3D wearable NIR spectral tomography for early prediction of breast cancer’s residual cancer burden after neoadjuvant chemotherapy

Prediction of Residual Cancer Burden (RCB) early in treatment can enhance long-term survival outcomes for breast cancer patients undergoing neoadjuvant chemotherapy (NAC) and their long-term survival.1-3 Near-infra- red spectral (NIRS) tomography (NIRST) provides operational advantages over other imaging technologies (e.g.,

mammography, ultrasound, Magnetic Resonance Imaging (MRI), CT, Positron Emission Tomography) in the NAC setting because it is noninvasive, portable, low cost, and does not use ionizing radiation or exogenous contrast agents. Crucially, NIRS captures biophysical changes in tissue occurring in the vascular, intra- and

extra-cellular matrix compartments. These subtle changes signal a tumor's early response to NAC, even before visible tumor size modifications occur.4-7 Compared to other optical imaging modalities that use reflectance light, which limits tissue imaging depth8, 9, or require high-power lasers posing safety risks10, NIRST captures diffused

tomographic signals, allowing for deeper breast tumor sensing. In our previous studies, we have pioneered a NIRST system that quantifies vascular changes in the breast during NAC, highlighting tumor response within about two minutes as the patient sits in a semi-reclined position.11,12 This system correlates tumor NIRST

changes with clinical outcomes based on residual cancer burden (RCB) classification.13 Our clinical NIRST data from 35 women on NAC persuades us of NIRST's potential to revolutionize the clinical approach to these patients. This belief stems from recently published results which showcase the normalized percentage change of total

hemoglobin in a tumor (ΔHbT%) by the end of the first cycle as a compelling biomarker that discerns either RCB-0 or RCB-II from all cases in other classes (p £ 0.001).13 A thorough multi-class Receiver Operating Curve analysis offers an Area Under the Curve of 0.80, emphasizing the precision of ΔHbT% in distinguishing between

RCB classes. We envisage elevating our system's efficiency by integrating flexible circuit strips through a sleek, wearable 64-channel breast interface and applying 3D image reconstruction that reflects full tumor response (instead of partial tumor volume) – the focal point of our proposed Aim 3. Our previous data further indicate that

NIRST's diagnostic performance remains undiminished when relying solely on continuous-wave signals and patient-specific scattering estimates derived from mammograms or MRI defined breast density.14 These insights have shaped the hardware platform proposed in Aim 1. However, the current iteration of our NIRST system faces

challenges from its dependency on large-core fiber bundles for receiving light from the patient's breast. By phas- ing out these cumbersome fiber bundles, we open avenues for novel 3D image reconstruction methods en- hanced with deep learning—as set out in Aim 2. In essence, this project focusses on technological innovations

aimed at creating a comprehensive NIRST breast imaging platform, designed to forecast RCB in early stages of NAC. Accordingly, we are transitioning this platform from academic exploration to a clinically practical solution, while concurrently seeking additional funding to facilitate expansive clinical studies in the foreseeable future.