Fraunhofer-Gesellschaft

Publica

Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Combination of sensor-embedded and secure server-distributed artificial intelligence for healthcare applications

 
: Gembaczka, Pierre; Heidemann, Burkhard; Bennertz, Bernhard; Gröting, Wolfgang; Norgall, Thomas; Seidl, Karsten

:

Biomedizinische Technik 64 (2019), Nr.s2, S.S14
ISSN: 0013-5585
ISSN: 1862-278X
German Society of Biomedical Engineering (Annual Meeting) <2019, Frankfurt/Main>
Bundesministerium fur Wirtschaft und Energie BMWi (Deutschland)
KI-Innovationswettbewerb; CarefuLKI
Verantwortete KI-Plattform für Gesundheit, Pflege und soziale Teilhabe
Englisch
Abstract
Fraunhofer IMS ()
Fraunhofer IIS ()
health avatar; embedded AI; neural network; artificial intelligence (AI); embedded systems

Abstract
The application of artificial intelligence (AI) in the areas of health, care and social participation offers great opportunities but also involves great challenges and risks. To counter them, extensive regulatory, ethical and data-security related requirements exist for data recording, storage and processing of respective personalized and patient-related data. “Artificial Intelligence as a Service” (AIaaS) is pushed for consumer applications by big companies which implies data storage on external database server. However, the available solutions do not meet the above mentioned requirements. Moreover, small and medium-sized enterprises (SMEs) in the field of healthcare fear the loss of data sovereignty and information outflow. In this paper, we propose a secure and resource-efficient approach by embedding AI directly close to the sensor in combination with secure and distributed data processing on local server or certified “Trusted Data Center”. For this puropose, we have developed the Artificial Intelligence for Embedded Systems (AIfES) platformindependent machine learning library in C programming language. It contains a fully configurable deep artificial neuronal network with feedforward structure. The library can be run directly on a microcontrollers and even allows to train the AI network on the microcontroller. Possible healthcare application include direct (pre-)processing of sensor data, calibration of sensors, pattern recognition and classification. Furthermore, virtual sensors can be developed by extracting a dependence of different measured variables (e.g., pulse and blood oxygen saturation using a pulse oximeter) to a new target variable (e.g., respiratory frequency). The work is part of the German Federal Ministry for Economic Affairs and Energy (BMWi) funded project Care[Ful]KI – Responsible AI platform for health, care and social participation. It aimes an integrated, on open standards based, legally certain and highly available AI data platform with competence and data pool. It shall allow innovative products and services applying anonymized or pseudonymized healthcare data.

: http://publica.fraunhofer.de/dokumente/N-582862.html