Funded by the European Union – Next Generation EU
In the biomedical sector, the quality verification of ultrasound probes is essential to ensure diagnostic accuracy. However, the tests currently in use are often performed manually, resulting in variability, slowness, and poor repeatability. These tests require a strong subjective component from the operator, generating results that are not standardized and difficult to replicate. The need for efficiency and precision has therefore driven the development of automated and more reliable solutions.
The Soundflex project introduced an innovative solution based on the combination of collaborative robotics and artificial intelligence to automate the entire process of testing ultrasound probes. After an in-depth analysis of manual operating methods, a system was developed that involves the automatic positioning of the probe through a robotic arm – cobot – and the analysis of ultrasound images through neural networks trained on specific data from different types of probes. In particular, algorithms were created for probe-cone alignment, initially with a classification and regression approach, then optimized into a purely classification version. The image quality test was also automated, using a ResNet18 neural network to compare the acquired images with a reference Phantom. The entire system operates locally, in a Docker environment, ensuring fast response times and data security.
Thanks to this solution, the testing process has been completely automated, eliminating the need for manual intervention and significantly increasing the repeatability and reliability of the results. The developed neural networks achieved an accuracy of over 99% in image classification, thus improving quality control. Furthermore, Prensilia’s ability to independently adapt the algorithms to new types of probes makes the system highly scalable and efficient, with tangible benefits in terms of reducing time and errors.
Funded by the European Union – Next Generation EU
Don’t miss any updates on tenders, job openings, and CIM news directly in your feed.
"*" indicates required fields