Diagnosis of breast cancer biopsies using quantitative phase imaging

Abstract

The standard practice in the histopathology of breast cancers is to examine a hematoxylin and eosin (H&E) stained tissue biopsy under a microscope. The pathologist looks at certain morphological features, visible under the stain, to diagnose whether a tumor is benign or malignant. This determination is made based on qualitative inspection making it subject to investigator bias. Furthermore, since this method requires a microscopic examination by the pathologist it suffers from low throughput. A quantitative, label-free and high throughput method for detection of these morphological features from images of tissue biopsies is, hence, highly desirable as it would assist the pathologist in making a quicker and more accurate diagnosis of cancers. We present here preliminary results showing the potential of using quantitative phase imaging for breast cancer screening and help with differential diagnosis. We generated optical path length maps of unstained breast tissue biopsies using Spatial Light Interference Microscopy (SLIM). As a first step towards diagnosis based on quantitative phase imaging, we carried out a qualitative evaluation of the imaging resolution and contrast of our label-free phase images. These images were shown to two pathologists who marked the tumors present in tissue as either benign or malignant. This diagnosis was then compared against the diagnosis of the two pathologists on H&E stained tissue images and the number of agreements were counted. In our experiment, the agreement between SLIM and H&E based diagnosis was measured to be 88%. Our preliminary results demonstrate the potential and promise of SLIM for a push in the future towards quantitative, label-free and high throughput diagnosis.

Identification and Staging of B-Cell Acute Lymphoblastic Leukemia Using Quantitative Phase Imaging and Machine Learning

“…3,7,8 This reagent-free modality has recently gained traction toward potential biomedical applications including cancer diagnosis. 9,10 However, earlier studies have primarily relied on interpreting the QPI results in terms of a few principal morphological characteristics. Over the last 5 years, however, several laboratories including our own have sought to shift the paradigm by utilizing machine learning (ML) and artificial intelligence (AI) in analyzing QPI data.…”

“…Quantitative phase imaging (QPI) is a rapidly emerging imaging modality that operates on unlabeled specimens, permits fast measurements even with low-intensity illumination (thereby reducing phototoxicity concerns), and, as such, is potentially promising for this task. , By constructing quantitative maps of optical path length delays introduced by the specimen, QPI facilitates the assessment of morphology and dynamics of live cells while obviating the need for exogenous contrast agents . Significantly, QPI lends itself to rapid imaging, and some of the earliest measurements focused on membrane fluctuations in healthy and diseased human red blood cells with nanometer-scale sensitivity at the millisecond time scale. − The interpretation of the phase signal has offered key insights into the physiological processes of live cells, such as cellular dry mass, transmembrane water flux, and water content changes. ,, This reagent-free modality has recently gained traction toward potential biomedical applications including cancer diagnosis. , However, earlier studies have primarily relied on interpreting the QPI results in terms of a few principal morphological characteristics. Over the last 5 years, however, several laboratories including our own have sought to shift the paradigm by utilizing machine learning (ML) and artificial intelligence (AI) in analyzing QPI data. − The ability of QPI to provide volumes of high-dimension imaging data makes it conducive to the application of ML for tasks involving cell classification.…”