Neuroimaging and computing for better diagnosis and treatment of Alzheimer’s disease
Longitudinal atrophy flow in Alzheimer's disease
Inria is organising MICCAI 2012, the 15th International Conference on Medical Image Computing and Computer Assisted Intervention, which takes place from the 1st to the 5th of October.
We spoke to Prof. Giovanni Frisoni, neurologist at IRCCS Fatebenefratelli in Italy, and Inria researcher Marco Lorenzi, both of whom are presenting a paper about brain image analysis for better diagnosis and treatment of Alzheimer’s disease.
When analysing brain images, doctors look for regions of the brain that show abnormal amounts of atrophy, which can be caused by neurodegenerative diseases like Alzheimer’s. A new computer analysis method to help better analyse these images, devised by Marco Lorenzi, is based on an existing dynamic model that now includes time-sequenced cerebral images.
A virtual neuroimaging laboratory
The researchers are developing a “virtual laboratory” of sorts for imaging the brains of patients with Alzheimer’s disease. They are currently analysing around 1000 images (taken from 300 patients and 200 control subjects) but hope to have ten times as many images soon. Analyses on databases of this size require high-performance computing and a number of user-friendly algorithms.
“We can ‘see’ if a patient has Alzheimer’s by looking at the images and by using so-called imaging biomarkers,” explains Prof. Frisoni. “Such techniques are very recent and it will take another five years or so, we believe, before patients can routinely be diagnosed this way. There are three main biomarkers: the first is brain atrophy (which is the most dramatic phenomenon) because the tissue in some parts of the brain can shrink by up to 30%, even in the early stages of the disease. We can see this shrinking using high-resolution magnetic resonance imaging.
The second marker is glucose hypometabolism in some brain regions, while the third is the accumulation of beta-amyloid plaques in the cortex. Both these phenomena can be seen with PET. “For instance, we can observe the plaques, thought to be the main culprit behind Alzheimer’s, by injecting small molecules into the brain that selectively bind to the amyloid.”
Better understanding Alzheimer’s disease
Marco Lorenzi has a degree in mathematics and writes the complex algorithms and software to help doctors like Prof. Frisoni better analyse the digital brain images by focusing on the biomarkers. These algorithms will also allow researchers to understand the mechanisms behind Alzheimer’s, and in particular how the brain structure changes as the disease progresses.
“Our main challenge is that the numerical tools we develop must address the needs of doctors at all times, and the mathematics should always target the clinical problem,” explains Marco. “We must thus work in close collaboration and provide each other with constant feedback.”
Marco will finish his PhD by the end of this year. “During my PhD, we succeeded in developing a set of instruments that have shown promising results when it comes to analysing digital brain images of Alzheimer’s disease patients,” he says. “These instruments are mainly based on non-rigid registration of magnetic resonance images - a computational technique for modelling the structural changes of the brain through deformations of its shape and size. In particular, we have improved how the model statistics are computed, something that will help us better understand the pathology. We have also developed different methods for interpreting the disease and for quantifying its severity in a given patient. This might help a doctor to see which of his patients has Alzheimer’s and at what stage it is, so he can take the appropriate action.”
Avoiding costly clinical trials
The main goals of the research are to use the software and images to help develop effective drugs that stop the progression of the disease - or better still, preventative drugs that might be given to healthy subjects.
“Pharmaceutical companies currently develop drugs by undertaking large clinical trials. In the case of Alzheimer’s, these can involve hundreds of patients that are followed over a time period of several years,” explains Prof. Frisoni. “This is expensive and time-consuming. By analysing a large number of images with the help of specific algorithms and by using the specific biomarkers mentioned above, we might be able to reduce the number of patients studied by ten-fold and the time span to months rather than years.”