Introduction/Background Ovarian cancer is the sixth most common cancer in women in the UK and a leading cause of death from gynaecological cancers, with effective screening and early detection remaining elusive. Vibrational spectroscopy is a rapid, low-cost, non-invasive analytical tool with the potential to classify benign and malignant pathologies according to their biochemical composition. We report on findings of infrared (IR) spectroscopy of ascitic fluid from women with or without ovarian cancer which, to our knowledge, has been conducted for the first time.

Methodology Attenuated total reflection Fourier-transform IR (ATR-FTIR) spectroscopy was conducted on ascites collected from 20 patients with benign gynaecological pathology, 5 patients with borderline ovarian tumours and 20 ovarian cancer patients. Ten spectra were acquired from each sample. The obtained spectra were pre-processed by cutting the fingerprint region and applying the Savitzky-Golay 2nd derivative and vector normalisation. Classification was conducted by partial least squares discriminant analysis (PLS-DA) with 9 latent variables (94% explained variance) and Support Vector Machines (SVM) algorithm.

Results Fitting accuracy for discrimination of the 3 classes with PLS-DA was 88% (training set with 70% of samples). When the model was fed with external test samples (30% of samples) the prediction accuracy was 85%. Sensitivity was 0.91, 0.80 and 0.79 for benign, borderline and cancer categories, while specificity was 0.85, 0.96 and 0.93, respectively. However, with SVM, accuracies for fitting and external test samples discrimination were 100%. Additionally, obtained sensitivities and specificities were 100% for the 3 different classes, confirming a better diagnosis response using SVM.

Conclusion ATR-FTIR spectroscopy demonstrates excellent potential for classifying ovarian cancer and non-cancerous gynaecological conditions in ascitic fluid.

Disclosure Nothing to disclose.