Exploration of machine learning techniques in predicting multiple sclerosis disease course

Zhao, Yijun; Healy, Brian C.; Rotstein, Dalia; Guttmann, Charles R.G.; Bakshi, Rohit; Weiner, Howard L.; Brodley, Carla E.; Chitnis, Tanuja

doi:10.1371/journal.pone.0174866

Cited by 135 publications

(119 citation statements)

References 31 publications

Supporting

Mentioning

111

Contrasting

Unclassified

Order By: Relevance

“…Stratification of individuals based on their EEG characteristics continues to move beyond group-level comparisons and towards sophisticated methodology to identify prognostic neuromarkers (31,66). The present study is consistent with the 'big data in psychiatry' approach that has potential to greatly benefit clinical research and practice (66,67,68,69). Our findings support the Specificity (%) 49 54 49 Note.…”

Section: Markovska-simoska and Pop-jordanova (32) Reported 68% Accurasupporting

confidence: 84%

EEG spectral power, but not theta/beta ratio, is a neuromarker for adult ADHD

Kiiski

Bennett

Rueda‐Delgado

et al. 2019

Preprint

View full text Add to dashboard Cite

Background. Adults with attention-deficit/hyperactivity disorder (ADHD) have been described as having altered resting-state electroencephalographic (EEG) spectral power and theta/beta ratio (TBR).However, a recent review (Pulini et al. 2018) identified methodological errors in neuroimaging, including EEG, ADHD classification studies. Therefore, the specific EEG neuromarkers of adult ADHD remain to be identified, as do the EEG characteristics that mediate between genes and behavior (mediational endophenotypes). Methods. Resting-state eyes-open and eyes-closed EEG were measured from 38 adults with ADHD, 45 first-degree relatives of people with ADHD and 51 unrelated controls. A machine learning classification analysis using penalized logistic regression (Elastic Net) examined if EEG spectral power (1-45 Hz) and TBR could classify participants into ADHD, first-degree relatives and/or control groups. Random-label permutation was used to quantify any bias in the analysis. Results. Eyes-open absolute and relative EEG power distinguished ADHD from control participants (area under receiver operating characteristic = .71-.77). The best predictors of ADHD status were increased power in delta, theta and low-alpha over centro-parietal regions, and in frontal low-beta and parietal mid-beta. TBR did not classify ADHD status. Elevated eyes-open power in delta, theta, low-alpha and low-beta distinguished first-degree relatives from controls (area under receiver operating characteristic = .68-.72), suggesting that these features may be a mediational endophenotype for adult ADHD.Conclusions. Resting-state EEG spectral power may be a neuromarker and mediational endophenotype of adult ADHD. These results did not support TBR as a diagnostic neuromarker for ADHD. It is possible that TBR is a characteristic of childhood ADHD.

show abstract

Section: Markovska-simoska and Pop-jordanova (32) Reported 68% Accurasupporting

confidence: 84%

EEG spectral power, but not theta/beta ratio, is a neuromarker for adult ADHD

Kiiski

Bennett

Rueda‐Delgado

et al. 2019

Preprint

View full text Add to dashboard Cite

show abstract

“…Further analyses will explore effects of specific treatments on NfL levels 12, 36, 37. Future studies should validate our findings and explore the additional predictors of long‐term disease course and MRI outcomes in multivariate and machine learning models38.…”

Section: Discussionmentioning

confidence: 74%

Neurofilament light chain serum levels correlate with 10‐year MRI outcomes in multiple sclerosis

Chitnis

Gonzalez

Healy

et al. 2018

Ann Clin Transl Neurol

Self Cite

119

141

View full text Add to dashboard Cite

ObjectiveTo assess the value of annual serum neurofilament light (NfL) measures in predicting 10‐year clinical and MRI outcomes in multiple sclerosis (MS).MethodsWe identified patients in our center's Comprehensive Longitudinal Investigations in MS at Brigham and Women's Hospital (CLIMB) study enrolled within 5 years of disease onset, and with annual blood samples up to 10 years (n = 122). Serum NfL was measured using a single molecule array (SIMOA) assay. An automated pipeline quantified brain T2 hyperintense lesion volume (T2LV) and brain parenchymal fraction (BPF) from year 10 high‐resolution 3T MRI scans. Correlations between averaged annual NfL and 10‐year clinical/MRI outcomes were assessed using Spearman's correlation, univariate, and multivariate linear regression models.ResultsAveraged annual NfL values were negatively associated with year 10 BPF, which included averaged year 1–5 NfL values (unadjusted P < 0.01; adjusted analysis P < 0.01), and averaged values through year 10. Linear regression analyses of averaged annual NfL values showed multiple associations with T2LV, specifically averaged year 1–5 NfL (unadjusted P < 0.01; adjusted analysis P < 0.01). Approximately 15–20% of the BPF variance and T2LV could be predicted from early averaged annual NfL levels. Also, averaged annual NfL levels with fatigue score worsening between years 1 and 10 showed statistically significant associations. However, averaged NfL measurements were not associated with year 10 EDSS, SDMT or T25FW in this cohort.InterpretationSerum NfL measured during the first few years after the clinical onset of MS contributed to the prediction of 10‐year MRI brain lesion load and atrophy.

show abstract

“…Beyond NLP, advances in machine learning have enabled new approaches for prediction of disease onset and future diseases [49]. This is in addition to the [81] exploration of machine learning techniques in predicting multiple sclerosis disease course. Another application is in image recognition for classification of radiology and pathology images [58].…”

Section: Issues With Baseline / Ground Truth Valuementioning

confidence: 99%

A Solution to the Hyper Complex, Cross Domain Reality of Artificial Intelligence: The Hierarchy of AI

Kear¹,

Folkes²

2020

IJACSA

View full text Add to dashboard Cite

Artificial Intelligence (AI) is an umbrella term used to describe machine-based forms of learning. This can encapsulate anything from Siri, Apple's smartphone-based assistant, to Tesla's autonomous vehicles (self-driving cars). At present, there are no set criteria to classify AI. The implications of which include public uncertainty, corporate scepticism, diminished confidence, insufficient funding and limited progress. Current substantial challenges exist with AI such as the use of combinationally large search space, prediction errors against ground truth values, the use of quantum error correction strategies. These are discussed in addition to fundamental data issues across collection, sample error and quality. The concept of cross realms and domains used to inform AI, is considered. Furthermore there is the issue of the confusing range of current AI labels. This paper aims to provide a more consistent form of classification, to be used by institutions and organisations alike, as they endeavour to make AI part of their practice. In turn, this seeks to promote transparency and increase trust. This has been done through primary research, including a panel of data scientists / experts in the field, and through a literature review on existing research. The authors propose a model solution in that of the Hierarchy of AI.

show abstract

Exploration of machine learning techniques in predicting multiple sclerosis disease course

Cited by 135 publications

References 31 publications

EEG spectral power, but not theta/beta ratio, is a neuromarker for adult ADHD

EEG spectral power, but not theta/beta ratio, is a neuromarker for adult ADHD

Neurofilament light chain serum levels correlate with 10‐year MRI outcomes in multiple sclerosis

A Solution to the Hyper Complex, Cross Domain Reality of Artificial Intelligence: The Hierarchy of AI

Contact Info

Product

Resources

About