Development and characterization of a dual-fiber Raman probe for accurate and rapid endometrial carcinoma detection

We report on the development and characterization of a dual-fiber optic Raman probe for enhancing real-time endometrial carcinoma biopsy. The fiber optic probe was designed with two miniaturized coating fibers with an outer diameter (OD) of 3.1 mm. We demonstrate that with the use of the Raman probe associated with the background subtraction algorithms, high-quality tissue Raman spectra covering the fingerprint (FP) (500-1800cm^-1) can be acquired from different tissue types (e.g., muscle, fat, parathyroid gland, and thyroid gland). Using an anti-reflection coating filter and background subtraction algorithms, our probe has better signal-to-noise (SNR) than a conventional coating probe with a comparable OD. Meanwhile, our probe has less OD and 58% SNR compared with the handheld probe (OD = 10 mm). We developed a Raman fiber probe with a small outer diameter and high SNR (>90 in fat) simultaneously and used the small-size probe for accurate and rapid endometrial carcinoma detection. Specifically, a total of 220 fingerprint Raman spectra was obtained from 22 endometrial carcinoma (EC) patients (benign: 10, malignant: 12) and divided into two batches (110 spectra from 11 patients for each) from different patients for robustness test. The mean area under the receiver operating characteristic curves (AUC) for differentiating benign and malignant EC reached 0.946, with a mean accuracy of 93.1% and a mean time expense of 5 seconds by the ResNet model. Furthermore, the specific Raman features corresponding to biomolecules (e.g., amide I II, lipid, and collagen) were found to contribute to the accurate classification.