Neural Biomarkers for Identifying Atopic Dermatitis and Assessing Acupuncture Treatment Response Using Resting-State fMRI

Abstract

In-Seon Lee,1 Da-Eun Yoon,1 Seoyoung Lee,2 Jae-Hwan Kang,3,4 Younbyoung Chae,1 Hi-Joon Park,1 Junsuk Kim5 1College of Korean Medicine, Kyung Hee University, Seoul, Republic of Korea; 2Department of Physics and Computational Radiology, Division of Radiology and Nuclear Medicine, Oslo University Hospital, Oslo, Norway; 3Digital Health Research Division, Korea Institute of Oriental Medicine, Daejeon, Republic of Korea; 4Aging Convergence Research Center, Korea Research Institute of Bioscience and Biotechnology, Daejeon, Republic of Korea; 5School of Information Convergence, Kwangwoon University, Seoul, Republic of KoreaCorrespondence: Junsuk Kim, School of Information Convergence, Kwangwoon University, 20 Kwangwoon-ro, Seoul, 01897, Republic of Korea, Tel [phone], Email [email protected] Hi-Joon Park, College of Korean Medicine, Kyung Hee University, 26 Kyunghee-daero, Seoul, 02447, Republic of Korea, Tel [phone], Email [email protected]: Only a few studies have focused on the brain mechanisms underlying the itch processing in AD patients, and a neural biomarker has never been studied in AD patients. We aimed to develop a deep learning model-based neural signature which can extract the relevant temporal dynamics, discriminate between AD and healthy control (HC), and between AD patients who responded well to acupuncture treatment and those who did not.Patients and Methods: We recruited 41 AD patients (22 male, age mean ± SD: 24.34 ± 5.29) and 40 HCs (20 male, age mean ± SD: 26.4 ± 5.32), and measured resting-state functional MRI signals. After preprocessing, 38 functional regions of interest were applied to the functional MRI signals. A long short-term memory (LSTM) was used to extract the relevant temporal dynamics for classification and train the prediction model. Bootstrapping and 4-fold cross-validation were used to examine the significance of the models.Results: For the identification of AD patients and HC, we found that the supplementary motor area (SMA), posterior cingulate cortex (PCC), temporal pole, precuneus, and dorsolateral prefrontal cortex showed significantly greater prediction accuracy than the chance level. For the identification of high and low responder to acupuncture treatment, we found that the lingual-parahippocampal-fusiform gyrus, SMA, frontal gyrus, PCC and precuneus, paracentral lobule, and primary motor and somatosensory cortex showed significantly greater prediction accuracy than the chance level.Conclusion: We developed and evaluated a deep learning model-based neural biomarker that can distinguish between AD and HC as well as between AD patients who respond well and those who respond less to acupuncture. Using the intrinsic neurological abnormalities, it is possible to diagnose AD patients and provide personalized treatment regimens.Keywords: Atopic Dermatitis, deep learning, functional MRI, biomarkers, personalized medicine

Keywords

• atopic dermatitis
• deep learning
• functional mri
• biomarkers
• personalized medicine