Vesicoureteral reflux detection with reliable probabilistic outputs

摘要: Vesicoureteral Reflux (VUR) is a pediatric disorder in which urine flows backwards from the bladder to upper urinary tract. Its detection of great importance as it increases risk Urinary Tract Infection, can then lead kidney infection since bacteria may have direct access kidneys. Unfortunately VUR requires rather painful medical examination, called voiding cysteourethrogram (VCUG), that exposes child radiation. In an effort avoid exposure radiation required by VCUG some recent studies examined use machine learning techniques for based on data be obtained without exposing This work takes one step further proposing approach provides lower and bounds conditional probability given having VUR. The important property these they are guaranteed (up statistical fluctuations) contain well-calibrated probabilities with only requirement observations independent identically distributed (i.i.d.). Therefore much more informative reliable than plain yes/no answers provided other techniques.