Optomechanical Imaging System for Breast Cancer Detection

Abstract

Detection of gross anatomical changes is currently the principal method used to diagnose breast cancer. However, thousands of breast cancer cases are missed and thousands of unnecessary biopsies are performed every year. In this thesis, a new functional in vivo breast imaging system is presented. Results demonstrate that the developed system is sensitive to breast cancer and can differentiate between malignant and benign breast lesions. Angiogenesis, hypoxia, and enhanced stiffness associated with breast cancer are intrinsic contrast mechanisms that can be exploited for detection. The hypothesis is that studying tissue’s hemodynamics under conditions of rest, controlled provocations, and carbogen inspiration will produce adequate evidence to reveal the presence of breast cancer tumors. My thesis work comprised building a near infrared optomechanical tomographic breast imaging system, conducting a clinical study to assess the system's capability for lesion characterization, and analyzing the optical and mechanical signals to extract predictors for breast cancer. The imaging system examines both breasts simultaneously under conditions of rest and controlled provocations. Novel elements of the system include incorporation of high-density optical sensing heads (8192 channels) that include feedback-controlled force sensing articulating elements. In this design, the experimental conditions for both breasts are similar and stable; therefore, comparing the hemodynamic responses between the two breasts with high precision is possible. In order to better interpret how the applied pressure alters hemodynamics; the optomechanical response of the breast was explored during fine articulation, as a function of the applied force protocol. Comparisons between calculated internal pressure and reconstructed hemodynamic images show strong correlations. A clinical study was conducted on healthy subjects (n=28), as well as those with benign (n=33) and cancerous (n=23) breast lesions. Data analyses included characterizing the mechanical properties of the breast from the measured force and deformation, and deriving breast cancer biomarkers from the hemodynamic images under conditions of rest, mechanical articulation and carbogen inspiration. Receiver operator characteristic (ROC) analysis and binary logistic regression with leave-one-out cross validation were used to assess the efficacy of these biomarkers to diagnose cancer. The detectability of breast cancer using carbogen inspiration was enhanced by showing a larger rise in HbSat in comparison to healthy tissue. During the resting state, higher amplitude of vasomotor rhythms and greater degree of spatial heterogeneity were found in the tumor-bearing breast than in the contralateral healthy breast. In addition, mechanical articulations provided selective temporal contrast for cancer relative to the surrounding normal breast. A diagnostic accuracy of 93% was achieved using a weighted combination of predictors from baseline, carbogen inspiration, and mechanical articulations.