Breast cancer FUS ablation monitoring and predicting BBB opening using multi-parameter classifiers

Focused ultrasound (FUS) is a non-invasive modality to treat breast cancer. In conjunction with microbubbles, FUS has also been shown capable of opening the blood–brain barrier (BBB) (BBBO). In this study, we provide an overview of employing multi-parameter classifiers to ensure best monitoring performance of FUS. In breast cancer, Harmonic Motion imaging-guided focused ultrasound (HMIgFUS) utilizes high-intensity, amplitude-modulated acoustic radiation force for ablation, with simultaneous monitoring of tissue stiffness. To improve FUS monitoring, we developed FUS-net to remove FUS interference from RF signals. FUS-net filtering successfully filtered continuous FUS output by preserving frequency information, compared with notch filtering. In addition, a multimodality segmentation network combining HMI and B-mode, HMINet, was developed to accurately segment clinical lesions in real time (2 Hz). HMInet was also shown to automatically detect lesions due to the higher mechanical contrast and quantify the lesion volume during ablation. In FUS-BBBO, neuronavigation-guided FUS and microbubbles were used to induce BBBO in Diffuse Midline Glioma patients and a multi-parameter classifier used cavitation metrics to predict outcomes. The multi-parameter classifier identified the harmonic component of PAM-derived cavitation dose as the most important factor in predicting BBBO, with ROC of up to 0.875 for a single-feature classifier.