Migration of Deep Learning Models Across Ultrasound Scanners.

Abstract

A transfer function approach has recently proven effective for calibrating deep learning (DL) algorithms in quantitative ultrasound (QUS), addressing data shifts at both the acquisition and machine levels. Expanding on this approach, we develop a strategy to acquire the functionality of a DL model from one ultrasound machine and implement it on another in a black-box setting, in the context of QUS. This demonstrates the ease with which the functionality of a DL model can be transferred between machines. While the proposed approach can also assist regulatory bodies in comparing and approving DL models, it also highlights the security risks associated with deploying such models in a commercial scanner for clinical use. The method is a black-box unsupervised domain adaptation technique that integrates the transfer function approach with an iterative schema. It does not utilize any information related to model internals but it solely relies on the availability of an input-output interface. Additionally, we assume the availability of unlabeled data from a testing machine. This scenario could become relevant as companies begin deploying their DL functionalities for clinical use. In the experiments, we used a SonixOne and a Verasonics machine. The model was trained on SonixOne data, and its functionality was then transferred to the Verasonics machine. The proposed method successfully transferred the functionality to the Verasonics machine, achieving a remarkable 98 percent classification accuracy in a binary decision task. This study underscores the need to establish security measures prior to deploying DL models in clinical settings.