The combination of ultrasound and AI is a relatively new branch in digital health. Ultrasound images contain a huge amount of data, and even experienced physicians sometimes struggle to cope with. The interpretation of ultrasound images requires extensive professional experience, and the training cycle of the ultrasound physicians is quite long. AI technology, therefore, has great potential to help with pattern detection to improve quality and efficiency, and to reduce human errors.


The lesion area is automatically identified and segmented from the ultrasound image, with the confident degree of malignancy. The diseases currently involved include thyroid nodules, mammary gland, carotid plaque, liver cancer and etc.

JF has established a database containing large-scale ultrasound image samples, which were used to train convolutional neural network powered by GPU-clusters to analyze ultrasound images. The computer-aided detection system tremendously reduces doctors' burden and enable patients to obtain more accurate diagnosis and personalized treatment plan recommendations.