Abstrakti
Electronic tongues (e-tongues) have been broadly employed in monitoring the quality of food, beverage, cosmetics, and pharmaceutical products, and in diagnosis of diseases, as the e-tongues can discriminate samples of high complexity, reduce interference of the matrix, offer rapid response. Compared to other analytical approaches using expensive and complex instrumentation as well as required sample preparation, the e-tongue is non-destructive, miniaturizable and on-site method with little or no preparation of samples. Even though e-tongues are successfully commercialized, their application in cancer diagnosis from urine samples is underestimated. In this review, we would like to highlight the various analytical techniques such as Raman spectroscopy, infrared spectroscopy, fluorescence spectroscopy, and electrochemical methods (potentiometry and voltammetry) used as e-tongues for urine analysis towards non-invasive cancer diagnosis. Besides, different machine learning approaches, for instance, supervised and unsupervised learning algorithms are introduced to analyze extracted chemical data. Finally, capabilities of e-tongues in distinguishing between patients diagnosed with cancer and healthy controls are highlighted.
| Alkuperäiskieli | Englanti |
|---|---|
| Artikkeli | 114810 |
| Julkaisu | Biosensors and Bioelectronics |
| Vuosikerta | 219 |
| DOI - pysyväislinkit | |
| Tila | Julkaistu - 1 tammik. 2023 |
| OKM-julkaisutyyppi | A2 Katsausartikkeli tiedejulkaisuussa (artikkeli) |
Rahoitus
TPH would like to acknowledge the Academy of Finland (Grant No. 323240 ). PY thanks the financial support from the EDUFI Fellowship. Supervised learning offers a powerful approach by presenting the algorithm with example inputs and their desired classes (outputs) to build a general model that maps inputs to classes (Moncayo et al., 2015). Supervised learning methodology could also be split up into two subgroups: classification and regression. Linear regression analysis is usually employed in making predictions by way of finding mathematical relationships between quantitative variables. Linear regression is considered as one of the most primitive ML methodologies which are still broadly utilized (Talabis et al., 2015). A wide variety of linear regression models include simple linear regression (SLR), multiple linear regression (MLR), principal component regression (PCR), partial least squares regression (PLSR), and support vector machine (SVM) (Yan and Ramasamy, 2019). Linear regression has been utilized in many applications, for examples, determination of the total polyphenols content in green tea (Chen et al., 2008), glucose, fructose, and sucrose in bayberry juice (Xie et al., 2009), prediction of microbial numbers on Atlantic salmon (Tito et al., 2012) and quantification of urea, creatinine, glucose, protein, and ketone in urine (Pezzaniti et al., 2001). Alternatively, nonlinear regression is a type of regression analysis in which experimental data are represented by a nonlinear function which is a combination of one or more independent variables and model parameters (Giddings and Ratkowsky, 1991). For instance, one of the nonlinear regression algorithms – neural network has the ability of learning complex nonlinear relationships from a training dataset. This ability makes it appropriate for pattern recognition problems involving the uncovering of convoluted tendencies in high-dimensional datasets (Guenther, 2001). Classification is established on a collection of formerly labeled inputs to generate a discrimination model capable of distinguishing between two or multiple classes. Classification is founded on the similarities and differences among individuals within the same and different classes, respectively, by means of discriminating a set of classes based on their measured features or variables. One of the main concerns encountered in classification is feature engineering, which deals with figuring out the most significant features. Over the past few decades, several classification algorithms have been developed, such as linear discriminant analysis (LDA), logistic regression (LR), partial least squares discriminant analysis (PLS-DA), naive Bayes, k-nearest neighbors (k-NN), artificial neural networks (ANNs), SVM, and decision tree algorithms (Subasi, 2020).TPH would like to acknowledge the Academy of Finland (Grant No. 323240). PY thanks the financial support from the EDUFI Fellowship.
Sormenjälki
Sukella tutkimusaiheisiin 'Analysis of urine using electronic tongue towards non-invasive cancer diagnosis'. Ne muodostavat yhdessä ainutlaatuisen sormenjäljen.Viittausmuodot
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver