Motion score for spectral quality control of optoacoustic-ultrasound data

Motion can markedly affect the image quality acquired by handheld multispectral optoacoustic-ultrasound (MS-OPUS) hybrid systems. Since images at different wavelengths are acquired sequentially, tissue and operator movement can affect the spectra assigned to each volume element (voxels) in the images. To improve the accuracy of MS-OPUS spectral measurements, we sought to develop a method that could quantify and mitigate motion effects. We introduce Motion score, a post-processing algorithm for motion quantification and automated frame selection in MS-OPUS. Our method estimates motion in frames by tracking the similarity of co-registered ultrasound images and automatically selects low-motion frames. Algorithmic performance was validated on a dataset of ten scans, i.e., eight tissue-mimicking agar phantoms and two in vivo scans of arteries featuring different motion patterns. We show that Motion score had exceptionally high agreement with human annotators in selecting low-motion MS-OPUS frames, the gold standard process, while speeding up the process notably. Moreover, Motion score outperformed other methods such as optical flow or cross-correlation by a large margin. Furthermore, we show that frames with low Motion score were free of motion-related artefacts and have the optimal spectral quality. Finally, we provide an open-source implementation of Motion score as an off-the-shelf tool that can bring an immediate benefit to any clinical or preclinical MS-OPUS analysis pipeline.