The goal of the cystectomy study is to use autofluorescence spectroscopy to diagnose normal mucosa from dysplasia, carcinoma-in-situ, invasive cancer, and papillary tumors in bladder. Principal component analysis (PCA) followed by least-squares regression (LR) was applied to the data set. For 27 patients taken with 400 nm excitation, a calibration set and a validation set were randomly formed with 13 and 14 patients respectively. With an imposed limitation of three categories of normal mucosa, CIS, and invasive cancer, an attempt with PCA and LR on the validation set produced 60% detection of invasive cancer, 0.8% of CIS, and 91.8% of normal mucosa. Changing the categories to normal vs. disease, with the disease category including CIS and invasive cancer, detected 86.7% of disease and 55.8% of normal mucosa. However, hemorrhage, inflammation, and other factors were not accounted for, yet later investigation showed hemorrhage decreased signal intensity by an average of around 25% and severe inflammation by an average of around 50%. Accounting for these effects is now ongoing. For 29 patients with 370 nm excitation, a calibration set and a validation set were randomly formed with 14 and 15 patients respectively, but these are also in the process of being redone to account for the effects of hemorrhage, inflammation, and edema.
Showing the most recent 10 out of 178 publications
