Prostate cancer histopathology with label-free multispectral deep UV microscopy quantifies phenotypes of tumor grade and aggressiveness

Abstract

Identifying prostate cancer patients that are harboring aggressive forms of prostate cancer remains a significant clinical challenge. To help address this problem, we develop an approach based on multispectral deep-ultraviolet (UV) microscopy that provides novel quantitative insight into the aggressiveness and grade of this disease. First, we find that UV spectral signatures from endogenous molecules give rise to a phenotypical continuum that differentiates critical structures of thin tissue sections with subcellular spatial resolution, including nuclei, cytoplasm, stroma, basal cells, nerves, and inflammation. Further, we show that this phenotypical continuum can be applied as a surrogate biomarker of prostate cancer malignancy, where patients with the most aggressive tumors show a ubiquitous glandular phenotypical shift. Lastly, we adapt a two-part Cycle-consistent Generative Adversarial Network to translate the label-free deep-UV images into virtual hematoxylin and eosin (H&E) stained images. Agreement between the virtual H&E images and the gold standard H&E-stained tissue sections is evaluated by a panel of pathologists who find that the two modalities are in excellent agreement. This work has significant implications towards improving our ability to objectively quantify prostate cancer grade and aggressiveness, thus improving the management and clinical outcomes of prostate cancer patients. This same approach can also be applied broadly in other tumor types to achieve low-cost, stain-free, quantitative histopathological analysis.

Deep UV Microscopy Identifies Prostatic Basal Cells: An Important Biomarker for Prostate Cancer Diagnostics

“…On the other hand, PC 2 and 4 resemble the absorption spectra of proteins, while PC 3 is similar to the inverted spectra of nucleic acid absorption [ 28 ]. However, as we have outlined previously [ 29 ], these PCs cannot solely be attributed to these molecular components, and we do not rule out contributions from other molecules.…”

“…To leverage the spectral response for molecular imaging, we take the projections of the spectra onto the first three principal components (which contain over 99% of the data variance) and do a coordinate transformation from Cartesian coordinates to spherical coordinates. With this transformation, the shape of the spectrum of each spatial pixel in the image, given by the biochemical composition within that pixel, can be accurately described using only the azimuth ( ) and elevation ( ) angles [ 29 ]. The radius in spherical coordinates then serves as a relative measure of concentration.…”

“…Other label-free methods have been proposed for virtual H&E staining based on deep learning (e.g., using phase contrast [49], autofluorescence [50], and brightfield [51]), but to our knowledge, deep UV microscopy is the only approach that has demonstrated the ability to provide both H&E and 9 BME Frontiers highly specific p63 IHC simultaneously (or just label-free p63 IHC virtual stains). Further, unlike virtual H&E staining, the proposed virtual IHC stain does not rely on deep learning methods which, while promising, have their own unique challenges [29,[50][51][52]. For instance, deep learning requires a lot of matched labeled data in order to train a conversion model, thus inherent artifacts in the images-like weak or negative p63 expression and faded staining-can be learned.…”

“…Recently, deep ultraviolet (UV) imaging has reemerged as an important tool for fast, simple, and reliable label-free molecular imaging based on light attenuation in this spectral region, providing a number of important advantages over conventional optical microscopy modalities (a comparison is provided in the supplemental material). For example, many biomolecules critical to cellular function and the development of diseases have distinct absorption spectral signatures in the deep UV region of the spectrum, which enable high-contrast, quantitative molecular imaging and phenotyping [ 28 , 29 ]. Further, unlike other label-free molecular imaging methods, deep UV microscopy is fast, offers high spatial resolution owing to its shorter wavelength, and does not require expensive laser systems.…”

“…(Note that the resolution, , of a microscope depends on the wavelength, , and numerical aperture, NA, of the collecting optics and is given by /NA.) Finally, unlike modalities that require fluorescent agents or dyes, deep UV imaging is label-free and quantitative, thus obviating the need for chemical staining [ 28 , 29 ]. This significantly simplifies laboratory procedures and reduces testing/operational costs and variability.…”