Advanced analysis of multiparametric volumetric ultrafast ultrasound: a novel approach for non-invasive breast cancer diagnosis

Introduction

Tumour development follows a diversity of complementary biological pathways, including the modification of the tissue structure and vascularization, which are not currently captured by imaging techniques in the clinic. Breast cancer patients undergo a series of imaging sessions using complementary modalities, including ionizing mammographies.

In many cases, this imaging is not precise enough for a diagnosis, so that tissue samples in the form of biopsies are used to further characterize the tumour and determine the appropriate treatment. Beyond the pain and stress associated with biopsies, this complex process is costly and time-consuming, and delays the time to diagnosis.

Current Diagnostic Techniques

Ultrasound B-mode imaging is largely used in the diagnostic process of breast cancer, in part because it is low cost, portable, and largely available, as well as non-ionizing and non-invasive.

Our laboratory, Institute Physics for Medicine Paris, has recently developed several quantitative techniques allowing for the measurement of:

• Tissue stiffness
• Fiber organization
• Vascular mapping

These factors are all relevant to tumour development.

Proposed Approach

In this project, I propose a new approach to diagnosing breast cancer non-invasively by applying machine learning analysis to rich volumetric multiparametric maps of complementary tumour aspects, obtained using these innovative ultrafast ultrasound techniques.

The project will tackle the technological challenge of:

1. Integrating these techniques into a common acquisition and analysis framework.
2. Collecting a large clinical dataset.
3. Developing and validating a predictive malignancy model informing on tumour characteristics for the diagnosis.

Impact

This approach will open the door to fully virtual biopsies, impacting society on a large scale in terms of:

• Cost
• Diagnostic efficacy
• Patient comfort