Response to Comment on "Feasibility of Raman spectroscopy as a potential in vivo tool to screen for pre-diabetes and diabetes".

This letter aims to reply to Bratchenko and Bratchenko's comment on our paper "Feasibility of Raman spectroscopy as a potential in vivo tool to screen for pre-diabetes and diabetes." Our paper analyzed the feasibility of using in vivo Raman measurements combined with machine learning techniques to screen diabetic and prediabetic patients. We argued that this approach yields high overall accuracy (94.3%) while retaining a good capacity to distinguish between diabetic (area under the receiver-operating curve [AUC] = 0.86) and control classes (AUC = 0.97) and a moderate performance for the prediabetic class (AUC = 0.76). Bratchenko and Bratchenko's comment focuses on the possible overestimation of the proposed classification models and the absence of information on the age of participants. In this reply, we address their main concerns regarding our previous manuscript.

Feasibility of Raman spectroscopy as a potential in vivo tool to screen for pre-diabetes and diabetes.

In this article, we investigated the feasibility of using Raman spectroscopy and multivariate analysis method to noninvasively screen for prediabetes and diabetes in vivo. Raman measurements were performed on the skin from 56 patients with diabetes, 19 prediabetic patients and 32 healthy volunteers. These spectra were collected along with reference values provided by the standard glycated hemoglobin (HbA1c) assay. A multiclass principal component analysis and support vector machine (PCA-SVM) model was created from the labeled Raman spectra and was validated through a two-layer cross-validation scheme. Classification accuracy of the model was 94.3% with an area under the receiver operating characteristic curve AUC of 0.76 (0.65-0.84) for the prediabetic group, 0.86 (0.71-0.93) for the diabetic group and 0.97(0.93-0.99) for the control group. Our results suggest the feasibility of using Raman spectroscopy for the classification of prediabetes and diabetes in vivo.

Use of Raman spectroscopy to screen diabetes mellitus with machine learning tools: reply to comment.

We show the spectra of advanced glycation products in response to recent comments made by Bratchenko et al. Our results suggest that information retrieved by Raman spectroscopy is relevant to screening diabetic patients, however, the comparison carried out in our paper, between ANN and SVM, was not fair, because of the erroneous PCA selection procedure and different sources of variation present in the analysis.

Use of Raman spectroscopy to screen diabetes mellitus with machine learning tools.

Type 2 diabetes mellitus (DM2) is one of the most widely prevalent diseases worldwide and is currently screened by invasive techniques based on enzymatic assays that measure plasma glucose concentration in a laboratory setting. A promising plan of action for screening DM2 is to identify molecular signatures in a non-invasive fashion. This work describes the application of portable Raman spectroscopy coupled with several supervised machine-learning techniques, to discern between diabetic patients and healthy controls (Ctrl), with a high degree of accuracy. Using artificial neural networks (ANN), we accurately discriminated between DM2 and Ctrl groups with 88.9-90.9% accuracy, depending on the sampling site. In order to compare the ANN performance to more traditional methods used in spectroscopy, principal component analysis (PCA) was carried out. A subset of features from PCA was used to generate a support vector machine (SVM) model, albeit with decreased accuracy (76.0-82.5%). The 10-fold cross-validation model was performed to validate both classifiers. This technique is relatively low-cost, harmless, simple and comfortable for the patient, yielding rapid diagnosis. Furthermore, the performance of the ANN-based method was better than the typical performance of the invasive measurement of capillary blood glucose. These characteristics make our method a promising screening tool for identifying DM2 in a non-invasive and automated fashion.