Features selection to distinguish Frontotemporal Lobar Degeneration syndromes using machine learning
Poster No:
1836
Submission Type:
Abstract Submission
Authors:
Mariana Teresa da Silva1, Tiago Azevedo1, Boyd Ghosh2, James Rowe1, Marta Correia1, Timothy Rittman1
Institutions:
1University of Cambridge, Cambridge, United Kingdom, 2University Hospital Southampton NHS trust, Cambridge, United Kingdom
First Author:
Co-Author(s):
Introduction:
Frontotemporal Lobar Degenerative Syndromes (FTLD) comprise a group of neurodegenerative disorders causing changes in frontal cognitive and motor functions. Four syndromes were assessed in this study: Progressive Supranuclear Palsy (PSP), Corticobasal Syndrome (CBS), behavioural variant Frontotemporal dementia (bvFTD) and Primary Progressive Aphasia (PPA). Accurate diagnosis is crucial for guiding clinical management and stratifying people appropriately for trials of disease-modifying drugs.
Machine learning algorithms applied to neuroimaging data have successfully achieved binary classification. In this study we got further to distinguish between the four FTLD clinical syndromes.
Using Diffusion Tensor Imaging (dMRI) has advantages over structural MRI for atypical parkinsonian disorders, including distinguishing PSP and CBS from Parkinson's disease (Correia et al., 2020).However, it remains unclear what combination of DTI neuroimaging features may provide optimal classification.
Machine learning algorithms applied to neuroimaging data have successfully achieved binary classification. In this study we got further to distinguish between the four FTLD clinical syndromes.
Using Diffusion Tensor Imaging (dMRI) has advantages over structural MRI for atypical parkinsonian disorders, including distinguishing PSP and CBS from Parkinson's disease (Correia et al., 2020).However, it remains unclear what combination of DTI neuroimaging features may provide optimal classification.
Methods:
A total of 299 patients divided between clinical diagnostic groups of PSP (106), CBS (52), PPA (46), bvFTD (27) were recruited from a regional tertiary referral specialist clinic at the Cambridge University Hospitals NHS Trust, UK. Age-matched health control subjects (68) were recruited. All subjects underwent a 3T MRI scan including diffusion MRI (dMRI) with diffusion sensitising gradients applied along 63 or 68 directions with b-values of 0 and 1000 s/mm2. Standard preprocessing with applied including motion correction and eddy correction. Fractional Anisotropy (FA) and Mean Diffusivity (MD) were extract from regions of interest (ROI) of the JHU atlas (Mori et al., 2005). The Addenbrooke's Cognitive Examination revised (ACE-r) was collected for all volunteers.
Results:
Table 1: showing performance results for multiple supervised machine learning algorithms on binary classification of disease vs controls.
Legend: PCA = Principal Component Analysis, Kbest = SelectKBest function, RFE = Recursive feature elimination, Acc = accuracy. FA = fractional anisotropy, MD = Mean diffusivity, ACE-r = Addenbrooke's Cognitive Examination revised. ** the highest value of F1_score per row
Considering the imbalance of the dataset, F1 score was used to compare the performance of each model followed by accuracy when two F1_scores were the same (e.g. PCA with RFE with features FA, MD and ACE-r). Results in table one demonstrates, firstly, that adding the clinical questionnaire ACE-r to the ROI data adds value for the differentiation of disease vs control. Secondly, performing feature selection without PCA resulted in the highest accuracy (92%) and F1_score (0.87) with Kbest as the optimal feature selection algorithm. Finally, Support Vector Machines slightly out-performed other algorithms.
Legend: PCA = Principal Component Analysis, Kbest = SelectKBest function, RFE = Recursive feature elimination, Acc = accuracy. FA = fractional anisotropy, MD = Mean diffusivity, ACE-r = Addenbrooke's Cognitive Examination revised. ** the highest value of F1_score per row
Considering the imbalance of the dataset, F1 score was used to compare the performance of each model followed by accuracy when two F1_scores were the same (e.g. PCA with RFE with features FA, MD and ACE-r). Results in table one demonstrates, firstly, that adding the clinical questionnaire ACE-r to the ROI data adds value for the differentiation of disease vs control. Secondly, performing feature selection without PCA resulted in the highest accuracy (92%) and F1_score (0.87) with Kbest as the optimal feature selection algorithm. Finally, Support Vector Machines slightly out-performed other algorithms.
Conclusions:
We demonstrate the utility of assessing multiple approaches to feature selection for the differential diagnosis of Frontotemporal Lobar Degenerative syndromes.
Disorders of the Nervous System:
Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) 2
Modeling and Analysis Methods:
Diffusion MRI Modeling and Analysis
Neuroinformatics and Data Sharing:
Informatics Other 1
Keywords:
Computational Neuroscience
Degenerative Disease
Machine Learning
WHITE MATTER IMAGING - DTI, HARDI, DSI, ETC
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using APA citation style.
Mori, S., Wakane, S., van Zijl, P.C.M., Nagae-Poetscher, L.M. (2005), MRI Atlas of Human White Matter. Elsevier, Amsterdam, The Netherlands.