Transductive Machine Learning for Reliable Medical Diagnostics

In the past decades Machine Learning tools have been successfully used in several medical diagnostic problems. While they often significantly outperform expert physicians (in terms of diagnostic accuracy, sensitivity, and specificity), they are mostly not being used in practice. One reason for this is that it is difficult to obtain an unbiased estimation of diagnoses reliability. We discuss how reliability of diagnoses is assessed in medical decision making and propose a general framework for reliability estimation in Machine Learning, based on transductive inference. We compare our approach with a usual (Machine Learning) probabilistic approach as well as with classical stepwise diagnostic process where reliability of diagnose is presented as its posttest probability. The proposed transductive approach is evaluated on several medical data sets from the UCI (University of California, Irvine) repository as well as on a practical problem of clinical diagnosis of the coronary artery disease. In all cases significant improvements over existing techniques are achieved.     

Variable presentations of Currarino syndrome in three members of the same family

The article presents an autosomal dominant Currarino syndrome with incomplete penetrance in three out of four members of the same family. The mother had only a bony sacral defect and no other signs. In the older daughter, the syndrome was completely developed with presacral cystic teratoma, a sacral defect and abdominal discomfort. The younger daughter had no clinical or imaging features of the disease. The only son harboured presacral meningocele, urinary stenosis and a sacral defect. The daughter and son with developed variants of the syndrome were successfully operated on and are now symptom free.     

Divergent Physical Activity and Novel Alternative Responses to High Fat Feeding in Polygenic Fat and Lean Mice

We determined whether altered physical activity levels might underlie the contrasting adiposity of a divergently selected polygenic murine model of metabolic syndrome (Fat; F) and leanness (Lean; L) mice. We measured physical activity with a long term running wheel experiment and performed an additional high fat diet intervention. Further, we measured posture allocation by visual monitoring within the home cage as a non-exercise correlate of ‘normal’ physical activity. Whilst initially similar, running wheel activity of the F line declined with age, while the activity of the L line increased. Food intake was higher in the L line and increased with wheel exposure. Vertical rearing measured by video quantification in the home cage, without the stimulus of a running wheel was also significantly higher in the L line. The two lines developed novel alternate strategies to defend their body weight when exposed to high fat diets with a running wheel. F mice increased their running wheel activity, and despite unaltered food intake, still gained weight. L mice reduced their food intake and maintained activity levels without a significant change in body weight. Phenotypic selection for divergence in body fat content has co-segregated with a genetic predisposition for divergent physical activity levels and different strategies for coping with exposure to high fat diets that will facilitate the discovery of the genes underlying these important obesity related traits. Edited by Stephen Maxson.