Currently, disorders such as dementia or Alzheimer's cannot be cured. If we rightly hope to live to be 85 years or older today, the risk of developing Alzheimer's or dementia is almost 50%. In Europe, Alzheimer's causes annual costs of around 800 billion euros, which could quintuple by 2050 due to the aging of the baby boomer generation. The course of the disease can only be slowed down if the necessary changes in personal lifestyle are made quickly and consistently after a very early diagnosis. Alzheimer's disease is one of the greatest medical and social challenges. Artificial Intelligence can make an important contribution here.
The DFKI spin-off ki elements supports doctors and patients in detecting early signs of incipient dementia. Non-invasive speech-based tests have been developed for cognitive health screening. AI tools are used for speech recognition, semantic processing, and analysis of paralinguistic phenomena (speech speed, word fluency, hesitations). Machine learning assists in the evaluation.
"We are very proud to be one of the ten winners of the 2019 competition 'Excellent Landmark in the Land of Ideas' and would like to express our sincere thanks for the award. We are particularly pleased to receive the award at a time when our technology is being used in the first medical practices, clinics and studies in Germany and France," said ki elements co-founder and DFKI researcher Nicklas Linz.
Neurocognitive examinations are still carried out with pen and paper and manually evaluated by the specialist in the clinic, which makes comparability more difficult and reduces the diagnostic significance. The software Δelta (pronounced: Delta) digitalizes neurocognitive tests such as word fluid tests ("Count as many animals as possible in 60 seconds") or image descriptions. The specialists are relieved of repetitive manual work so that the decision of the specialist staff can be made in the best possible and most efficient way.
Δelta uses automatic language processing and machine learning methods to "hear" indicators for various cognitive disorders from the spoken language. Language is recorded directly in the expert app and processed using signal processing methods. Spoken words are recognized and converted into text. Clinically relevant markers can be extracted by automatic linguistic analysis. Both automatic speech recognition and linguistic feature extraction are based on deep learning procedures, which makes it easy to integrate multiple languages and extract clinically relevant features without manual effort.
The basis for a diagnostic tool for identifying cognitive deficits was laid in the innovation project "ELEMENT", which was funded by the European Institute for Innovation and Technology (EIT). The technology was developed and tested together with the partners INRIA (Institut National de Recherche en Informatique et en Automatique), the Association Innovation Alzheimer in Nice, and the Saarland University Hospital. Since 2018, Δelta has been entitled to carry the CE mark.
Further information
land-der-ideen.de/en
ki-elements.de
Contact ki-elements
Nicklas Linz
Saarland Informatics Campus, Geb. D 3_2
D-66123 Saarbrücken, Germany
Tel.: +49 681 85775 5073
E-Mail: Nicklas.Linz@dfki.de