Machine Learning in Healthcare Habehh, Hafsa; Gohel, Suril
Current genomics,
12/2021, Volume:
22, Issue:
4
Journal Article
Peer reviewed
Open access
Recent advancements in Artificial Intelligence (AI) and Machine Learning (ML) technology have brought on substantial strides in predicting and identifying health emergencies, disease populations, and ...disease state and immune response, amongst a few. Although, skepticism remains regarding the practical application and interpretation of results from ML-based approaches in healthcare settings, the inclusion of these approaches is increasing at a rapid pace. Here we provide a brief overview of machine learning-based approaches and learning algorithms including supervised, unsupervised, and reinforcement learning along with examples. Second, we discuss the application of ML in several healthcare fields, including radiology, genetics, electronic health records, and neuroimaging. We also briefly discuss the risks and challenges of ML application to healthcare such as system privacy and ethical concerns and provide suggestions for future applications.
Studies of resting-state fMRI have shown that blood oxygen level dependent (BOLD) signals giving rise to temporal correlation across voxels (or regions) are dominated by low-frequency fluctuations in ...the range of ∼ 0.01-0.1 Hz. These low-frequency fluctuations have been further divided into multiple distinct frequency bands (slow-5 and -4) based on earlier neurophysiological studies, though low sampling frequency of fMRI (∼ 0.5 Hz) has substantially limited the exploration of other known frequency bands of neurophysiological origins (slow-3, -2, and -1). In this study, we used resting-state fMRI data acquired from 21 healthy subjects at a higher sampling frequency of 1.5 Hz to assess the presence of resting-state functional connectivity (RSFC) across multiple frequency bands: slow-5 to slow-1. The effect of different frequency bands on spatial extent and connectivity strength for known resting-state networks (RSNs) was also evaluated. RSNs were derived using independent component analysis and seed-based correlation. Commonly known RSNs, such as the default mode, the fronto-parietal, the dorsal attention, and the visual networks, were consistently observed at multiple frequency bands. Significant inter-hemispheric connectivity was observed between each seed and its contra lateral brain region across all frequency bands, though overall spatial extent of seed-based correlation maps decreased in slow-2 and slow-1 frequency bands. These results suggest that functional integration between brain regions at rest occurs over multiple frequency bands and RSFC is a multiband phenomenon. These results also suggest that further investigation of BOLD signal in multiple frequency bands for related cognitive processes should be undertaken.
Approximately 700,000 people in the USA have chronic kidney disease requiring dialysis. Protein-energy wasting (PEW), a condition of advanced catabolism, contributes to three-year survival rates of ...50%. PEW occurs at all levels of Body Mass Index (BMI) but is devastating for those people at the extremes. Treatment for PEW depends on an accurate understanding of energy expenditure. Previous research established that current methods of identifying PEW and assessing adequate treatments are imprecise. This includes disease-specific equations for estimated resting energy expenditure (eREE). In this study, we applied machine learning (ML) modelling techniques to a clinical database of dialysis patients. We assessed the precision of the ML algorithms relative to the best-performing traditional equation, the MHDE.
This was a secondary analysis of the Rutgers Nutrition and Kidney Database. To build the ML models we divided the population into test and validation sets. Eleven ML models were run and optimized, with the best three selected by the lowest root mean squared error (RMSE) from measured REE. Values for eREE were generated for each ML model and for the MHDE. We compared precision using Bland-Altman plots.
Individuals were 41.4% female and 82.0% African American. The mean age was 56.4 ± 11.1 years, and the median BMI was 28.8 (IQR = 24.8 − 34.0) kg/m
2
. The best ML models were SVR, Linear Regression and Elastic net with RMSE of 103.6 kcal, 119.0 kcal and 121.1 kcal respectively. The SVR demonstrated the greatest precision, with 91.2% of values falling within acceptable limits. This compared to 47.1% for the MHDE. The models using non-linear techniques were precise across extremes of BMI.
ML improves precision in calculating eREE for dialysis patients, including those most vulnerable for PEW. Further development for clinical use is a priority.
Graph-based computational network analysis has proven a powerful tool to quantitatively characterize functional architectures of the brain. However, the test-retest (TRT) reliability of graph metrics ...of functional networks has not been systematically examined. Here, we investigated TRT reliability of topological metrics of functional brain networks derived from resting-state functional magnetic resonance imaging data. Specifically, we evaluated both short-term (<1 hour apart) and long-term (>5 months apart) TRT reliability for 12 global and 6 local nodal network metrics. We found that reliability of global network metrics was overall low, threshold-sensitive and dependent on several factors of scanning time interval (TI, long-term>short-term), network membership (NM, networks excluding negative correlations>networks including negative correlations) and network type (NT, binarized networks>weighted networks). The dependence was modulated by another factor of node definition (ND) strategy. The local nodal reliability exhibited large variability across nodal metrics and a spatially heterogeneous distribution. Nodal degree was the most reliable metric and varied the least across the factors above. Hub regions in association and limbic/paralimbic cortices showed moderate TRT reliability. Importantly, nodal reliability was robust to above-mentioned four factors. Simulation analysis revealed that global network metrics were extremely sensitive (but varying degrees) to noise in functional connectivity and weighted networks generated numerically more reliable results in compared with binarized networks. For nodal network metrics, they showed high resistance to noise in functional connectivity and no NT related differences were found in the resistance. These findings provide important implications on how to choose reliable analytical schemes and network metrics of interest.
There is a growing interest in studies of human brain networks using resting-state functional magnetic resonance imaging (fMRI). However, it is unclear whether and how brain networks measured during ...the resting-state exhibit comparable properties to brain networks during task performance. In the present study, we investigated meta-analytic coactivation patterns among brain regions based upon published neuroimaging studies, and compared the coactivation network configurations with those in the resting-state network. The strength of resting-state functional connectivity between two regions were strongly correlated with the coactivation strength. However, the coactivation network showed greater global efficiency, smaller mean clustering coefficient, and lower modularity compared with the resting-state network, which suggest a more efficient global information transmission and between system integrations during task performing. Hub shifts were also observed within the thalamus and the left inferior temporal cortex. The thalamus and the left inferior temporal cortex exhibited higher and lower degrees, respectively in the coactivation network compared with the resting-state network. These results shed light regarding the reconfiguration of the brain networks between task and resting-state conditions, and highlight the role of the thalamus in change of network configurations in task vs. rest.
Various studies have indicated that the thalamus is involved in controlling both cortico-cortical information flow and cortical communication with the rest of the brain. Detailed anatomy and ...functional connectivity patterns of the thalamocortical system are essential to understanding the cortical organization and pathophysiology of a wide range of thalamus-related neurological and neuropsychiatric diseases. The current study used resting-state fMRI to investigate the topography of the human thalamocortical system from a functional perspective. The thalamus-related cortical networks were identified by performing independent component analysis on voxel-based thalamic functional connectivity maps across a large group of subjects. The resulting functional brain networks were very similar to well-established resting-state network maps. Using these brain network components in a spatial regression model with each thalamic voxel’s functional connectivity map, we localized the thalamic subdivisions related to each brain network. For instance, the medial dorsal nucleus was shown to be associated with the default mode, the bilateral executive, the medial visual networks; and the pulvinar nucleus was involved in both the dorsal attention and the visual networks. These results revealed that a single nucleus may have functional connections with multiple cortical regions or even multiple functional networks, and may be potentially related to the function of mediation or modulation of multiple cortical networks. This observed organization of thalamocortical system provided a reference for studying the functions of thalamic sub-regions. The importance of intrinsic connectivity-based mapping of the thalamocortical relationship is discussed, as well as the applicability of the approach for future studies.
Convergence insufficiency (CI) is the most common binocular vision problem, associated with blurred/double vision, headaches, and sore eyes that are exacerbated when doing prolonged near work, such ...as reading. The Convergence Insufficiency Neuro-mechanism Adult Population Study (NCT03593031) investigates the mechanistic neural differences between 50 binocularly normal controls (BNC) and 50 symptomatic CI participants by examining the fast and slow fusional disparity vergence systems. The fast fusional system is preprogrammed and is assessed with convergence peak velocity. The slow fusional system optimizes vergence effort and is assessed by measuring the phoria adaptation magnitude and rate. For the fast fusional system, significant differences are observed between the BNC and CI groups for convergence peak velocity, final position amplitude, and functional imaging activity within the secondary visual cortex, right cuneus, and oculomotor vermis. For the slow fusional system, the phoria adaptation magnitude and rate, and the medial cuneus functional activity, are significantly different between the groups. Significant correlations are observed between vergence peak velocity and right cuneus functional activity (p = 0.002) and the rate of phoria adaptation and medial cuneus functional activity (p = 0.02). These results map the brain-behavior of vergence. Future therapeutic interventions may consider implementing procedures that increase cuneus activity for this debilitating disorder.
Heterogeneous response to Enzalutamide, a second-generation androgen receptor signaling inhibitor, is a central problem in castration-resistant prostate cancer (CRPC) management. Genome-wide systems ...investigation of mechanisms that govern Enzalutamide resistance promise to elucidate markers of heterogeneous treatment response and salvage therapies for CRPC patients. Focusing on the de novo role of MYC as a marker of Enzalutamide resistance, here we reconstruct a CRPC-specific mechanism-centric regulatory network, connecting molecular pathways with their upstream transcriptional regulatory programs. Mining this network with signatures of Enzalutamide response identifies NME2 as an upstream regulatory partner of MYC in CRPC and demonstrates that NME2-MYC increased activities can predict patients at risk of resistance to Enzalutamide, independent of co-variates. Furthermore, our experimental investigations demonstrate that targeting MYC and its partner NME2 is beneficial in Enzalutamide-resistant conditions and could provide an effective strategy for patients at risk of Enzalutamide resistance and/or for patients who failed Enzalutamide treatment.
Conscious attempts to regulate alcohol and drug use are often undermined by automatic attention and arousal processes that are activated in the context of salient cues. Response to these cues ...involves body and brain signals that are linked via dynamic feedback loops, yet no studies have targeted the cardiovascular system as a potential conduit to alter automatic neural processes that maintain cue salience. This proof-of-concept study examined within-person changes in neural response to parallel but unique sets of visual alcohol-related cues at two points in time: prior to versus following a brief behavioral intervention. The active intervention was resonance breathing, a rhythmical breathing task paced at 0.1 Hz (6 breaths per minute) that helps normalize neurocardiac feedback. The control intervention was a low-demand cognitive task. Functional magnetic resonance imaging (fMRI) was used to assess changes in brain response to the cues presented before (A1) and after (A2) the intervention in 41 emerging adult men and women with varying drinking behaviors. The resonance breathing group exhibited significantly less activation to A2 cues compared with A1 cues in left inferior and superior lateral occipital cortices, right inferior lateral occipital cortex, bilateral occipital pole, and temporal occipital fusiform cortices. This group also showed significantly greater activation to A2 cues compared with A1 cues in medial prefrontal, anterior and posterior cingulate, and precuneus cortices, paracingulate, and lingual gyri. The control group showed no significant changes. Thus, following resonance breathing, activation in brain regions involved in visual processing of cues was reduced, while activation in brain areas implicated in behavioral control, internally directed cognition, and brain–body integration was increased. These findings provide preliminary evidence that manipulation of the cardiovascular system with resonance breathing alters neural activation in a manner theoretically consistent with a dampening of automatic sensory input and strengthening of higher-level cognitive processing.
There is a paucity of research on treatment-related neurotoxicity in older adults with multiple myeloma (MM) treated with high-dose chemotherapy (HDC) and autologous SCT (HDC/ASCT), despite the ...increasing use of this regimen. We examined resting state functional connectivity (RSFC), gray matter (GM) volume, neurocognitive function (NF), and proinflammatory cytokines (PCy) in older patients with MM pre- and post-HDC/ASCT. Eighteen patients underwent MRI, NF tests, and serum PCy measurements prior to HDC/ASCT, and fifteen patients completed a follow up five-months post-HDC/ASCT. There were significant decreases in RSFC post-HDC/ASCT in (1) the central executive network (CEN) involving the left dorsolateral prefrontal cortex and right posterior parietal cortex (p = 0.022) and (2) the CEN involving the right posterior parietal cortex and the salience network involving the right dorsal anterior cingulate cortex (p = 0.029). There were no significant changes in GM or NF, except for improvements in attention (Digit Span Backward, p = 0.03). There were significant increases in several PCy post-HDC/ASCT (p ≤ 0.05). In conclusion, RSFC decreased in frontal, parietal, and cingulate cortices post-HDC/ASCT, NF was relatively stable, and several PCy increased. These findings are congruent with other studies in cancer patients and provide supporting evidence for the vulnerability of frontoparietal regions to chemotherapy’s adverse effects.