Mobile ultrasound technology developer Socionext and SOINN said that the Japanese artificial intelligence (AI) start-up's SOINN (Self-Organizing Incremental Neural Network) AI technology has produced favorable results in learning to calculate subcutaneous fat thickness on abdominal ultrasound images.
Trained via deep learning on about 700 images produced by Socionext's viewphii mobile ultrasound system, the SOINN software was able to accurately determine fat tissue thickness in 80% of images within a 5% margin of error, according to the companies. Noting that human errors such as numerical input mistakes were a common occurrence in the study, the firms said they believe the AI software has the potential to assist in reading the exams and detecting human errors in handling the images.
Socionext and SOINN said they will present the initial results this week at Medtec Japan in Tokyo.
