Proposal for pre-treatment CT/MRI derived radiomics and machine learning (ML) models for predicting treatment and immune-response in patients undergoing image-guided ablation and histotripsy for small renal masses and liver tumours. A proof-of-concept study.

Background: Histotripsy is a novel therapeutic technique that employs focused ultrasound waves to create precise, nonthermal, mechanical tissue destruction.

Histotripsy has recently received FDA approval for use in liver cancer and has been awarded one of the eight Innovative Devices Access Pathways by the government in February 2024.

In contrast to histotripsy, image-guided ablation is an established treatment for small renal cancers and liver cancers and is shown to have comparable oncological control and superior complication profile when compared to traditional surgical intervention.

Despite that, incomplete treatment response is still common, especially in non-clear cell renal cancers, hepatocellular carcinoma and colorectal liver metastases.

Furthermore, treatment response is highly dependent on tumour type, location, and complexity, as well as patient factors such as skin-to-lesion distance and co-morbidities.

Finally, image-guided ablation and histotripsy is known to trigger specific immune responses against targeted tumours within the body.

Radiomics, a subset of machine learning, involves the quantitative extraction of features from images such as CT and MRI scans, which enables the prediction of clinical outcomes and treatment response.

Aims: This study aims to develop a machine learning model using radiomics to predict treatment response for image-guided ablation in small renal cancer and liver cancers.

Additionally, to explore the possibility of a proof-of-concept to transfer and apply the model and learning to histotripsy as a novel treatment modality to predict treatment and immune-response after treatment. Methods: This is a retrospective case series and a proof-of-concept study.

Over 200 patients undergoing cryoablation for renal tumours and over 200 patients undergoing microwave ablation for liver tumours will be included.

Separate patients enrolled in the CAIN trial and the #HOPE4LIVER trial in Leeds will be included in the histotripsy arm.

A machine-learning model will be developed using pre-treatment and follow-up CT or MRI scans to predict treatment response.

The model will be evaluated on unseen data to ensure validity and then applied to the histotripsy cohort as a feasibility, proof-of-concept study. Radiomic features related to immune response will be evaluated descriptively.

How the results of this research will be used: A model to predict treatment response and, if feasible, immune response is important for informed decision-making during multidisciplinary team meetings to determine the best approach to personalised treatment.

The research results will be used to determine the feasibility of applying a predictive model to histotripsy data to design a future study validating or refining the model on histotripsy.