Characterisation of ductal carcinoma in situ (DCIS) using mass spectrometry imaging towards near realtime margin assessment.

3140 Background: Imprecision in breast-conserving surgery leads to high national average high rates of reoperative intervention. In line with updated margin guidelines, accurate differentiation between non-invasive and invasive breast cancer is essential. This study aimed to assess whether mass spectrometry can distinguish between normal breast tissue, benign, non-invasive and invasive disease towards the development of an intraoperative margin assessment tool. Methods: Breast tissue samples were collected from patients undergoing mastectomy. Samples were flash-frozen, sectioned, and analysed using a Xevo G2-XS QTof mass spectrometer (Waters Corp.). Selected sections were ionised using a pulsed optical parametric oscillator laser (OpoletteTM 2731/3034, OPOTEK) which operated at 2940 nm wavelength and 20 Hz repetition rate. The laser focused on the tissue through a 20 mm focal distance convex lens generating aerosol which was aspirated into the spectrometer. The data was combined using spatial distribution and chemical information from characteristic ions to generate 2D chemical images and labelled using consecutive H&E-stained sections annotated by a Consultant Histopathologist for ground truth cross-validation. Results: Over 1 million mass spectra were collected from imaging 52 breast tissue sections. This includes 720 mass spectra from 31 DCIS breast tissue sections, compared to 6 spectra from 2 DCIS breast tissue samples in previous work. A pixel size of 50 μm and scan rate was 250 μm/s was utilised. An ex-vivo classification model was built using n=6,796 and achieved >99% sensitivity for tumour detection (DCIS and IBC) and 100% specificity for identifying normal tissue. Principal Component Analysis demonstrated accurate separation of IBC, DCIS, benign breast disease, and normal breast tissue. Six possible metabolites were identified following Recursive Feature Elimination (RFE) was used to identify the most significant features which differentiate the tissue types, these were annotated using the Lipid Maps database (http://www.lipidmaps.org/) (Table 1). Cancerous tissue showed higher levels of structural lipids (600-900 Da), while normal/benign breast tissue had higher levels of small metabolites (50-300 Da) and fatty acids (200-400 Da). Conclusions: Mass spectrometry imaging enables accurate differentiation of IBC, DCIS, benign breast disease, and normal breast tissue. Biological features identified from the most significant RFE selected features. m/z value Annotation Delta Theoretical m/z Ion Class 255.2324 Palmitic acid 0.0006 255.2330 M-H Fatty acid 297.2751 FA 19:0 0.0037 297.2799 M-H Fatty acid 307.2019 FA 16:0 0.0026 307.2046 M+Cl Fatty acid 766.5392 PE 38:4 0 766.5392 M-H PE 843.5053 PI 35:4 0.0025 843.5029 M-H PI 891.7444 TG 52:3 0.0003 891.7447 M+Cl TG