From magnetic resonance brain structure imagery, the software program automatically rates Alzheimer’s disease patients with 70% assertiveness.
Universidad Nacional de Colombia (UNal) in Manizales Engineering doctoral candidate David Cárdenas Peña says that to develop the tool they used a database of approximately 1,000 magnetic resonances of patients diagnosed with Alzheimer’s disease.
The database was designed in three steps; in the first they debugged the images by filtering, lowering and deleting noise (digital effects of the image), without the cleaning process impacting the calculations. This was made possible by using Freesurfer, an open source software suite.
The second step was “parceling” which consisted of extracting brain structures of the 1,000 magnetic resonances. From the obtained images Cárdenas carried out a series of measurements (numeric vector) based on morphological features such as width, area and volume. This procedure was essential because Alzheimer’s causes neuronal death and loss of brain tissue and with time it dramatically transforms, inclusively shrinking, affecting all its functions.
In the third step they began to “teach” the software to distinguish if a person had cognitive deficiency, Alzheimer’s disease or was healthy. This process is known as machine learning, a branch of artificial intelligence so computers can generalize behaviors (learn) from information provided as examples, in this particular case imagery of people with the disease.
Universidad de Caldas Professor and Unisalud, UNal-Manizales IPS Physician Félix Peláez Cortés says that patients with signs of Alzheimer’s are advised to take the Mini–Mental State Examination (MMSE).
“The test consists of a questionnaire which involves superior cognitive functions such as space-time orientation; attention capacity, concentration and memory; calculation and attention; language capability and visual spatial perception; and the capability to follow basic instructions,” he added.
After they request a series of neuropsychological tests to assess learning and memory functions and then take tomography’s and magnetic resonances which help to determine anatomic alterations of the brain.
However currently, analysis of the images is performed in a standard manner, i.e. through visual inspection which turns it into a subjective process which requires the expertise of a specialist physician and an excellent quality of the image to avoid false diagnosis.
This situation was the cause of developing the software which effectiveness was compared to CADDementia, an online software evaluation framework for computer-aided diagnosis of dementia.
After verifying the effectiveness of the software in its first stage, the second stage will consist of a local medical team to validate the proposal as a diagnosis support tool for Colombian patients. With this goal and looking for funding for the research project the initiative was presented to the Colombian Administrative Department for Science Technology and Innovation (Colciencias, for its Spanish acronym) healthcare CTel project summons.
Agencia de Noticias UNAL