Machine Learning of Multi-Modal Tumor Imaging Reveals Trajectories of Response to Precision Treatment
Nesrin Mansouri1, Daniel Balvay1, Omar Zenteno1
1INSERM, PARCC, Université Paris Cité, F-75015 Paris, France.
Abstract:
The standard assessment of response to cancer treatments is based on gross tumor characteristics, such as tumor size or glycolysis, which provide very indirect information about the effect of precision treatments on the pharmacological targets of tumors. Several advanced imaging modalities allow for the visualization of targeted tumor hallmarks. Descriptors extracted from these images can help establishing new classifications of precision treatment response. We propose a machine learning (ML) framework to analyze metabolic-anatomical-vascular imaging features from positron emission tomography, ultrafast Doppler, and computed tomography in a mouse model of paraganglioma undergoing anti-angiogenic treatment with sunitinib. Imaging features from the follow-up of sunitinib-treated (n = 8, imaged once-per-week/6-weeks) and sham-treated (n = 8, imaged once-per-week/3-weeks) mice groups were dimensionally reduced and analyzed with hierarchical clustering Analysis (HCA). The classes extracted from HCA were used with 10 ML classifiers to find a generalized tumor stage prediction model, which was validated with an independent dataset of sunitinib-treated mice. HCA provided three stages of treatment response that were validated using the best-performing ML classifier. The Gaussian naive Bayes classifier showed the best performance, with a training accuracy of 98.7 and an average area under curve of 100. Our results show that metabolic-anatomical-vascular markers allow defining treatment response trajectories that reflect the efficacy of an anti-angiogenic drug on the tumor target hallmark.
Insights
Precision cancer treatment response can be better assessed using advanced imaging and machine learning. This study developed a framework to analyze metabolic, anatomical, and vascular features for improved tumor staging.
Area of Science:
- Oncology
- Medical Imaging
- Machine Learning
Background:
- Standard cancer treatment response assessment relies on indirect tumor characteristics like size, which are insufficient for evaluating precision therapies.
- Advanced imaging modalities offer visualization of tumor-specific hallmarks, enabling new classifications for precision treatment response.
- Precision medicine requires novel methods to accurately track treatment efficacy at the molecular and cellular levels.
Purpose of the Study:
- To propose a machine learning (ML) framework for analyzing metabolic-anatomical-vascular imaging features.
- To develop a generalized tumor stage prediction model for precision cancer treatment response.
- To validate the framework using a mouse model of paraganglioma treated with anti-angiogenic sunitinib.
Main Methods:
- Utilized positron emission tomography, ultrafast Doppler, and computed tomography imaging in a mouse model.
- Applied dimensionality reduction and hierarchical clustering analysis (HCA) to imaging features.
- Employed 10 ML classifiers for tumor stage prediction and validated with an independent dataset.
Main Results:
- Hierarchical clustering analysis identified three distinct stages of treatment response.
- The Gaussian naive Bayes classifier demonstrated superior performance with 98.7% training accuracy and 100% average area under the curve.
- The ML framework successfully predicted tumor response stages based on imaging features.
Conclusions:
- Metabolic-anatomical-vascular imaging markers can define treatment response trajectories.
- These markers accurately reflect the efficacy of anti-angiogenic drugs on tumor targets.
- The developed ML framework provides a robust method for assessing precision cancer treatment response.
