Objectives Classification of malignant tissue is performed by histopathology, results of which may be obtained within weeks. Faster histopathology analysis is performed using frozen section (FS) analysis, results of which are available within less than an hour. However, the accuracy of the FS test is lower compared to histopathology. We are proposing a real time method based on IR spectroscopy of liquid biopsy from fresh samples of gynecologic cancer tissues as an alternative to the conventional methods.

Methods 27 biopsies (ovarian and uterine, 17 classified as benign and 10 malignant) were extracted from suspected tumor sites during gynecologic surgical procedures and sent for both pathological and FTIR analyses. Tissue samples were lightly pressed against the surface of an ATR crystal, leaving on its’ surface impression smears. Mid-IR absorption spectra were obtained within 15 minutes of excision. Histopathological results of these very same samples were used to develop discriminant models from the absorbance data of the measured smears using machine learning techniques (PCA-LDA and SVM).

Results IR absorbance spectra of malignant smears were consistently higher from spectra of benign smears in the 850–1450 cm-1 range and they were consistently lower in the 3200–3600cm-1 range. The PCA-LDA discrimination model correctly classified the samples with a sensitivity and specificity of 100%, and the SVM showed a training accuracy of 100% and a cross validation accuracy of 91.3%.

Conclusions These preliminary results suggest that ATR-FTIR spectroscopy of tissue smears may have an important role in the development of next-generation techniques for intra-operative tumor classification.