Objectives To investigate the prevalence of axial spondyloarthritis (axSpA) in patients with chronic back pain (CBP) of less than 2 years (2y) duration referred to the rheumatologist, the... Show moreObjectives To investigate the prevalence of axial spondyloarthritis (axSpA) in patients with chronic back pain (CBP) of less than 2 years (2y) duration referred to the rheumatologist, the development of diagnosis over time, and patient characteristics of those developing definite (d-)axSpA over 2y. Methods We analysed the 2y data from SPondyloArthritis Caught Early, a European cohort of patients (<45 years) with CBP (>= 3 months, <= 2y) of unknown origin. The diagnostic workup comprised evaluation of clinical SpA features, acute phase reactants, HLA-B27, radiographs and MRI (sacroiliac joints and spine), with repeated assessments. At each visit (baseline, 3 months, 1y and 2y), rheumatologists reported a diagnosis of axSpA or non-axSpA with level of confidence (LoC; 0-not confident at all to 10-very confident). Main outcome: axSpA diagnosis with LoC >= 7 (d-axSpA) at 2y. Results In 552 patients with CBP, d-axSpA was diagnosed in 175 (32%) at baseline and 165 (30%) at 2y. Baseline diagnosis remained rather stable: at 2y, baseline d-axSpA was revised in 5% of patients, while 8% 'gained' d-axSpA. Diagnostic uncertainty persisted in 30%. HLA-B27+ and baseline sacroiliitis imaging discriminated best 2y-d-axSpA versus 2y-d-non-axSpA patients. Good response to non-steroidal anti-inflammatory drugs and MRI-sacroiliitis most frequently developed over follow-up in patients with a new d-axSpA diagnosis. Of the patients who developed MRI-sacroiliitis, 7/8 were HLA-B27+ and 5/8 male. Conclusion A diagnosis of d-axSpA can be reliably made in nearly one-third of patients with CBP referred to the rheumatologist, but diagnostic uncertainty may persist in 5%-30% after 2y. Repeated assessments yield is modest, but repeating MRI may be worthwhile in male HLA-B27+ patients. Show less
Background: Recent advances in data-driven computational approaches have been helpful in devising tools to objectively diagnose psychiatric disorders. However, current machine learning studies... Show moreBackground: Recent advances in data-driven computational approaches have been helpful in devising tools to objectively diagnose psychiatric disorders. However, current machine learning studies limited to small homogeneous samples, different methodologies, and different imaging collection protocols, limit the ability to directly compare and generalize their results. Here we aimed to classify individuals with PTSD versus controls and assess the generalizability using a large heterogeneous brain datasets from the ENIGMA-PGC PTSD Working group. Methods: We analyzed brain MRI data from 3,477 structural-MRI; 2,495 resting state-fMRI; and 1,952 diffusion-MRI. First, we identified the brain features that best distinguish individuals with PTSD from controls using traditional machine learning methods. Second, we assessed the utility of the denoising variational autoencoder (DVAE) and evaluated its classification performance. Third, we assessed the generalizability and reproducibility of both models using leave-one-site-out cross-validation procedure for each modality. Results: We found lower performance in classifying PTSD vs. controls with data from over 20 sites (60 % test AUC for s-MRI, 59 % for rs-fMRI and 56 % for D-MRI), as compared to other studies run on single-site data. The performance increased when classifying PTSD from HC without trauma history in each modality (75 % AUC). The classification performance remained intact when applying the DVAE framework, which reduced the number of features. Finally, we found that the DVAE framework achieved better generalization to unseen datasets compared with the traditional machine learning frameworks, albeit performance was slightly above chance. Conclusion: These results have the potential to provide a baseline classification performance for PTSD when using large scale neuroimaging datasets. Our findings show that the control group used can heavily affect classification performance. The DVAE framework provided better generalizability for the multi-site data. This may be more significant in clinical practice since the neuroimaging-based diagnostic DVAE classification models are much less site-specific, rendering them more generalizable. Show less
Helderman, N.C.; Suerink, M.; Kilinc, G.; Berg, J.G. van den; Nielsen, M.; Tesselaar, M.E.T. 2023
Practice of neuroendocrine neoplasms (NENs) of the digestive tract, which comprise a highly diverse group of tumors with a rising incidence, faces multiple biological, diagnostic, and therapeutic... Show morePractice of neuroendocrine neoplasms (NENs) of the digestive tract, which comprise a highly diverse group of tumors with a rising incidence, faces multiple biological, diagnostic, and therapeutic issues. Part of these issues are due to misuse and misinterpretation of the classification and terminology of NENs of the digestive tract, which make it increasingly challenging to evaluate and compare literature. For instance, grade 3 neuroendocrine tumors (NETs) are frequently referred to as neuroendocrine carcinomas (NECs) and vice versa, while NECs are by definition high grade and therefore constitute a separate entity from NETs. Moreover, the term NETs is regularly misused to describe NENs in general, and NETs are frequently referred to as benign, while they should always be considered malignancies as they do have metastatic potential. To prevent misconceptions in future NEN-related research, we reviewed the most recent terminology used to classify NENs of the digestive tract and created an overview that combines the classification of these NENs according to the World Health Organization (WHO) with location- and functionality-based classifications. This overview may help clinicians and researchers in understanding current literature and could serve as a guide in the clinic as well as for writing future studies on NENs of the digestive tract. In this way, we aim for the universal use of terminology, thereby providing an efficient foundation for future NEN-related research. Show less
Arend, B.W.H. van der; Verhagen, I.E.; Leeuwen, M. van; Arend, M.Q.T.P. van der; Casteren, D.S. van; Terwindt, G.M. 2023
BackgroundThere is a need for standardization of the definition of a migraine day for clinical and research purposes. MethodsWe prospectively compared different definitions of a migraine day with E... Show moreBackgroundThere is a need for standardization of the definition of a migraine day for clinical and research purposes. MethodsWe prospectively compared different definitions of a migraine day with E-diary data of n = 1494 patients with migraine. We used a baseline definition based on migraine characteristics with a duration of >= 4 hours OR triptan intake (independently from its effect) OR (visual) aura lasting 5-60 minutes. ResultsOf all migraine days defined by triptan intake only, 66.2% had a duration <4 hours. Adjusting the headache duration criterion to >= 30 minutes led to a decrease in days defined by triptan intake only and resulted in a 5.4% increase in total migraine days (equals 0.45 migraine day increase in monthly migraine days). These additional migraine days had a median duration of 2.5 hours. ConclusionWe propose to define a migraine day as follows: 1) (a) headache duration >= 30 minutes; (b) matching >= 2 of four characteristics: unilateral, pulsating, moderate to severe pain, aggravation by or causing avoidance of routine physical activity; and (c) during headache >= 1 of the following: nausea and/or vomiting, photophobia and phonophobia or 2) (visual) aura duration 5-60 minutes or 3) a day with headache for which acute migraine-specific medication is used irrespective of its effect. Show less
Based on recent achievements in phylogenetic studies of the Brassicaceae, a novel infrafamilial classification is proposed that includes major improvements at the subfamilial and supertribal levels... Show moreBased on recent achievements in phylogenetic studies of the Brassicaceae, a novel infrafamilial classification is proposed that includes major improvements at the subfamilial and supertribal levels. Herein, the family is subdivided into two subfamilies, Aethionemoideae (subfam. nov.) and Brassicoideae. The Brassicoideae, with 57 of the 58 tribes of Brassicaceae, are further partitioned into five supertribes, including the previously recognized Brassicodae and the newly established Arabodae, Camelinodae, Heliophilodae, and Hesperodae. Additional tribus-level contributions include descriptions of the newly recognized Arabidopsideae, Asperuginoideae, Hemilophieae, Schrenkielleae, and resurrection of the Chamireae and Subularieae. Further detailed comments on 17 tribes in need of clarifications are provided. Show less
There are long-standing unsolved issues regarding the diagnosis and classification of central disorders of hypersomnolence. These include delineating and identifying phenotypes and unique... Show moreThere are long-standing unsolved issues regarding the diagnosis and classification of central disorders of hypersomnolence. These include delineating and identifying phenotypes and unique conditions (“sui generis”), sleep deprivation’s impact on phenotypes and how to separate sleep deprivation as a trigger from other causes, as well as the association of excessive sleepiness with other disorders. We discuss these issues and present a novel, straightforward classification system with consistent terminology to get out of the impasse and do justice to people with hypersomnolence. Show less
Monitoring the illegal trade of wool fibres of wild vicun~a (Vicugna vicugna) and guanaco (Lama guanicoe) is highly desirable. The high market value of fleece from these camelid species poses a... Show moreMonitoring the illegal trade of wool fibres of wild vicun~a (Vicugna vicugna) and guanaco (Lama guanicoe) is highly desirable. The high market value of fleece from these camelid species poses a threat to their wild populations. A previous study showed that direct analysis in real time time-of-flight mass spectrometry (DART-TOFMS) effectively identifies wool fibres to species. Producing high-resolution data in a short period of time makes DART-TOFMS a reliable identification tool, even though data analysis can still be improved. The present study proposes a novel data analysing pipeline based on Convolutional Neural Networks (CNN), applicable to any kind of DART-TOF MS data. We tested our proposed method on keratin fibres of four camelid species (Vicugna vicugna: n 1⁄4 19; Vicugna pacos: n 1⁄4 20; Lama guanicoe: n 1⁄4 20, and Lama glama: n 1⁄4 20). Analyses showed that selecting 512 ions with the highest relative intensity provides the best resolution and yields 100% accuracy for species identification. Show less
This paper introduces the Bradley-Terry regression trunk model, a novel probabilistic approach for the analysis of preference data expressed through paired comparison rankings. In some cases, it... Show moreThis paper introduces the Bradley-Terry regression trunk model, a novel probabilistic approach for the analysis of preference data expressed through paired comparison rankings. In some cases, it may be reasonable to assume that the preferences expressed by individuals depend on their characteristics. Within the framework of tree-based partitioning, we specify a tree-based model estimating the joint effects of subject-specific covariates over and above their main effects. We, therefore, combine a tree-based model and the log-linear Bradley-Terry model using the outcome of the comparisons as response variable. The proposed model provides a solution to discover interaction effects when no a-priori hypotheses are available. It produces a small tree, called trunk, that represents a fair compromise between a simple interpretation of the interaction effects and an easy to read partition of judges based on their characteristics and the preferences they have expressed. We present an application on a real dataset following two different approaches, and a simulation study to test the model's performance. Simulations showed that the quality of the model performance increases when the number of rankings and objects increases. In addition, the performance is considerably amplified when the judges' characteristics have a high impact on their choices. Show less
This International evidence-based nomenclature and classification consensus on the congenital bicuspid aortic valve and its aortopathy recognizes 3 types of bicuspid aortic valve: 1. Fused type,... Show moreThis International evidence-based nomenclature and classification consensus on the congenital bicuspid aortic valve and its aortopathy recognizes 3 types of bicuspid aortic valve: 1. Fused type, with 3 phenotypes: right-left cusp fusion, right-non cusp fusion and left-non cusp fusion; 2. 2-sinus type with 2 phenotypes: Latero-lateral and antero-posterior; and 3. Partial-fusion or forme fruste. This consensus recognizes 3 bicuspid-aortopathy types: 1. Ascending phenotype; root phenotype; and 3. extended phenotypes. Show less
This International evidence-based nomenclature and classification consensus on the congenital bicuspid aortic valve and its aortopathy recognizes 3 types of bicuspid aortic valve: 1. Fused type,... Show moreThis International evidence-based nomenclature and classification consensus on the congenital bicuspid aortic valve and its aortopathy recognizes 3 types of bicuspid aortic valve: 1. Fused type, with 3 phenotypes: right-left cusp fusion, right-non cusp fusion and left-non cusp fusion; 2. 2-sinus type with 2 phenotypes: Latero-lateral and antero-posterior; and 3. Partial-fusion or forme fruste. This consensus recognizes 3 bicuspid-aortopathy types: 1. Ascending phenotype; root phenotype; and 3. extended phenotypes. Show less
This International Consensus Classification and Nomenclature for the congenital bicuspid aortic valve condition recognizes 3 types of bicuspid valves: 1. The fused type (right-left cusp fusion,... Show moreThis International Consensus Classification and Nomenclature for the congenital bicuspid aortic valve condition recognizes 3 types of bicuspid valves: 1. The fused type (right-left cusp fusion, right-non-coronary cusp fusion and left-non-coronary cusp fusion phenotypes); 2. The 2-sinus type (latero-lateral and antero-posterior phenotypes); and 3. The partial-fusion (forme fruste) type. The presence of raphe and the symmetry of the fused type phenotypes are critical aspects to describe. The International Consensus also recognizes 3 types of bicuspid valve-associated aortopathy: 1. The ascending phenotype; 2. The root phenotype; and 3. Extended phenotypes. Show less
Rutgers, J.J.; Banki, T.; Kamp, A. van der; Waterlander, T.J.; Scheijde-Vermeulen, M.A.; Heuvel-Eibrink, M.M. van den; ... ; Krijger, R.R. de 2021
Background Histopathological classification of Wilms tumors determines treatment regimen. Machine learning has been shown to contribute to histopathological classification in various malignancies... Show moreBackground Histopathological classification of Wilms tumors determines treatment regimen. Machine learning has been shown to contribute to histopathological classification in various malignancies but requires large numbers of manually annotated images and thus specific pathological knowledge. This study aimed to assess whether trained, inexperienced observers could contribute to reliable annotation of Wilms tumor components for classification performed by machine learning. Methods Four inexperienced observers (medical students) were trained in histopathology of normal kidneys and Wilms tumors by an experienced observer (pediatric pathologist). Twenty randomly selected scanned Wilms tumor-slides (from n = 1472 slides) were annotated, and annotations were independently classified by both the inexperienced observers and two experienced pediatric pathologists. Agreement between the six observers and for each tissue element was measured using kappa statistics (kappa). Results Pairwise interobserver agreement between all inexperienced and experienced observers was high (range: 0.845-0.950). The interobserver variability for the different histological elements, including all vital tumor components and therapy-related effects, showed high values for all kappa-coefficients (> 0.827). Conclusions Inexperienced observers can be trained to recognize specific histopathological tumor and tissue elements with high interobserver agreement with experienced observers. Nevertheless, supervision by experienced pathologists remains necessary. Results of this study can be used to facilitate more rapid progress for supervised machine learning-based algorithm development in pediatric pathology and beyond. Show less
Renewed calls for decolonizing anthropology in the 21st century raise the question of what work earlier waves of decolonization since the 1960s have left undone. Some of this work should focus on... Show moreRenewed calls for decolonizing anthropology in the 21st century raise the question of what work earlier waves of decolonization since the 1960s have left undone. Some of this work should focus on the classification of human differences, which figured prominently in all phases of the discipline’s history: as a methodology in its racist phases, as an object of study during its late colonial phase of professionalization, as self-critical reflexivity in the 1980s and 1990s, and as a renewed critique in the 21st century. Can a universal methodology of studying classifications of human kinds arise from the discipline’s past of colonial stereotyping? I argue affirmatively, through an approach that recognizes time as the epistemic condition that connects past and present positions to present and future methodologies. Firstly, my analysis distinguishes the parochial embedding in colonial culture of Durkheim and Mauss’ ideas about classification from their more universal intentions. This is then developed into a threefold reflexive and timeful methodology of studying classification’s nominal-descriptive, constructive, and interventionist dimensions—a process of adding temporality to the study of classification. Subsequently, Antenor Firmin’s 19th-century critique of racial classifications, and W. E. B. Du Bois’s theory of double consciousness help to show how this threefold methodology addresses the insufficiently theorized process of being classified and discriminated against through racial categories wielded by the powers that be. These arguments radicalize the essay’s timeful perspective by concluding that we need to avoid modernist uses of time as classification and adopt the aforementioned threefold methodology in order to put time in classifications of human kinds. This reverses modern positivism’s subordination to methodological rules of the epistemic conditions posed by contingent history and shows instead that the universal goals of methodology should be understood as a future ideal. Show less
Purpose To describe the clinical characteristics and outcome of polypoidal choroidal vasculopathy (PCV), also known as aneurysmal type 1 (sub-retinal pigment epithelium (RPE)) neovascularization,... Show morePurpose To describe the clinical characteristics and outcome of polypoidal choroidal vasculopathy (PCV), also known as aneurysmal type 1 (sub-retinal pigment epithelium (RPE)) neovascularization, in Caucasian patients. Methods Single-centre study in 66 Caucasian patients with a diagnosis of PCV based on optical coherence tomography scan and indocyanine green angiography. Clinical characteristics and multimodal imaging were collected and assessed by an experienced retina specialist. Results This study involved 74 eyes of 66 patients with PCV, with a mean age at onset of 73 years and a female preponderance of 66%. The mean number of polypoidal lesions per eye was 1 (range: 1-5 lesions), out of which 75% was located in the macula and 19% in the peripapillary region. Of the 74 eyes, 37 eyes (50%) had PCV associated with a drusenoidal neovascular age-related macular degeneration (AMD) phenotype (PCV-AMD) and 18 eyes (24%) had PCV associated with non-polypoidal type 1 choroidal neovascularization/branching vascular network (PCV-BVN) without signs of drusenoidal AMD, while 19 eyes (26%) had idiopathic, isolated PCV (iPCV). The mean subfoveal choroidal thickness measured in 22 patients was 245 mu m (range: 71-420 mu m). In 51% of patients, the initially performed therapy showed good anatomical recovery (resolution of intra- and subretinal fluid). Conclusions A spectrum of PCV (aneurysmal type 1/sub-RPE neovascularization) can be seen in Caucasian patients. PCV associated with a drusenoidal neovascular AMD phenotype in Caucasians is phenotypically and presumably pathophysiologically more associated with neovascular AMD (PCV-AMD: type A PCV). However, this may not be the case for patients with PCV with non-polypoidal type 1 choroidal neovascularization or BVN and no signs of drusenoidal AMD (PCV-BVN: type B PCV), and for patients with idiopathic PCV without associated drusen or BVN (iPCV; type C PCV). Most patients have a thin choroid, even when drusen are absent. For the entire patient group, a moderate anatomical recovery was observed after treatment. Show less
The aim of this European initiative is to facilitate a structured discussion to improve the next edition of the International Classification of Sleep Disorders (ICSD), particularly the chapter on... Show moreThe aim of this European initiative is to facilitate a structured discussion to improve the next edition of the International Classification of Sleep Disorders (ICSD), particularly the chapter on central disorders of hypersomnolence.The ultimate goal for a sleep disorders classification is to be based on the underlying neurobiological causes of the disorders with clear implication for treatment or, ideally, prevention and or healing. The current ICSD classification, published in 2014, inevitably has important shortcomings, largely reflecting the lack of knowledge about the precise neurobiological mechanisms underlying the majority of sleep disorders we currently delineate. Despite a clear rationale for the present structure, there remain important limitations that make it difficult to apply in routine clinical practice. Moreover, there are indications that the current structure may even prevent us from gaining relevant new knowledge to better understand certain sleep disorders and their neurobiological causes.We suggest the creation of a new consistent, complaint driven, hierarchical classification for central disorders of hypersomnolence; containing levels of certainty, and giving diagnostic tests, particularly the MSLT, a weighting based on its specificity and sensitivity in the diagnostic context.We propose and define three diagnostic categories (with levels of certainty): 1/"Narcolepsy" 2/"Idiopathic hypersomnia", 3/"Idiopathic excessive sleepiness" (with subtypes). (C) 2020 The Author(s). Published by Elsevier Ltd. Show less
Introduction: We recently proposed a scale for assessment of patient-relevant functional limitations following an episode of venous thromboembolism (VTE). Further development of this post-VTE... Show moreIntroduction: We recently proposed a scale for assessment of patient-relevant functional limitations following an episode of venous thromboembolism (VTE). Further development of this post-VTE functional status (PVFS) scale is still needed.Methods: Guided by the input of VTE experts and patients, we refined the PVFS scale and its accompanying manual, and attempted to acquire broad consensus on its use.Results: A Delphi analysis was performed involving 53 international VTE experts with diverse scientific and clinical backgrounds. In this process, the number of scale grades of the originally proposed PVFS scale was reduced and descriptions of the grades were improved. After these changes, a consensus was reached on the number/definitions of the grades, and method/timing of the scale assessment. The relevance and potential impact of the scale was confirmed in three focus groups totaling 18 VTE patients, who suggested additional changes to the manual, but not to the scale itself. Using the improved manual, the.-statistics between PVFS scale self-reporting and its assessment via the structured interview was 0.75 (95%CI 0.58-1.0), and 1.0 (95%CI 0.83-1.0) between independent raters of the recorded interview of 16 focus groups members.Conclusion: We improved the PVFS scale and demonstrated broad consensus on its relevance, optimal grades, and methods of assessing among international VTE experts and patients. The interobserver agreement of scale grade assignment was shown to be good-to-excellent. The PVFS scale may become an important outcome measure of functional impairment for quality of patient care and in future VTE trials. Show less
Nicholson, A.G.; Sauter, J.L.; Nowak, A.K.; Kindler, H.L.; Gill, R.R.; Remy-Jardin, M.; ... ; Galateau-Salle, F. 2020
Introduction: Molecular and immunologic breakthroughs are transforming the management of thoracic cancer, although advances have not been as marked for malignant pleural mesothelioma where... Show moreIntroduction: Molecular and immunologic breakthroughs are transforming the management of thoracic cancer, although advances have not been as marked for malignant pleural mesothelioma where pathologic diagnosis has been essentially limited to three histologic subtypes.Methods: A multidisciplinary group (pathologists, molecular biologists, surgeons, radiologists, and oncologists), sponsored by European Network for Rare Adult Solid Cancers/International Association for the Study of Lung Cancer, met in 2018 to critically review the current classification.Results: Recommendations include: (1) classification should be updated to include architectural patterns and stromal and cytologic features that refine prognostication; (2) subject to data accrual, malignant mesothelioma in situ could be an additional category; (3) grading of epithelioid malignant pleural mesotheliomas should be routinely undertaken; (4) favorable/unfavorable histologic characteristics should be routinely reported; (5) clinically relevant molecular data (programmed death ligand 1, BRCA 1 associated protein 1 [BAP1], and cyclin dependent kinase inhibitor 2A) should be incorporated into reports, if undertaken; (6) other molecular data should be accrued as part of future trials; (7) resection specimens (i.e., extended pleurectomy/decortication and extrapleural pneumonectomy) should be pathologically staged with smaller specimens being clinically staged; (8) ideally, at least three separate areas should be sampled from the pleural cavity, including areas of interest identified on pre-surgical imaging; (9) image-acquisition protocols/imaging terminology should be standardized to aid research/refine clinical staging; (10) multidisciplinary tumor boards should include pathologists to ensure appropriate treatment options are considered; (11) all histologic subtypes should be considered potential candidates for chemotherapy; (12) patients with sarcomatoid or biphasic mesothelioma should not be excluded from first-line clinical trials unless there is a compelling reason; (13) tumor subtyping should be further assessed in relation to duration of response to immunotherapy; and (14) systematic screening of all patients for germline mutations is not recommended, in the absence of a family history suspicious for BAP1 syndrome.Conclusions: These multidisciplinary recommendations for pathology classification and application will allow more informative pathologic reporting and potential risk stratification, to support clinical practice, research investigation and clinical trials. (C) 2019 International Association for the Study of Lung Cancer. Published by Elsevier Inc. All rights reserved. Show less
Vos, F. de; Schouten, T.M.; Koini, M.; Bouts, M.J.R.J.; Feis, R.A.; Lechner, A.; ... ; Rombouts, S.A.R.B. 2020
Anatomical magnetic resonance imaging (MRI), diffusion MRI and resting state functional MRI (rs-fMRI) have been used for Alzheimer's disease (AD) classification. These scans are typically used to... Show moreAnatomical magnetic resonance imaging (MRI), diffusion MRI and resting state functional MRI (rs-fMRI) have been used for Alzheimer's disease (AD) classification. These scans are typically used to build models for discriminating AD patients from control subjects, but it is not clear if these models can also discriminate AD in diverse clinical populations as found in memory clinics.To study this, we trained MRI-based AD classification models on a single centre data set consisting of AD patients (N = 76) and controls (N = 173), and used these models to assign AD scores to subjective memory complainers (N = 67), mild cognitive impairment (MCI) patients (N = 61), and AD patients (N = 61) from a multi-centre memory clinic data set. The anatomical MRI scans were used to calculate grey matter density, subcortical volumes and cortical thickness, the diffusion MRI scans were used to calculate fractional anisotropy, mean, axial and radial diffusivity, and the rs-fMRI scans were used to calculate functional connectivity between resting state networks and amplitude of low frequency fluctuations. Within the multi-centre memory clinic data set we removed scan site differences prior to applying the models.For all models, on average, the AD patients were assigned the highest AD scores, followed by MCI patients, and later followed by SMC subjects. The anatomical MRI models performed best, and the best performing anatomical MRI measure was grey matter density, separating SMC subjects from MCI patients with an AUC of 0.69, MCI patients from AD patients with an AUC of 0.70, and SMC patients from AD patients with an AUC of 0.86. The diffusion MRI models did not generalise well to the memory clinic data, possibly because of large scan site differences. The functional connectivity model separated SMC subjects and MCI patients relatively good (AUC = 0.66). The multimodal MRI model did not improve upon the anatomical MRI model.In conclusion, we showed that the grey matter density model generalises best to memory clinic subjects. When also considering the fact that grey matter density generally performs well in AD classification studies, this feature is probably the best MRI-based feature for AD diagnosis in clinical practice. Show less