Fast and Noninvasive Diagnosis of Cervical Cancer by Coherent Anti-Stokes Raman Scattering

Abstract

Cervical cancer is the fourth most common cancer in women worldwide, and early detection of its precancerous lesions can decrease mortality. Cytopathology, HPV testing, and histopathology are the most commonly used tools in clinical practice. However, these methods suffer from many limitations such as subjectivity, cost, and time. Therefore, there is an unmet clinical need to develop new noninvasive methods for the early detection of cervical cancer. Here, a novel noninvasive, fast, and label-free approach with high accuracy is presented using liquid-based cytology Pap smears. CARS and SHG/TPF imaging was performed at one wavenumber on the Pap smears from patients with specimens negative for intraepithelial lesions or malignancy (NILM), and low-grade (LSIL) and high-grade (HSIL) squamous intraepithelial lesions. The normal, LSIL, and HSIL cells were selected on the basis of the ratio of the nucleus to the cytoplasm and cell morphology. Raman spectral imaging of single cells from the same smears was also performed to provide integral biochemical information of cells. Deep convolutional neural networks (DCNNs) were trained independently with CARS, SHG/TPF, and Raman images, taking into account both morphotextural and spectral information. DCNNs based on CARS, SHG/TPF, or Raman images have discriminated between normal and cancerous Pap smears with 100% accuracy. These results demonstrate that CARS/SHG/TPF microscopy has a prospective use as a label-free imaging technique for the fast screening of a large number of cells in cytopathological samples. C cancer is the fourth most common cancer in women worldwide, with 266 000 deaths in 2012. The peak rate of cervical cancer cases is found in middle-aged women between 35 and 44 years of age. Early detection of the precancer stage is necessary to reduce the mortality associated with cervical cancer significantly. Mostly, cervical cancer develops in the basal layer of cells lining the cervix and progresses gradually, revealing several dysplastic changes that can lead to invasive cancer. The Papanicolaou (Pap) test or socalled Pap smear is the most common screening method for identifying an abnormality in the cervix. Abnormal Pap smears are followed by colposcopy, biopsy, and histopathological investigation to confirm the diagnosis. The Pap test is a noninvasive method, extensively accepted, and its results include the following categories according to the Bethesda system: negatives for intraepithelial lesions or malignancy (NILM), atypical squamous cells of undetermined significance (ASCUS), low-grade squamous intraepithelial lesions (LSIL), and high-grade squamous intraepithelial lesions (HSIL). This cytology method depends on the visual evaluation of individual cell morphology and detects cancer and precancer cells, making it highly subjective with a large variation in the sensitivity (50−96%). Persistent infection with human papillomavirus (HPV) is the major risk factor for the development of cervical cancer. An HPV-DNA test is used to screen for HPV-DNA fragments and determine whether the patient is infected with one of the HPV high-risk types. Although the HPV-DNA test has a higher sensitivity (∼95%), it suffers from low specificity (∼84%) and is also expensive. Therefore, there is an unmet clinical need to develop a new noninvasive method for cervical cancer screening. Raman spectroscopic methods including conventional Raman spectroscopy, coherent anti-Stokes Raman scattering (CARS), and stimulated Raman scattering (SRS) are emerging biophotonic tools in the bioanalysis and imaging of biomaterials such as body fluids, cells, and tissues. These methods are nondestructive and label-free approaches with a spectroscopic capability to probe different biomolecules and monitor their changes with the progression of cancer or neurodegenerative diseases. Conventional Raman spectroscopy has been applied extensively to cervical cancer specimens, especially tissues and cell lines; however, limited applications Received: July 26, 2019 Accepted: September 4, 2019 Published: September 4, 2019 Article pubs.acs.org/ac Cite This: Anal. Chem. XXXX, XXX, XXX−XXX © XXXX American Chemical Society A DOI: 10.1021/acs.analchem.9b03395 Anal. Chem. XXXX, XXX, XXX−XXX D ow nl oa de d vi a Sa m ir E lM as ht ol y on S ep te m be r 17 , 2 01 9 at 1 8: 49 :5 2 (U T C ). Se e ht tp s: //p ub s. ac s. or g/ sh ar in gg ui de lin es f or o pt io ns o n ho w to le gi tim at el y sh ar e pu bl is he d ar tic le s. on cervix cytology were reported. For instance, Raman spectroscopy was used to differentiate between HPV-positive and HPV-negative Pap smears with high accuracy. On the other hand, the classification of normal and cancerous Pap smears was achieved with lower accuracy (80%). This is most likely because Raman spectra were acquired using cell pellets instead of individual cells, leading to sample heterogeneity. Recent studies have shown that few Raman spectra from the nuclei of single cells can discriminate between normal and abnormal Pap smears with high accuracy using principal component analysis−linear discriminant analysis (PCA−LDA) and partial least-squares discriminant analysis (PLS−DA). However, these few spectra neither represent the integral biochemical composition of cells nor allow access to the morphological or textural features of cells because of the lack of a Raman imaging modality. In addition, the use of conventional chemometric approaches such as PCA-LDA and PLS-DA does not consider the morphological or textural features of cells. Coherent Raman techniques such as CARS and SRS imaging have been used recently in many biomedical applications. These methods are much faster than conventional Raman spectroscopy and can be performed at a speed of up to a video rate, allowing fast diagnosis. We have recently used a combination of CARS and second harmonic generation (SHG) imaging as a fast tool for prescreening urothelial cells in urine sediments. Afterward, Raman spectral imaging of the selected cells was performed to differentiate between noncancerous and cancerous urothelial cells using deep convolutional neural networks (DCNNs). The results have shown the advantage of DCNN for Raman data compared to the conventional chemometric methods. This is because the DCNN classifications were based on not only the spectral information but also the morphological features of the cell. Thus, a combination of CARS imaging, much faster than Raman imaging, and DCNNs would be a perfect candidate for a fast label-free imaging approach for the diagnosis of cancer using cytopathological samples. Such combination was used for the detection of lung and head and neck carcinoma using only tissue sections that were collected invasively from patients but not using liquid-based cytology. Here, we report for the first time fast CARS, SHG, and twophoton excited autofluorescence (TPF) imaging of liquidbased cytology including normal, LSIL, and HSIL Pap smears that were collected noninvasively from patients. Cells were screened within a very short time. Raman spectral imaging of single cells from the same Pap smears was also acquired, and the results provided not only integral biochemical information of single cells but also morphological features of cells. DCNNs were used to discriminate with very high accuracy among normal, LSIL, and HSIL cells in Pap smears based on morphological features extracted from CARS, SHG/TPF, and Raman microscopic images. Finally, the results demonstrate that CARS and SHG/TPF imaging has the potential to be a fast and noninvasive method for the diagnosis of cervical cancer with high accuracy. ■ EXPERIMENTAL SECTION Pap Smears. Pap smears were collected from 10 healthy women, 10 patients diagnosed with LSIL, and 10 patients diagnosed with HSIL by ZYDOLAB (Institute for Clinical Cytology and Immune Cytochemistry; Dortmund, Germany). Institutional review board approval (IRB 16-5654) and written informed consent from all patients have been obtained. Pap smears were provided using liquid-based cytology. For this method, samples from the cervix uteri were collected using a cervical sampler and deposited into preservative liquid. This technique allows accurate results because of the removal of mucus, blood, and other elements. The preparation of Pap smears for Raman and CARS/SHG/TPF measurements is shown in the Supporting Information (SI). CARS and SHG Microscopic Imaging. CARS and SHG images were acquired using a commercial setup (TCS SP5 II CARS; Leica Microsystems, Heidelberg, Germany) consisting of a picosecond pulsed laser setup (APE picoEmerald, Berlin, Germany). It generates and synchronizes two collinearly aligned beams to a confocal inverted microscope as reported previously. The pump and Stokes wavelengths were adjusted to 810.5 and 1064 nm, respectively. Laser beams are focused on the microscope using a water-immersion objective (HCX IRAPO L, 25X/0.95 W, Leica Microsystems). CARS and SHG/TPF images of Pap smears from 30 patients were acquired simultaneously at 2935 cm−1 with a pixel dwell time of 180 μs and a pixel resolution of 250 nm. CARS imaging is displayed in one channel, while SHG and TPF imaging is shown in another channel. Raman Spectral Imaging. Raman spectral imaging of cells in Pap smears was measured using an alpha300 RA confocal Raman microscope (WITec, Ulm, Germany) as described previously. A frequency-doubled Nd:YAG laser operating at 532 nm (Crystal Laser, Reno, NV, USA) is the Raman excitation source. The laser beam is coupled into a microscope using a single-mode optical fiber, and it is collimated and then focused on the sample through a Nikon NIR APO (60x/1.00 NA) water-immersion objective. Raman measurements were performed using a raster scanning laser beam over cervical cells in order to measure full Raman spectra (0.5 s per pixel) with a pixel resolution of 500 nm. The Raman measured 82 normal cells, 32 LSIL cells, and 41 HSIL cells that were selected fr