Philips Research Technology Backgrounder

Computer Aided Diagnosis of Neurodegenerative Disease

An innovative system jointly developed by the University Medical Center Hamburg-Eppendorf and Philips

As a result of improved living standards and medical care, people now live longer than ever before. The average age expectancy for men and women in Western Europe, for example, is already almost 80 years. However, many elderly people are not disease free. A large number of them suffer chronic disease, and one of the most debilitating in terms of quality-of-life is dementia.

Dementia currently affects well over 25 million people worldwide, and with the demographic shift to older populations it is set to reach epidemic proportions unless effective treatments can be found. The cost of today’s treatment is already putting massive burdens on healthcare authorities and the societal impact on patients and their caregivers is immense.

Dementia is the end result of a number of progressive degenerative diseases of the brain. These diseases are associated with changes in brain chemistry that are thought to begin ten or more years before patients suffer symptoms of cognitive impairment. The most common neurodegenerative diseases are Alzheimer’s Disease, Lewy-body Dementia and Frontotemporal Dementia. These three diseases account for around 60%, 15% and 10% of all dementia cases respectively and are at the moment incurable. Current treatments, such as Cholinesterase inhibitors, provide symptomatic relief in the mild to moderate stages of the disease but do not arrest its progression. Alternative therapeutic options, currently under development, attempt to interrupt the disease process at an earlier stage. There is therefore a growing need for the early detection of neurodegenerative disease and reliable diagnosis of its underlying type.

“It is quite clear that in the next few years there will be new medications for Alzheimer’s disease and other forms of dementia that will have to be given much earlier on if they are to be effective,” says Dr. Holger Jahn, Consultant in Clinical Psychiatry at the University Medical Center Hamburg-Eppendorf (UKE). “Unfortunately, at the moment we make the diagnosis of dementia quite late on in the course of the disease.”

The highly effective blood-brain barrier in the human body makes it currently not possible to detect biomarkers for neurodegenerative disease in blood samples. In-vitro analysis of cerebrospinal fluid (obtained via a lumbar puncture) and in-vivo FDG-PET imaging (Positron Emission Tomography using the tracer Fluorodeoxyglucose) are therefore used as diagnostic tools.

By providing a quantitative indication of the amount of glucose being used in different parts of the brain to fuel brain activity, FDG-PET scans can reveal abnormal brain conditions. However, the resultant FDG-PET images are difficult to interpret. Particularly in the early stages of neurodegenerative disease, it requires the expertise of a highly skilled specialist to make an accurate diagnosis. The limited number of such specialists means that an easier, faster and more convenient method of diagnosing neurodegenerative disease is required.

Computer Aided Diagnosis (CAD)
The solution jointly developed by the University Medical Center Hamburg-Eppendorf (UKE) and Philips Research combines advanced image processing and computer learning techniques with a database of reference brain scans to create an expert system that interprets FDG-PET images automatically to assist clinicians in the accurate differential diagnosis of neurodegenerative diseases.

The first stage in the process is automatic alignment of FDG-PET brain-scan images to a reference model of the human brain – a process known as ‘elastic registration’. This not only corrects for the orientation of a particular patient’s head in the scanner but also for anatomical variation in the brains of different patients (different shape and/or size etc.). The ‘elastically registered’ FDG-PET image is then compared with a set of reference brain scans that represent typical disease patterns, such as that for Alzheimer’s disease, Lewy-body Dementia or Frontotemporal Dementia. Depending on the fit between the patient’s scan and these reference scans, the system provides clinicians with a probable diagnosis.

In developing this automated diagnostic system, emphasis has been placed on making it as highly integrated and user-friendly as possible so that less experienced clinicians can use it as a ‘second-reader’ to achieve the same high level of accuracy in their diagnoses as clinical experts.

The system has already been retrospectively tested using historical brain-scan images of patients with known disease outcomes. It is now being clinically evaluated by running it alongside UKE’s existing dementia diagnosis procedures. This clinical evaluation will also be used to fine-tune the system’s ability to differentiate between the three most common types of dementia by expanding the database of reference images that the system can refer to and by incorporating more of UKE’s expert knowledge into the computer program.

The system is also being extended to combine the metabolic information obtained from FDG-PET scans with structural information revealed by imaging the patient’s brain with an MRI (Magnetic Resonance Imaging) scanner. MRI scans provide very clear and detailed information about brain structure and this can be used both to improve the elastic registration of FDG-PET images and to provide additional diagnostic clues.

Screening and therapy monitoring
The potential of the UKE/Philips computer aided diagnostic system to detect dementia-related diseases well before patients begin to suffer symptoms could make it a powerful screening tool. Although it is unlikely to be used for whole population screening, it could be used to screen people who are considered at risk of contracting neurodegenerative diseases such as Alzheimer’s – for example, those with specific gene constellations or a family history of neurodegenerative disease.

The system could also provide a convenient and easy-to-use means of therapy response monitoring, making it a powerful tool in the development of new drugs to control or cure neurodegenerative diseases of the brain. Many of the new drugs under development are specifically targeted for use in the very early stages of diseases such as Alzheimer’s, which means that their efficacy cannot be evaluated by symptom-based studies.

Personalized medicine
Because many of these drugs are powerful ‘mind-changing’ substances, the development of a pathology-based system that can be used to monitor individual patient responses would help doctors to devise patient-specific drug therapies that maximize efficacy while minimizing side effects. This will bring the current trend of personalized medicine to yet another area of clinical care, improving patient outcomes and reducing healthcare costs for both the treatment and aftercare of dementia related diseases. It will also help to prevent today’s high incidence of dementia in elderly people from reaching the epidemic proportions that so many people fear.