Premium
Feasibility of desorption electrospray ionization mass spectrometry for diagnosis of oral tongue squamous cell carcinoma
Author(s) -
D'Hue Cedric,
Moore Michael,
Summerlin DonJohn,
Jarmusch Alan,
Alfaro Clint,
Mantravadi Avinash,
Bewley Arnaud,
Gregory Farwell D.,
Cooks R. Graham
Publication year - 2017
Publication title -
rapid communications in mass spectrometry
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.528
H-Index - 136
eISSN - 1097-0231
pISSN - 0951-4198
DOI - 10.1002/rcm.8019
Subject(s) - tongue , chemistry , epithelium , linear discriminant analysis , pathology , mass spectrometry , principal component analysis , nuclear medicine , chromatography , artificial intelligence , medicine , computer science
Rationale Desorption electrospray ionization mass spectrometry (DESI‐MS) has demonstrated utility in differentiating tumor from adjacent normal tissue in both urologic and neurosurgical specimens. We sought to evaluate if this technique had similar accuracy in differentiating oral tongue squamous cell carcinoma (SCC) from adjacent normal epithelium due to current issues with late diagnosis of SCC in advanced stages. Methods Fresh frozen samples of SCC and adjacent normal tissue were obtained by surgical resection. Resections were analyzed using DESI‐MS sometimes by a blinded technologist. Normative spectra were obtained for separate regions containing SCC or adjacent normal epithelium. Principal Component Analysis and Linear Discriminant Analysis (PCA‐LDA) of spectra were used to predict SCC versus normal tongue epithelium. Predictions were compared with pathology to assess accuracy in differentiating oral SCC from adjacent normal tissue. Results Initial PCA score and loading plots showed clear separation of SCC and normal epithelial tissue using DESI‐MS. PCA‐LDA resulted in accuracy rates of 95% for SCC versus normal and 93% for SCC, adjacent normal and normal. Additional samples were blindly analyzed with PCA‐LDA pixel‐by‐pixel predicted classifications as SCC or normal tongue epithelial tissue and compared against histopathology. The m/z 700–900 prediction model showed a 91% accuracy rate. Conclusions DESI‐MS accurately differentiated oral SCC from adjacent normal epithelium. Classification of all typical tissue types and pixel predictions with additional classifications should increase confidence in the validation model.