Abstract Details

Title Identifying The Role Of A Hippocampal-thalamic Memory Network In Ms Using A Machine Learning Approach

Topic Multiple Sclerosis

Presentation(s) P2 - Poster Session 2 (8:00 AM-9:00 AM)

Poster/Presentation
Number 9-002

Objective To extract a brain pattern that distinguishes episodic memory (EM) impaired and preserved MS patients using a machine learning method applied to multimodal MRI data.

Background EM impairment is common in MS. The thalamus, hippocampus, and fornix comprise a primary memory network of the brain; the extent to which MS-related changes in this network relate to memory impairment is not known.

Design/Methods We segregated 182 MS patients (RADIEMS cohort: 65.9% female, mean age 34.4±7.6 years, disease duration 2.1±1.4 years) into EM impaired (N=44) and preserved groups using a computerized nonverbal test of episodic memory (CANTAB PAL). In addition to volumes of 12 hippocampal subfields and 26 thalamic nuclei, resting state functional connectivity (FC) between hippocampus and thalamus, and DTI [fractional anisotropy (FA) and mean diffusivity (MD)] of bilateral fornix were examined. An ensemble model was created by taking an average of the predictions of a machine learning method: Random Forest (RF). Predictions were obtained using two different datasets: 1-demographic dataset (age, sex, education level, and IQ), and 2-imaging dataset. Average of the area under ROC curve (AUC) and average balanced accuracy (BA) were used to assess the performance of RF with various datasets over 500 models.

Results Ensemble model applied to RF method gave high classification results (AUC=0.679±0.114 and BA=0.686±0.114) using demographics and imaging data. In the demographic dataset, age and IQ were the most important variables. In the imaging dataset, volume of the right and left fimbria of the hippocampus, left lateral posterior nucleus of the thalamus, and MD of the left fornix were the most important four predictors.

Conclusions Along with age, IQ and select thalamic nuclei and hippocampal subfiled volumes, microstructural damage to the fornix importantly contribute to EM impairment in early MS. To understand memory network changes and EM impairment better, future studies evaluating the thalamic-hippocampal memory network are warranted.

Authors/Disclosures
Korhan Buyukturkoglu, PhD (Columbia University) PRESENTER	Dr. Buyukturkoglu has nothing to disclose.
	No disclosure on file
	No disclosure on file
Amy Kuceyeski, PhD (Weill Cornell Medical College, Radiology)	An immediate family member of Dr. Kuceyeski has received personal compensation for serving as an employee of Heyer Physical Therapy. The institution of Dr. Kuceyeski has received research support from National Institutes of Health. Dr. Kuceyeski has received personal compensation in the range of $500-$4,999 for serving as a Grant Panel Member with Department of Defense.
James F. Sumowski (Icahn School of Medicine At Mount Sinai)	Mr. Sumowski has nothing to disclose.
Victoria Leavitt, PhD, F�鶹��ýӳ�� (Columbia University Irving Medical Center)	Dr. Leavitt has received personal compensation in the range of $500-$4,999 for serving on a Scientific Advisory or Data Safety Monitoring board for Biogen. Dr. Leavitt has received personal compensation in the range of $500-$4,999 for serving on a Speakers Bureau for Novartis. The institution of Dr. Leavitt has received research support from National Institutes of Health. The institution of Dr. Leavitt has received research support from National Multiple Sclerosis Society. The institution of Dr. Leavitt has received research support from Department of Defense. Dr. Leavitt has received intellectual property interests from a discovery or technology relating to health care.

�鶹��ýӳ��

�鶹��ýӳ��