Skip to main content

Dementia modelling

Periodic Reporting for period 1 - DEMO (Dementia modelling)

Reporting period: 2016-09-01 to 2018-08-31

Dementia constitutes a major burden on society, both in monetary costs and the suffering of patients and their relatives. Alzheimer’s disease (AD), the most common form of dementia, is one of the most devastating healthcare problems faced by western society. The prevalence increases with the average lifetime year by year, and in most countries the population is aging.

Different types of dementia require different treatments. Cardiovascular treatments may for example prevent or slow vascular dementia (VaD), which is the second most common type of dementia. There are currently no disease-modifying treatments for AD, despite numerous promising drugs in development. Another challenge is that many of the dementias have clinically similar presentations and may co-exist (as many as 50% of AD patients may have VaD as well), making clinical diagnosis challenging. However, signs of dementia-related pathology are measurable using biomarkers far before clinical presentation, and the various types of dementia have quite different pathological bases.

The overall objective of the Dementia Modeling (DeMo) project is to develop new models for AD, VaD and mixed AD and VaD progression using, among others, novel imaging biomarkers of vascular pathology alongside already established AD imaging biomarkers. Such models enable identification and differentiation of at-risk subjects for effective treatment, and they can be used to assess disease progression for among others treatment monitorization.

The project emphasizes the use of historically widely used MRI scan types to enable use of historical imaging data in modeling. Algorithms are simultaneously developed for (1) robust disease progression modeling, and for the underlying necessary brain imaging biomarkers of VaD to be measured at each time point and entered to the disease progression model; i.e. (2) white matter lesion imaging biomarkers and (3) vascular pathology imaging biomarkers.

Three early stage researchers (ESRs) are trained while performing these three objectives (1-3) of the project.
The individual ESRs drive the progress of each of the respective objectives of the project, (1) robust disease progression modeling, (2) imaging biomarkers of white matter pathology, and (3) imaging biomarkers of vascular pathology. The ESRs each produced results during the reporting period that were disseminated by different means. Information about the venues and the results are, when available to the general public, provided via the links.

Scientific posters were presented at the Medical Image Computing Summer School (medICSS, https://medicss.cs.ucl.ac.uk/programme-3-2/); the Medical Imaging CDT Summer School ( http://www.onbicdt.ox.ac.uk/resources/imaging-cdt-summer-school-2018.html; the Medical Imaging Summer School (MISS, http://iplab.dmi.unict.it/miss/); the Grand Opening of the SCIENCE AI Centre at University of Copenhagen (https://di.ku.dk/english/event-calendar-2018/grand-opening-science-ai-centre/); the Second International Workshop on Modelling the Progression Of Neurological Disease (POND2018) (http://europond.eu/pond2018/); and the International conference on Medical Imaging with Deep Learning (MIDL, https://midl.amsterdam/). An oral presentation was given at MIDL and this presentation has been made available on Youtube by the conference organizers (https://youtu.be/k1xKKlvY14k). Two conference papers were published at MIDL and are both freely available (https://arxiv.org/pdf/1808.06519 https://arxiv.org/pdf/1808.05500). One of these papers has been selected for a special issue in the scientific journal Medical Image Analysis and will appear early 2019. Another conference paper has been accepted for the conference SPIE Medical Imaging, and the pre-print is freely available (https://arxiv.org/pdf/1810.01928). This paper will be presented as an oral at SPIE Medical Imaging in February 2019. Two medical image analysis challenge participations were performed. One ESR submitted predictions for The Alzheimer's Disease Prediction Of Longitudinal Evolution (TADPOLE) Challenge https://tadpole.grand-challenge.org/. The challenge results will appear early 2019. Another ESR submitted white matter hyperintensity segmentations for the Grand Challenge on MR Brain Segmentation (MRBrainS18, https://mrbrains18.isi.uu.nl/). Results will appear autumn 2018.

The main results within each of the 3 objectives so far are:
Objective 1) A new method for training recurrent deep neural networks while gracefully handling missing data which is an often-occurring problem. For details please see the following MIDL conference paper https://arxiv.org/pdf/1808.06519. See also the associated presentation on Youtube https://youtu.be/k1xKKlvY14k.
Objective 2) A new method for using specialized MRI sequences to steer the training of deep neural networks for white matter lesion segmentation in standard MRI sequence types. This results in better trained networks. For details please see the following MIDL conference paper https://arxiv.org/pdf/1808.05500.
Objective 3) A new method for handling extreme data imbalance when training a deep neural network for segmentation of microbleeds in standard MRI sequence types. Classical approach fails to detect small and rare structures such as microbleeds, as it is more cost-effective for an algorithm to miss such structures all the time rather than over-segment them. The proposed approach, that relies on two key concepts in the field of machine learning: curriculum learning and distillation losses, has been submitted for presentation at the Medical Imaging Computing and Computer Aided Intervention conference.
Progress beyond state-of-the-art for the three project objectives is

Objective 1) The new method makes training less influenced by missing data, and it can be applied to different types of recurrent neural networks without any need for changing these which makes it widely applicable. Moreover, contrary to the state-of-the-art, this method does not try to learn and utilize re-occurring missing value patterns, but instead tries to learn the true disease progression signal in the data.
Objective 2) The new method allows incorporating more advanced MRI sequence types during training to arrive at a better trained model, but without the need for these MRI sequences when the trained model is applied. State-of-the-art does not rely on the simple MRI sequences, but instead require the advanced as input when the model is applied, limiting the applicability in cohorts that does not have these.
Objective 3) Due to the rarity of open-access database that include manually annotated microbleed, we have been liaising with another European Project: “European Prevention of Alzheimer's Dementia Consortium” (http://ep-ad.org/) and now have access to a large set of data that can be used to develop novel methodologies.

It is expected that by the end of the project, the above progress is document in journal publications, and their combination has been realized and documented in a journal publication representing the overall project vision of novel joint modeling of AD and VaD using existing state-of-the-art AD imaging biomarkers and the novel VaD imging biomarkes. Such a model will represent significant new knowledge of the co-evolution of AD and VaD and provide a valuable future diagnostic and monitorization tool for clinics with potential for improved dementia treatment.
The goal of the Dementia Modelling (DEMO) project