex COVID-hematology connection, the focus is on interpretable machine learning to provide valuable insights to clinicians and biologists. METHODS: The study evaluated a dataset with 1166 patients with hematological diseases. The output was the achievement or non-achievement of a serological response after full COVID-19 vaccination. Various machine learning methods were applied, with the best model selected based on metrics such as the Area Under the Curve (AUC), Sensitivity, Specificity, and Matthew Correlation Coefficient (MCC). Individual SHAP values were obtained for the best model, and Principal Component Analysis (PCA) was applied to these values. The patient profiles were then analyzed within identified clusters. RESULTS: Support vector machine (SVM) emerged as the best-performing model. PCA applied to SVM-derived SHAP values resulted in four perfectly separated clusters. These clusters are characterized by the proportion of patients that generate antibodies (PPGA). Cluster 1, with the second-highest PPGA (69.91%), included patients with aggressive diseases and factors contributing to increased immunodeficiency. Cluster 2 had the lowest PPGA (33.3%), but the small sample size limited conclusive findings. Cluster 3, representing the majority of the population, exhibited a high rate of antibody generation (84.39%) and a better prognosis compared to cluster 1. Cluster 4, with a PPGA of 66.33%, included patients with B-cell non-Hodgkin's lymphoma on corticosteroid therapy. CONCLUSIONS: The methodology successfully identified four separate patient clusters using Machine Learning and Explainable AI (XAI). We then analyzed each cluster based on the percentage of HM patients who generated antibodies after COVID-19 vaccination. The study suggests the methodology's potential applicability to other diseases, highlighting the importance of interpretable ML in healthcare research and decision-making.

Prediction of manifest refraction using machine learning ensemble models on wavefront aberrometry data.

PURPOSE: To assess the performance of machine learning (ML) ensemble models for predicting patient subjective refraction (SR) using demographic factors, wavefront aberrometry data, and measurement quality related metrics taken with a low-cost portable autorefractor. METHODS: Four ensemble models were evaluated for predicting individual power vectors (M, J0, and J45) corresponding to the eyeglass prescription of each patient. Those models were random forest regressor (RF), gradient boosting regressor (GB), extreme gradient boosting regressor (XGB), and a custom assembly model (ASB) that averages the first three models. Algorithms were trained on a dataset of 1244 samples and the predictive power was evaluated with 518 unseen samples. Variables used for the prediction were age, gender, Zernike coefficients up to 5th order, and pupil related metrics provided by the autorefractor. Agreement with SR was measured using Bland-Altman analysis, overall prediction error, and percentage of agreement between the ML predictions and subjective refractions for different thresholds (0.25 D, 0.5 D). RESULTS: All models considerably outperformed the predictions from the autorefractor, while ASB obtained the best results. The accuracy of the predictions for each individual power vector component was substantially improved resulting in a ± 0.63 D, ±0.14D, and ±0.08 D reduction in the 95% limits of agreement of the error distribution for M, J0, and J45, respectively. The wavefront-aberrometry related variables had the biggest impact on the prediction, while demographic and measurement quality-related features showed a heterogeneous but consistent predictive value. CONCLUSIONS: These results suggest that ML is effective for improving precision in predicting patient's SR from objective measurements taken with a low-cost portable device.

Machine Learning for Mortality Analysis in Patients with COVID-19.

This paper analyzes a sample of patients hospitalized with COVID-19 in the region of Madrid (Spain). Survival analysis, logistic regression, and machine learning techniques (both supervised and unsupervised) are applied to carry out the analysis where the endpoint variable is the reason for hospital discharge (home or deceased). The different methods applied show the importance of variables such as age, O2 saturation at Emergency Rooms (ER), and whether the patient comes from a nursing home. In addition, biclustering is used to globally analyze the patient-drug dataset, extracting segments of patients. We highlight the validity of the classifiers developed to predict the mortality, reaching an appreciable accuracy. Finally, interpretable decision rules for estimating the risk of mortality of patients can be obtained from the decision tree, which can be crucial in the prioritization of medical care and resources.

Cerebral hunger maps in rodents and humans by diffusion weighted MRI.

We design, implement and validate a novel image processing strategy to obtain in vivo maps of hunger stimulation in the brain of mice, rats and humans, combining Diffusion Weighted Magnetic Resonance Imaging (DWI) datasets from fed and fasted subjects. Hunger maps were obtained from axial/coronal (rodents/humans) brain sections containing the hypothalamus and coplanar cortico-limbic structures using Fisher's Discriminant Analysis of the combined voxel ensembles from both feeding situations. These maps were validated against those provided by the classical mono-exponential diffusion model as applied over the same subjects and conditions. Mono-exponential fittings revealed significant Apparent Diffusion Coefficient (ADC) decreases through the brain regions stimulated by hunger, but rigorous parameter estimations imposed the rejection of considerable number of pixels. The proposed approach avoided pixel rejections and provided a representation of the combined DWI dataset as a pixel map of the "Hunger Index" (HI), a parameter revealing the hunger score of every pixel. The new methodology proved to be robust both, by yielding consistent results with classical ADC maps and, by reproducing very similar HI maps when applied to newly acquired rodent datasets. ADC and HI maps demonstrated similar patterns of activation by hunger in hypothalamic and cortico-limbic structures of the brain of rodents and humans, albeit with different relative intensities, rodents showing more intense activations by hunger than humans, for similar fasting periods. The proposed methodology may be easily extended to other feeding paradigms or even to alternative imaging methods.

Hypothalamic metabolic compartmentation during appetite regulation as revealed by magnetic resonance imaging and spectroscopy methods.

We review the role of neuroglial compartmentation and transcellular neurotransmitter cycling during hypothalamic appetite regulation as detected by Magnetic Resonance Imaging (MRI) and Spectroscopy (MRS) methods. We address first the neurochemical basis of neuroendocrine regulation in the hypothalamus and the orexigenic and anorexigenic feed-back loops that control appetite. Then we examine the main MRI and MRS strategies that have been used to investigate appetite regulation. Manganese-enhanced magnetic resonance imaging (MEMRI), Blood oxygenation level-dependent contrast (BOLD), and Diffusion-weighted magnetic resonance imaging (DWI) have revealed Mn(2+) accumulations, augmented oxygen consumptions, and astrocytic swelling in the hypothalamus under fasting conditions, respectively. High field (1)H magnetic resonance in vivo, showed increased hypothalamic myo-inositol concentrations as compared to other cerebral structures. (1)H and (13)C high resolution magic angle spinning (HRMAS) revealed increased neuroglial oxidative and glycolytic metabolism, as well as increased hypothalamic glutamatergic and GABAergic neurotransmissions under orexigenic stimulation. We propose here an integrative interpretation of all these findings suggesting that the neuroendocrine regulation of appetite is supported by important ionic and metabolic transcellular fluxes which begin at the tripartite orexigenic clefts and become extended spatially in the hypothalamus through astrocytic networks becoming eventually MRI and MRS detectable.

Imaging hypothalamic activity using diffusion weighted magnetic resonance imaging in the mouse and human brain.

Hypothalamic appetite regulation is a vital homeostatic process underlying global energy balance in animals and humans, its disturbances resulting in feeding disorders with high morbidity and mortality. The objective evaluation of appetite remains difficult, very often restricted to indirect measurements of food intake and body weight. We report here, the direct, non-invasive visualization of hypothalamic activation by fasting using diffusion weighted magnetic resonance imaging, in the mouse brain as well as in a preliminary study in the human brain. The brain of fed or fasted mice or humans were imaged at 7 or 1.5 Tesla, respectively, by diffusion weighted magnetic resonance imaging using a complete range of b values (10

A new information processing measure for adaptive complex systems.

This paper presents an implementation-independent measure of the amount of information processing performed by (part of) an adaptive system which depends on the goal to be performed by the overall system. This new measure gives rise to a theoretical framework under which several classical supervised and unsupervised learning algorithms fall and, additionally, new efficient learning algorithms can be derived. In the context of neural networks, the framework of information theory strives to design neurally inspired structures from which complex functionality should emerge. Yet, classical measures of information have not taken an explicit account of some of the fundamental concepts in brain theory and neural computation, namely that optimal coding depends on the specific task(s) to be solved by the system and that goal orientedness also depends on extracting relevant information from the environment to be able to affect it in the desired way. We present a new information processing measure that takes into account both the extraction of relevant information and the reduction of spurious information for the task to be solved by the system. This measure is implementation-independent and therefore can be used to analyze and design different adaptive systems. Specifically, we show its application for learning perceptrons, decision trees and linear autoencoders.

Why do olfactory neurons have unspecific receptive fields?

Biological olfactory neurons are deployed as a population, most responding to a large variety of chemical compounds, that is, they possess unspecific receptive fields. The question of whether this unspecificity results from some physical constraint placed upon chemical transduction, or on the other hand, is beneficial to system performance is unclear. In this paper we employ the notion of Fisher information to address this question by quantifying how both the distribution and the tunings of the receptive fields within olfactory receptor populations affect the optimal estimation performance of the system. Our results show that overlapping sensory neuron tunings that respond to common chemical compounds have better estimation performance than perfectly specific tunings. Our results suggest two phenomena that might represent general principles of organization within biological sensory systems responding to multiple stimuli: maximization of the diversity of tunings and homogeneity in the distribution of these different receptive fields across the stimulus space (independent of the statistics of the input stimuli). Our model predicts that a local randomized mechanism controlling receptor specificities generates optimal multidimensional stimulus estimation, for which there is some experimental evidence from the biology.