Introduction/Background Frozen section (FS) is typically used in cases of suspicious pelvic masses during surgical exploration as a preliminary step to the formal, ‘Gold Standard’ (GS), histopathologic examination, providing needed information for intraoperative decision-making process.

Fourier-transform infrared (FTIR) Spectroscopy is a physical technique, highly sensitive to molecular changes, its conjugation with Attenuated Total Reflection (ATR) concept, enables a rapid sample examination directly from tissues. In the current study, we assessed the potential of ATR-FTIR spectroscopy as a rapid intraoperative clinical decision-making support tool.

Methodology 51 samples were collected intraoperatively during gynecological surgeries, enabling 85 different ATR-FTIR measurements in less than 5 minutes per sample. Machine learning (ML) techniques including- Principal Component Analysis, Linear Discriminant Analysis, Support Vector Machine, and Decision Tree among other models, were implemented to classify samples into malignant and benign groups, as well as determine the sample’s origin and histological subgroups.

T-test was used to establish statistical significance, in reference to final GS pathology.

Results 26 benign & 25 malignant, adnexal, and uterine samples, were the base for 40 and 45 measurements respectively. ML models presented a significant distinction between malignancy and benign samples with 92.9% accuracy (N=51, measurements=85, sensitivity 91.1%, specificity 95%, AUC 0.94). The ability to accurately determine the origin of malignant samples (uterine vs. ovarian) was 91.9% (N=22, measurements=37, sensitivity 83.3%, specificity 96%, AUC 0.95), and in benign samples 95% (N=16, measurements=25, sensitivity 92.3%, specificity 100%, AUC 0.96). Of note, High-grade serous carcinoma was discriminated from Endometrioid carcinoma in both endometrial and ovarian origin with 93.1% accuracy (N=14, measurements=26, sensitivity 86.7%, specificity 100%, AUC 0.98).

Conclusion ATR-FTIR spectroscopy provides an accurate mean for rapid sample classification into malignant vs. benign, as well as determining the histological subtype and tissue origin.

Further developments of these high-potential, novel findings for utilization as complementary intraoperative techniques are currently in process.

Disclosures No conflict of interest.