This gitbook aims at motivating the application of Artificial Intelligence (AI) algorithms to the echOpen project. It is also meant to be a starting kit for whomever wants to contribute to the project, or simply wants to know more about machine learning algorithms applied to ultra-sound imaging in a real-time embedded environment.

Despite the technological advances a large part of world enjoys today, it remains that two third of the world population have no access whatsoever to a medical diagnosis device (TODO: ref). This situation is mainly due to the exorbitant price of these medical tools as well as the lack of a proper number of trained medical practitioners in some regions. Our goal is to produce `! We believe that open source, collective intelligence and artificial intelligence, combined together, could change this aberrant situation and literally save millions of lives.

echOpen is an open-source ultrasound probe connected to an android smartphone. Here, we are interested in the device as a platform for implementing novel AI applications for medical diagnosis. Not only modern smartphones offer a substantial amount of computing power for running deep models, both on CPU and GPU, but one can also take advantage of the variety of available sensors in order to improve the user experience, in addition of using the network capabilities for (consented) data repatriation, remote diagnosis, continuous update of the AI system etc.

The short-term objective is to endow echOpen with an AI-aided diagnosis system, capable of detecting organs, anomalies and in general, improving the image quality and the user experience of the medical practitioner. The long-term objectif is to extend its usability to non-practitioners, make deployable with little or no training.

echOpen benefits from several academic partnerships. As for the data necessary for training learning algorithms, we are closely working with our partner and incubator, the AP-HP in order to exploit anonymous data from their data lake, consisting of imaging their corresponding anonymous medical report.

More coming