WS09‐04Ultrasound diagnosis of adnexal masses: computer modelling
Juan Luis Alcázar · Ultrasound in Obstetrics and Gynecology · 2000
The differential diagnosis of adnexal masses still represents a challenge. The main goal is to determine whether a given mass is malignant or benign, in order to establish the best therapautical approach. When a clinician face to a woman diagnosed as having an adnexal tumor several parameters may be used to determine the nature of the mass, such as patient's age, menopausal status, clinical complaints, physical examination, tumor's sonographic and Doppler features and serum tumor markers. When these data are collected may be difficult to interpret them all together and the clinician has to rely in one or two of them to predict the diagnosis and to decide the management strategy, guided many times by own experience or strict protocols. A possible solution to this question may be the use of computer modelling. This means the use of statistical or mathematical models developed, using computer aid, to predict adnexal malignancy. Currently, there are two methods that may be used: Multiple Logistic Regression Analysis (MLRA) and Artificial Neural Networks (ANN). Both techniques allow to calculate the probability of a particular event has occurred, in this case a mass to be malignant or benign. By computer modelling, actual independent predictors parameters are identified and may be used combined and simultaneously to predict malignancy. Each method has advantages and disadvantages, being MLRA the statistical technique of choice and ANN a potential newer alternative to MLRA. Several studies have been published using one or both technique to predict adnexal malignancy. The results of these studies are encouraging showing a high diagnostic performance. All studies comparing MLRA to ANN showed the latter seems to be superior. However, it should be taken into account that most of these models have not tested prospectively.