Concurrent Optical Spectroscopy and Breast MRI to Improve Diagnosis without Contrast Injection

Dartmouth College, Hanover NH

Investigators

Linked publications, trials & patents

Abstract

We request support to continue developing and evaluating a combined near-infrared spectroscopic tomography (NIRST) and magnetic resonance imaging (MRI) platform. Our aim is to explore whether this concurrent modal- ity—which does not require contrast injection or involve ionizing radiation—can achieve acceptable diagnostic performance. Research during the current funding period yielded: (i) An MRI-guided (MRg)-NIRST system that can acquire MRI and NIRST data from up to 4,096 source-detector positions over the entire breast at six wave- lengths1; (ii) An MRI-compatible, wearable breast optical interface featuring eight flexible optical strips, each equipped with six photodetectors (PDs) and six side-firing fiber probes, designed to accommodate breasts of various sizes and shapes2; (iii) A new type of tissue phantom that is permanent, easily molded into various shapes, has similar elastic and spectral properties (in the wavelength range of 600–850 nm) to breast tissue, and provides MRI or CT contrast for validating the NIRST system;3 (iv) Reconstructed images of phantoms with inclusions ranging from 10–25 mm in diameter with small errors in estimated inclusion diameter and con- trast in total hemoglobin (HbT); (v) HbT estimates from reconstructed images of healthy subjects which ranged between 8.0–25.2 μM and aligned with previous imaging studies;1,4 and (vi) A deep learning approach (DL) that reconstructs 3D MRg-NIRST images in 1.4 seconds after the network is trained on synthetic data. These achievements have provided valuable technical insights, demonstrating the potential to overcome barriers to clinical acceptance of MRg-NIRST and motivating us to continue the project. The hardware platform proposed in Aim 1 resolves issues caused by MRI surface artifacts (due to the strips) and estimates additional NIRST parameters—such as oxy-hemoglobin, water, lipid and collagen content—by increasing the number of laser sources and extending the wavelength range. Advanced 3D image reconstruction methods, using either deep learning or traditional techniques or their combination, will be developed in Aim 2, and validated on phantoms, healthy volunteers, and patients with known breast abnormalities. Finally, we will evaluate the diagnostic perfor- mance of MRg-NIRST, with and without contrast injection, in a clinical study involving 112 patients with breast abnormalities, as part of Aim 3. This project builds upon an extensive translational breast imaging infrastructure we have developed through which collaborations with experts in diagnostic radiology and biomedical engineering occur that will advance the proposed imaging technology.