A new artificial intelligence (AI) technique aimed at preventing medical imaging cyberthreats will be presented on August 26 at the 2020 International Conference on Artificial Intelligence in Medicine (AIME 2020).
The method analyzes the instructions sent from a host computer that controls CT, MRI, and ultrasound scanners to detect malicious or abnormal operating directives and human and system errors. The tool detects context-free (CF) anomalous instructions, which are unlikely values or directions to give very high radiation doses, as well as context-sensitive (CS) anomalous instructions, which are normal values but abnormal to a specific context, like mismatching a potential diagnosis, according to researchers from Ben-Gurion University.
Adding the CS method to the CF layer improved the overall anomaly detection performance from an F1 score of 71.6% to between 82% and 99%, depending on the clinical objective or the body part, according to the researchers.
The CF method evaluated 8,277 recorded CT instructions using 14 unsupervised anomaly detection algorithms. Then, they evaluated the CS layer for four different types of clinical objective contexts, using five supervised classification algorithms for each context.
