Compare the receiver operating characteristic (ROC) and linear discriminant analysis (LDA) for acromegaly detection by three-dimensional facial measurements

Excessive growth hormone secretion will result in acromegaly affect metabolic function. Patients with acromegaly is 2–4 times greater risk of death than the normal. Early diagnosis is the key follow-up treatment of acromegaly. The clinical diagnosis is based on typical acromegaly the face and body features, endocrine and radiological. However, acromegaly diagnosis is still quite deferred. Typical acromegaly, with the symptoms and appearance, the physician can diagnose. Obvious early symptoms, diagnosis is not easy. As imaging technology advances, one after another to explore the diagnosis of acromegaly, however, did not the size of the stereoscopic 3D image. The aim of this study is to compare the compare the Receiver operating characteristic (ROC) and discriminant analysis for acromegaly detection by three dimensional facial measurements. To explore the difference of detection rate between the two analysis methods. The result shows that the accuracies of three categories from the univariate discriminant analysis, the lateral angles displayed the highest accuracy between all three categories in the female but the lowest rate for the ROC analysis. However, the lateral angles displayed the lowest accuracy between all three categories in the male and the lowest rate for the ROC analysis. The lateral angles, calculated from the two prominent variables, made a larger difference than the other two categories. From the result, it shows that the accuracy difference analysis between the two analysis methods in both genders. The difference could come from the different operation of the analysis methods. It could use the different analysis method to analyze the different facial dimensions for the acromegaly detection in the future and increase the accuracy for disease detection.