Serial sections from paraffin-embedded pCA tissue were collected. One was stained with hematoxylin and eosin and Gleason scored; FTIR spectra were collected from malignant locations using a second unstained section. FTIR spectra, representing different Gleason grades, were used to construct a diagnostic classifier for pCA using linear discriminant analysis (LDA). This model was blind tested using 383 IR spectra from 36 biopsies.
Using a three-band Gleason criteria, we obtained sensitivity of ≥70%for the FTIR-LDA model to predict Gleason <7, = 7, and >7, with specificities of ≥81%. Using a threshold of Gleason/FTIR-LDA score of ≥8, we obtained a sensitivity and specificity of 71%and 67%, respectively, for the correlation with metastatic tumours using the FTIR-LDA system and 85%and 63%, respectively, for the correlation of metastatic tumours using the Gleason system.
There is a correlation between tissue architecture using Gleason score with tissue biochemistry using FTIR-LDA. Both systems are similar in their performance in predicting metastatic behaviour in tumours from individual patients.