Classification of cervical cancer cells using FTIR data

High false-negative rates of the Papanicolauo (so-called 'Pap') smear test and the shortage of colposcopists have led to the desire to find alternative non-expert (automated) approaches for accurately testing cervical smears for signs of cancer. Fourier-Transform Infra-Red (FTIR) spectroscopy has been shown to offer the potential for improving the accuracy (i.e. sensitivity and specificity) of these tests. This paper details the application of the machine learning methodology of Support Vector Machines (SVM) using FTIR data to enhance and improve upon the standard Pap test. A cohort of 53 subjects was used to test the veracity of both the Pap smear results and the FTIR based classifier against the findings of the colposcopists. The Pap test achieved an overall classification of 43 %, whereas our method achieved a rate of 72%