Abstract
BACKGROUND: The mechanism of esophageal pain in patients with nutcracker esophagus (NE) and other esophageal motor disorders is not known. Our recent study shows that baseline esophageal mucosal perfusion, measured by laser Doppler perfusion monitoring, is lower in NE patients compared to controls. The goal of our current study was to perform a more detailed analysis of esophageal mucosal blood perfusion (EMBP) waveform of NE patients and controls to determine the optimal EMBP biomarkers that combined with suitable statistical learning models produce robust discrimination between the two groups. METHODS: Laser Doppler recordings of 10 normal subjects (mean age 43 ± 15 years, 8 males) and 10 patients (mean age 47 ± 5.5 years., 8 males) with NE were analyzed. Time and frequency domain features were extracted from the first twenty-minute recordings of the EMBP waveforms, statistically ranked according to four independent evaluation criterions, and analyzed using two statistical learning models, namely, logistic regression (LR) and support vector machines (SVM). KEY RESULTS: The top three ranked predictors between the two groups were the 0.5 and 0.75 perfusion quantile values followed by the surface of the EMBP power spectrum in the frequency domain. ROC curve ranking produced a cross-validated AUC (area under the curve) of 0.93 for SVM and 0.90 for LR. CONCLUSIONS & INFERENCES: We show that as a group NE patients have lower perfusion values compared to controls, however, there is an overlap between the two groups, suggesting that not all NE patients suffer from low mucosal perfusion levels.