Article Text
PDF
Abstract
Introduction Anorexia nervosa is a refractory psychiatric disorder with a mortality rate of 5.9% and standardised mortality ratio of 5.35, which is much higher than other psychiatric disorders. The standardised mortality ratio of bulimia nervosa is 1.49; however, it is characterised by suicidality resulting in a shorter time to death. While there is no current validated drug treatment for eating disorders in Japan, cognitive–behavioural therapy (CBT) is a well-established and commonly used treatment. CBT is also recommended in the Japanese Guidelines for the Treatment of Eating Disorders (2012) and has been covered by insurance since 2018. However, the neural mechanisms responsible for the effect of CBT have not been elucidated, and the use of biomarkers such as neuroimaging data would be beneficial.
Methods and analysis The Eating Disorder Neuroimaging Initiative is a multisite prospective cohort study. We will longitudinally collect data from 72 patients with eating disorders (anorexia nervosa and bulimia nervosa) and 70 controls. Data will be collected at baseline, after 21–41 sessions of CBT and 12 months later. We will assess longitudinal changes in neural circuit function, clinical data, gene expression and psychological measures by therapeutic intervention and analyse the relationship among them using machine learning methods.
Ethics and dissemination The study was approved by The Ethical Committee of the National Center of Neurology and Psychiatry (A2019-072). We will obtain written informed consent from all patients who participate in the study after they had been fully informed about the study protocol. All imaging, demographic and clinical data are shared between the participating sites and will be made publicly available in 2024.
Trial registration number UMIN000039841
- eating disorders
- psychiatry
- neurobiology
- neurophysiology
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
Strengths and limitations of this study
The research project will seek to clarify diagnostic and therapeutic markers and identify predictive markers of treatment response in patients with eating disorders using a longitudinal approach.
This should contribute to the early detection and early intervention of eating disorders.
We will also create a brain imaging database in Japan for those with eating disorders.
MRI equipment and treatment protocols are not completely unified.
Introduction
Eating disorders (EDs) are a psychiatric disorder with a focus on body shape, weight and abnormal eating behaviours. They are mainly classified as anorexia nervosa (AN), bulimia nervosa (BN) and binge-eating disorder (BED).1 In Japan, the number of patients with EDs has increased about 10-fold between the 1980s and 1990s.2 Based on a meta-analysis including 36 studies, AN was found to have a standardised mortality ratio (SMR) of 5.13 and the highest mortality rate among mental disorders.4–7 In addition, renal function decreases as the duration of AN increases,8 and patients often suffer from various physical complications associated with prolonged symptoms and low weight.9–12 The rate of chronicity for patients with AN for up to 10 years is approximately 10%–20%, and the long-term prognosis is poor.13 14 Furthermore, current treatments may not be effective in severe cases of chronic AN.15 Regarding BN, the SMR is 1.49; however, it is characterised by suicidality resulting in a shorter time to death.3 Therefore, early detection and early intervention are important.
Cognitive–behavioural therapy (CBT) is a psychotherapy that aims to improve psychiatric symptoms by focusing on cognitive processes and behavioural patterns. CBT for EDs identifies the psychopathology that contributes to the disease while advancing regular eating and weight-regulating behavioural modifications.16 Cognitive styles such as fear of weight gain and adherence to body shape and behavioural patterns such as dietary restrictions and excessive exercise are often seen in EDs17; CBT can be used to address these psychopathologies.16 CBT has been shown to be effective as a psychotherapy for EDs,18–20 and CBT is also recommended in the National Institute for Health and Care Excellence (NICE) guidelines.21 Meta-analyses of Randomised Controlled Trials (RCTs) that include waitlist controls report that CBT is particularly effective for BN and BED.22–26 In Japan, there are two reports on the effectiveness and feasibility of CBT for BN or BED,27 28 and the national health insurance has covered CBT in Japan since 2018. However, according to a meta-analysis of RCTs, the dropout rate for CBT is approximately 24%,29 and there are large individual differences in treatment responsiveness.
To the best of our knowledge, there are no reports of neural mechanisms that regulate the effects and treatment responses of CBT for EDs. Therefore, accumulating evidence using brain imaging and biomarkers of gene expression would further our understanding of CBT effectiveness for EDs. The identification of biomarkers may provide clues to the development of diagnostic and therapeutic methods based on the elucidation of the pathophysiology and pathogenic mechanism of EDs. It may contribute to social implementation of early detection and early intervention with diagnostic methods using objective indicators.
Since individuals with EDs show abnormal eating behaviours, structural and functional brain imaging studies have been conducted to elucidate the neural basis of these abnormalities through brain imaging research. There are various reports on brain imaging research in patients with ED that have revealed abnormal reward systems,30 reductions in grey matter of various brain regions (frontal lobes,31 frontal gyrus,32 parietal and temporal lobes,33 occipital lobe34) and limited reductions in white matter volume.35 In Japan, the neural basis for the pathogenesis of EDs has been studied.36–39 However, brain imaging research for EDs has only been conducted in the form of small cross-sectional studies in a single facility, and the reproducibility and validity of these results remain questionable. Therefore, research at a larger scale is needed to elucidate the cognitive brain science of ED pathology and progression and to identify clinically useful diagnostic and therapeutic markers and prognostic predictors.
The ENIGMA-Eating Disorders should be noted as a multicentre study outside of Japan. However, it is only registered by researchers in Europe, the USA and Africa and mainly takes place in Germany. For the time being, they have used methods to meta-analyse anatomical images and diffusion tensor-weighted images of patients with AN and BN that have already been imaged at participating facilities. The ENIGMA-Eating Disorders study is expected to build a framework for joint research to be developed in the future. However, ED brain image multicentre research has not been found to have been conducted, even when searching the world’s major clinical research registration sites such as ClinicalTrials.gov in the USA and EU register in the EU. Furthermore, an ED brain image database has not yet been developed globally.
Aims and hypotheses
We aim to generate neuroscientific evidence for the effect of CBT. First, we will longitudinally collect brain MRI images and clinical data in observational studies before and after CBT for EDs. Next, we will identify clinical biomarkers of EDs through analytical studies using longitudinal image data before and after CBT for EDs and create neural evidence for the effect of CBT. We aim to identify brain image biomarkers that can be used as clinical markers (diagnostic markers, therapeutic markers and therapeutic response prediction markers) for EDs. We hypothesise the following: (1) characteristic brain circuit abnormalities exist for each type of ED; (2) there are brain circuit changes that correlate with changes in severity before and after treatment for ED; and (3) there are brain circuit features that define the therapeutic response of CBT for ED. Furthermore, the brain image data collected for this project will be integrated into an international brain database prepared by the Japan Agency for Medical Research and Development (AMED). In the future, after the AMED Brain/MINDS Beyond human brain MRI database is constructed, it will be possible to access the database. At the moment, the portal site (https://brainminds-beyond.jp/ja/resources/2020/05/amedmri.html) is being created (Details of The Brain / MINDS Beyond human brain MRI project: Koike et al, 2020 preprint).
Methods and analysis
Study design
This is a multisite, observational cohort study. For patients with ED who have received structured CBT, the following will be performed before and after CBT: brain MRI (T1-weighted image (T1WI), T2-weighted image (T2WI), resting state functional MRI and diffusion tensor imaging (DTI)); blood collection for gene polymorphism and gene expression analysis; and psychological evaluation.Table 1 shows the study design. In the pretreatment evaluation, subjects will visit each treatment facility for treatment, and a doctor or psychologist (recruiter) will evaluate eligibility and exclusion criteria for CBT introduction. Among the subjects who meet the CBT eligibility criteria, the eligibility criteria and exclusion criteria of this study will be evaluated. Subjects who have obtained informed consent for this study will be registered as study subjects. After registration, the schedule for the first session of CBT will be determined. In principle, brain MRI scans and other tests will be carried out within 4 weeks before the session date. The post-treatment evaluation is based on the day when CBT for EDs is completed, and in principle, tests such as brain MRI are performed within 4 weeks. In addition, a similar assessment 1 year after treatment (50±8 weeks) will be performed for subjects who are willing.
Standard protocol items
Study setting
This trial will be conducted by seven facilities (National Center of Neurology and Psychiatry (NCNP), Chiba University, Tohoku University, Tokyo University, Kyoto University, Kyushu University and University of Occupational and Environmental Health) in Japan.
Recruitment
Those who wish to participate in the study will visit the NCNP, Chiba University, Tohoku University, Tokyo University, Kyoto University, Kyushu University and University of Occupational or Environmental Health. The study participants will be recruited through posters and leaflets placed, through the official web-based advertisements, and by referrals from their primary care doctors or psychiatrists to receive CBT treatment. The doctor or psychologist (recruiter) evaluates the eligibility criteria and exclusion criteria for CBT introduction. Among the subjects who meet the CBT eligibility criteria, the eligibility criteria and exclusion criteria of this study will be evaluated. Subjects who have obtained informed consent for this study will be registered as subjects for this study.
Study participants
Patients will be enrolled in the study if they meet the following selection criteria without meeting the exclusion criteria (see table 2).
Inclusion and exclusion criteria
Patient and public involvement
Patients and public were not involved in the design of this study.
Interventions
In principle, CBT ‘improved version’20 for EDs, 20–40 session version will be implemented (UMIN000031625). However, structured CBT performed at each treatment facility is acceptable (UMIN000039485, UMIN000036825 and so on). Therapists are specialists (psychiatrists and certified public psychologists) who have attended workshops and have been trained in conducting CBT. In addition, each CBT is manualised and structured. Treatment compliance/fidelity is basically maintained by each therapist receiving supervision and following manuals.
Primary outcome
The primary outcome was ED symptoms and severity assessed by the Eating Disorder Examination Edition 17.0D (EDE 17D)40 41 or Eating Disorders Examination Questionnaire (EDE-Q).41 42 The EDE 17D is a semistructured interview, whereas the EDE-Q is a self-contained, 28-item questionnaire derived from the Eating Disorder Examination Edition (EDE). The EDE-Q is scored on a 7-point Likert scale (0–6) on which a score of ≥4 indicates a clinical range. The global score on the EDE-Q is the sum of the four subscale scores (for restraint, eating concern, shape concern and weight concern) divided by 4.
Secondary outcomes
Remission at the end of treatment (state that does not meet DSM-5 criteria).
Treatment completion rate (completed when 75% or more of 21–41 treatment sessions are received).
ED-specific indicators: current BMI; the lowest and highest BMI in the past; history of AN and BN.
Non-specific psychological indicators (see details as a online supplemental file 1): the Mini-International Neuropsychiatric Interview43 44 for comorbidities; Japanese Adult Reading Test45 46 for intellectual ability; Beck Depression Inventory-Second Edition47 48 for severity of depressive symptoms; State-Trait Anxiety Inventory49 50 for anxiety; NEO Five‐Factor Inventory (NEO‐FFI)51 52 for personality scores; 20-Item Toronto Alexithymia Scale53 54 for alexithymia; adult version facial expression recognition test55 for adult facial expression recognition ability; Edinburgh handedness inventory56 57 for determining objectively whether one is left or right handed; Childhood Trauma Questionnaire58 59 for childhood trauma; Post-traumatic Diagnostic Scale60 61 for trauma; Visual Analogue Scale62 for expectation to treatment; General Functioning scale of Family Assessment Device63 64 for family assessment; help-seeking preferences65 for attitude to seek help for others; Multidimensional Assessment of Interoceptive Awareness66 67 for relevant dimensions of bodily awareness; Barratt Impulsiveness Scale68 69 for impulsivity scale; Maudsley Obsessional Compulsive Inventory70 71 for obsessive–compulsive symptoms scale; Autism-Spectrum Quotient72 73 for autism tendencies; Rosenberg Self-Esteem Scale (SES)74 75 for global self-esteem; Tri-axial Coping Scale76 for evaluation of stress coping strategies; Working Alliance Inventory77 for aspects of the therapeutic alliance; EuroQol-5 Dimension78 79 for quality of life; the Generalized Anxiety Disorder Assessment80 81 for severity of anxiety; Patient Health Questionnaire81 82 for depression module; the WHO - Five Well-Being Index83 for mental health scale; socioeconomic status for education history; and Stanford Sleepiness Scale84 for subjective sleepiness levels.
Brain image data, gene polymorphism/gene expression analysis data.
Supplemental material
Demographic data
At the time of pre-CBT, we will collect the following demographic data: age, educational background, marital status, cohabitation/family/presence of partner, hospitalisation history, age of onset, medical history, comorbidities, family history and medication content.
Imaging acquisition
MRI scans will be obtained in all participants on 3 Tesla scanners: Siemens MAGNETOM Prisma (University of Tokyo), Skyrafit (NCNP) and Verio (Kyoto University); GE Discovery MR750 3.0T (Chiba University), Premier (University of Occupational and Environmental Health); and Phillips Ingenia 3.0T CX (Tohoku University and Kyushu University). The brain MRI examination takes T1WI, T2WI, resting-state fMRI and diffusion tensor-weighted images. The imaging protocol is according to Harmonized Protocol (HARP)85 for clinical MRI studies in Brain/MINDS Beyond when multiband echo-planar imaging option is enabled in each site. If not, the imaging protocol will be adopted based on the Strategic Research Program for Brain Science (SRPB) protocol. Resting-state fMRI will be acquired in an open-eye condition. MRI model, coil and imaging parameters are shown in tables 3 and 4.
3T MRI specification and imaging protocols
Scanning parameters for HARP and SRPB
Biological measures from blood samples
For blood tests, approximately 20 mL of blood will be collected at a time, with a total of approximately 60 mL during the study period. For shock symptoms due to vasovagal reactions, physical condition will be checked on the day of blood collection, and patients will be closely observed for 5 min after the start of blood collection when this reaction is likely to occur. Blood samples will include blood plasma, serum, blood sampling for DNA methylome and blood sampling for RNA transcriptomics.
Data logistics
The brain imaging data collected at all facilities will be anonymised at each facility and then aggregated in the NCNP through the Integrative Brain Imaging Support System. The primary analysis will be performed in the NCNP. The results of the primary analysis will be shared by all facilities. Blood samples collected at all facilities will be anonymised, and all samples for gene polymorphism/gene expression analysis will be sent to NCNP where the samples will be extracted and stored in a −80°C freezer. As soon as approximately 20 cases are collected, the gene expression analysis will be outsourced to a contractor. In addition, psychological and clinical data will be anonymised at each institution and then shared by all institutes (see the figure 1). The correspondence table for all data will be managed at the facility where the data were collected.
Data logistics. CHB, Chiba University; HCP, Human Connectome Project; IBISS, Integrative Brain Imaging Support System; KYS, Kyushu University; KYU, University of Kyoto; NCNP, National Center of Neurology and Psychiatry; THK, Tohoku University; UOEH, University of Occupational and Environmental Health; UTI, The University of Tokyo IRCN.
Statistical analysis
Brain imaging data collected at all facilities are subjected to a primary analysis by the analysis pipeline from NCNP. From the brain images of each individual, the data representing brain grey matter mass, white matter integrity and resting functional connection of the region of interest will be extracted. In addition, we will evaluate functional connectivity in various intracerebral networks and between networks such as the default mode network, salience network, dorsal attention network, cognitive control network and affective network in resting brain activity. For example, we plan to perform seed-based analysis with CONN toolbox (www.nitrc.org/projects/conn).86 We plan to verify the presence or absence of network abnormalities using the orbitofrontal cortex and anterior cingulate cortex, which have been pointed out as seeds in ED, and whether these network abnormalities have changed after CBT.
At the secondary analysis stage, strategies (harmonisation) for adjusting interfacility factors will be considered. At present, we are planning to use corrections based on average values between facilities or software such as Combat. The ComBat harmonisation tool87–89 uses Bayesian regression to find systematic differences among multiple data collected using different scanners. The tool performs additive and multiplicative corrections to produce distortions that eliminate these systematic differences from the data; however, as new harmonisation strategies are developed, they will be tested as appropriate. At the time when about 20 longitudinal data are collected, the data will be fixed after being integrated with psychological/clinical data other than brain images and gene polymorphism/gene expression analysis results, and the following analysis will begin: (1) identification of diagnostic markers (baseline data 20 people): compare clinical symptoms and brain images, psychology, gene polymorphism and gene expression data at baseline in AN and BN disease types, and search for cognitive, psychological and behavioural indicators associated with disease type diagnosis. At this time, a comparison with the healthy group will also be performed. By comparing these groups, diagnostic markers based on conventional diagnoses will be identified (categorical approach). Furthermore, we will identify diagnostic markers on searching for psychopathological features in brain images, gene polymorphisms and gene expression data related to core disease state indicators and other psychological indicators across all disease types (dimensional approach); (2) identification of therapeutic effect (longitudinal data set 40 people): we will analyse the relationship between changes in ED symptoms before and after CBT and changes (rates) in brain images, psychology, gene polymorphisms and gene expression data, and identify the therapeutic effects of CBT; (3) identification of therapeutic response markers (longitudinal data set 60): analysing associations with baseline brain images, psychology, gene polymorphisms and gene expression data to predict the completion and remission of EDs before and after CBT, and identify treatment predictive markers; (4) in addition to the above, analysis that is judged necessary or meaningful at the time of analysis will be performed. In particular, we plan to conduct analysis using machine learning techniques. If there are missing values in the dataset to be analysed, they will, in principle, be excluded from the analysis.
Machine-learning approach
We are planning to use the python Scikitlearn library (V.0.22.1) to perform support vector machine (SVM) classification.90 To prevent overfitting (ie, the classifier works perfectly on the training data but is poorly generalisable to new data), we will perform a feature relevance evaluation and dimensionality reduction using a tree-based feature selection approach with a nested cross-validation design.91–93
The nested cross-validation consists of an inner loop for model building and parameter estimation and an outer loop for model testing. Consequently, the dataset will be divided into two parts: a training plus validation subset and a test subset. In the inner loop, SVM models will have been trained with varying SVM hyperparameters (ie, cost parameters C and kernel function) based on a grid search, and a feature selection will have been performed using classification trees or random forests. The validation set will be used to determine the SVM hyperparameters over the grid of possible values. The performance of the resulting model, with optimised SVM hyperparameters and features, will be subsequently evaluated on the test set in the outer loop. For this outer loop, we will use a leave-one-out cross-validation scheme so that each sample will be used once as a test set (singleton) while the remaining samples form the training/validation set. The training/validation set will be divided into five equally sized parts. Four of these will be used as the training set and one as the validation set. This process is repeated until all subjects are left out at least once and prediction results are aggregated as well as reported.
Sample size
The number of cases needed to detect an improvement in ED symptoms, which is the main outcome, was calculated. If calculated based on interpersonal therapy (n=65, mean=2.37, SD=1.25) versus CBT (n=65, mean=1.57, SD=1.25),18 a single group of 27 cases is required based on a comparison of average outcomes of two independent samples at the time of postintervention (t-test); two tails, α=0.05, 1−β=0.80. Hence, a total of 54 patients is required (27 cases for AN; 27 cases for BN). We plan to complete 72 cases among the joint research facilities. For 5 years, considering the dropout rate of CBT for ED treatment (about 40%) and the consent rate for MRI (about 60%), and differences in CBT for each facility, we plan to recruit 200 patients with EDs who are introduced to CBT at all facilities. We expect 120 cases to obtain consent for MRI examination and 72 cases to complete CBT. Regarding the disease type, the number of patients with AN and BN is expected to be about 1:1 in the end. Tohoku University and Kyoto University have mainly implemented CBT for patients with BN; conversely, Tokyo University, University of Occupational and Environmental Health and Kyushu University have mainly implemented CBT for patients with AN, whereas NCNP and Chiba University have implemented CBT for patients with AN and BN almost equally. Furthermore, we will collect 70 cases in the healthy control group. In particular, in order to verify the effects of low body weight, we will collect data for 30 healthy women with a BMI between 16 and 17.5 (BMI values satisfy one of the diagnoses of AN) and no ED pathology.
Trial status
The first participant was included in this study on 16 June 2020. The end of the recruitment phase is currently scheduled for 31 July 2024.
Ethics and dissemination
Data distribution
At the end of the project period, the data will be publicly distributed to researchers through public databases. All imaging, demographic and clinical data are shared between the participating sites and will be made publicly available in 2024. To the best of our knowledge, this is one of the first multisite human brain MRI projects investigating multiple mental and neurological disorders throughout life. The Brain/MINDS Beyond human brain MRI project will help identify common and disease-specific pathophysiological features of brain disease and develop imaging biomarkers for clinical practice.
Ethical regulation
We follow the ethical regulation of the previous study.85 Sharing neuropsychiatric patient data that can contain information linked to subjects’ privacy requires special attention.94 Hence, the Brain/MINDS Beyond project has made NCNP the core site for supporting ethical considerations. Before participating in the project, all institutions are required to have their research plans approved by their ethical review committee. This includes the following points and ethical documentation: (1) MR images and clinical data of the participants can be shared within the Brain/MINDS Beyond Project or Japanese/international scientific institutions. Anonymised MR images with limited clinical data may become publicly available on an open database for research purposes, (2) MR images of the participants may be compared with nonhuman primate MRI data and (3) the intellectual property rights derived from the research of the Brain/MINDS Beyond project shall be attributed to the researcher’s institute, not the participants. All participants, after receiving the full description of the experiment, are required to provide written informed consent to participate in this project. The Japanese regulations for the sharing of personal information used for research purposes require attention when handling two types of data: ‘individual identification codes’ and ‘special care-required personal information’ (http://www.japaneselawtranslation.go.jp/law/detail/?id=2781&vm=04&re=01). The individual identification code is a direct identifier and is sufficient to identify a particular individual. Special care-required personal information represents indirect identifiers that need special care in handling in order to avoid potential disadvantages to the participants. In consideration of these regulations, data accompanied with the MR images are limited in the publicly accessible open database and include only 5-year age bins, sex, diagnostic information, handedness, simple socioeconomic status, clinical scale scores and sleepiness scale scores. In the Brain/MINDS Beyond project, we exclude the datasets of MR images containing facial information from the data in the publicly accessible open database.
Discussion
The strengths of this research are listed below. First, we aimed to construct a brain image database of EDs using the longitudinal brain imaging scans of ED patients from multiple facilities. To date, brain imaging research for ED has been limited to small-scale cross-sectional studies.36–39 For this reason, it is important to establish a brain imaging database of EDs using brain images from ED patients across multiple facilities and longitudinally to identify clinically useful diagnostic and therapeutic markers and prognostic predictors. Second, data collected before and after CBT treatment will be collected longitudinally, and treatment responsiveness can be input as a variable. In addition, it is possible to explore the effects of treatment responsiveness by treating childhood trauma as a treatment resistance factor and covariate. Third, the machine learning method can be used to develop into analytical research that leads to an integrated understanding of the pathology of EDs and even neuropsychiatric disorders. Thus far, studies that apply machine learning techniques to brain imaging data from patients with ED have begun to gradually appear but remain limited to cross-sectional studies using samples of 15–24 patients in a single group.95 96 To date, there remains no longitudinal data from before and after treatment and the association with other clinical markers. Lastly, we will apply an omics analysis to identify clinical markers of ED. Thus far, previous studies have been reported in Japan addressing the susceptibility gene for feeding regulators: the relationship between Ghrelin polymorphism and BN97; the association with young female dissatisfaction of the body, physique and high-density lipoprotein cholesterol98; and the association with AN and the fatty acid amide hydrolase (FAAH) polymorphism.99 A genome-wide correlation analysis (GWAS) using the world’s first microsatellite marker for AN was performed, and single nucleotide polymorphisms (SNPs) showing AN sensitivity were identified in at least three gene regions (exon 9 of the CNTN5 gene, the 3′-downstream region of the SPATA17 gene and TOX3 gene).100 Furthermore, in a GWAS using SNP markers, a previous study has suggested that having the minor 385A allele of the FAAH gene may be protective against restricting AN. Using the knowledge of ED genetic research, it is possible to narrow down the genes targeted in an omics analysis for ED carried out in this research project. Omics research for EDs has only been reported in a cross-sectional study of a few groups for AN,101 and no longitudinal study before and after treatment including CBT has been conducted. ED omics research is now beginning to gain traction. Hence, we will design to handle multivariates that includes omics analysis results in brain imaging data. We reduced the dimensions using deep learning techniques and will perform a multiple regression analysis using machine learning algorithms. This is expected to generate evidence for clinical markers of EDs (diagnostic markers, therapeutic effects and therapeutic response prediction markers).
This study has some limitations. First, the number of participants by sites may not be uniform depending on the recruitment of patients and other situations. Also, the number of total participants is too small, and therefore, it is not possible to use machine learning approaches to explore appropriate models without overfitting the available data. Second, although harmonisation is applied, the measurement variability may increase because MRI scanners and imaging protocols can differ between manufacturers. For example, an image scanned using the SRPB imaging protocol has a lower signal-to-noise ratio and resolution compared with an image scanned using HARP. Also, because of the presence or absence of the multiband option, the sampling rate can be up to three times different among facilities. Despite these limitations, such challenging and exploratory approaches are necessary to construct a brain imaging database of patients with EDs from multiple institutions as a first step. Hopefully, by sharing data between other research teams who are collecting brain imaging data on patients with EDs, we will be able to collect a sufficient amount of data.
Acknowledgments
We are very grateful to the patients for participating in the trial dissemination. We would also like to thank the authors of this study and our colleagues supporting the study.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors SH, YH and NK participated in the design, drafted and modified the manuscript. YH, MI, NK, KazufY, YM, TA, KazuhY and YS are the coinvestigators of the study who obtained funding and significantly contributed to the conception or design. AyS, YE, JT, NN, TT, HH, TN, KT, KW, HA, MG, ST, SF and ES participated in the design and substantially contributed to the different stages of the study’s development towards its practical conduction. AtS is the principal investigator of the study who obtained funding, modified the manuscript and provided substantial contributions to the study conception and design. All authors read and approved the final manuscript.
Funding This work was supported by AMED under grant number JP20dm0307104.
Competing interests None declared.
Patient consent for publication Not required.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.