SVM for FT‐MIR prostate cancer classification: An alternative to the traditional methods

Abstract

In this paper, principal component analysis (PCA), successive projections algorithm (SPA), and genetic algorithm (GA) followed by support vector machines (SVM), combined with Fourier-transform mid-infrared (FT-MIR) spectroscopy were presented as complementary or alternatives tools to the traditional methods for prostate cancer screening and classification. These approaches were applied to analyze tissue samples, and their performances were compared within dependent SVM models and with traditional methods of diagnosis, according to class separation interpretability, time consumption, and figures of merit. The results showed that variable reduction and selection methods followed by SVM can reduce drawbacks of independent SVM analysis. The potential biomarkers indicated by PCA-SVM, SPA-SVM, and GA-SVM were amide I, II, and III; as well as protein regions (1400-1585 cm −1 ), followed by DNA/RNA (O-P-O symmetric stretch) (1080 cm −1 ) and DNA (O-P-O asymmetric stretch) (1230 cm −1 ) regions. GA-SVM was the best classification approach, with higher sensitivity (100%) and specificity (80%), particularly in early stages, being better than traditional methods of diagnosis. KEYWORDSFT-MIR, prostate cancer, SVM, tissue 1 | INTRODUCTION Currently, prostate cancer recognition follows some phases. First of all, the proctologist evaluates the prostate by Digital Rectal Examination (DRE), which allows palpation of only 40% to 50% of the prostate, and it is mainly affected by intraobserver and inter-observer variability, resulting in DRE sensitivity and specificity of about 21% to 37% and 71% to 91%, respectively. 1-5 Concomitantly, a measurement of serum prostatic specific antigen levels is performed which have about 21% and 64% sensitivity and specificity in earliest stages, respectively, and about 32% and 93% sensitivity and specificity in high grade prostate cancers, respectively. When combined with DRE, sensitivity and specificity increase from 51% to 68% and from 92% to 94%, respectively 1-8 ; however, this measurement can be affected by other factors beyond prostate cancer, such as ejaculation, bacterial prostatitis, biopsy, and acute urinary retention, which may elevate prostatic specific antigen levels. 3,[5][6][7][8] According to the results of these exams, a biopsy coupled with an anatomopathological examination can be indicated in order to identify cancer stage. Trans-rectal ultrasound (TRUS)-guided biopsy is currently the gold standard, characterized by 39% to 52% sensitivity and 81% to 82% specificity. 9 The anatomopathological examination by Gleason score system classifies biopsy samples according to tumor aggressiveness and provides a prognostic idea. It defines Gleason 1 (best differentiation and most favorable prognosis) to Gleason 5 (least differentiation and poor prognosis). [10][11][12] The sum of the primary (most architectural atypia) and secondary (least architectural atypia) patterns leads to the Gleason score (eg, Gleason 3 + Gleason 4 is equivalent to a Gleason score 3 + 4 = 7). 12 In 2014, a new Gleaso...

Multivariate classification techniques and mass spectrometry as a tool in the screening of patients with fibromyalgia

“…The LDA supervised models are recurrent in several clinical studies of this nature, such as in cancer 36 , 37 and viral diseases 38 – 40 , because they work with processes with less complex characteristics. It is also common in the literature to find similar studies that expand the classifications to more robust models such as QDA 41 , 42 and SVM 43 , other researches seek to further expand the treatment of spectral samples, submitting them to these three types of techniques (LDA, QDA and SVM) and comparing the better accuracy results given the complexities of multivariate data 2 , 24 . This last approach is worked on and observed in this article.…”

Tutorial: multivariate classification for vibrational spectroscopy in biological samples

“…The variables selected by these techniques can be used as input for the classification methods described previously, which is important since these techniques reduce the data size and collinearity, hence, improving the model accuracy and analysis time. Adaptations of PCA, PLS, SPA and GA (as feature extraction/selection techniques) to LDA, QDA, SVM, KNN and ANN (as classifiers) are well known [114][115][116] .…”

Data‐Driven Approaches for Diagnosis of Incipient Faults in Cutting Arms of the Roadheader

“…e results showed that the new method greatly compensated for the defects of the traditional methods [21]. Jung et al used the PCA-assisted SVM method for reliable channel reservoir characterization.…”