From on-line prediction wiki

Main: OnlinePredictionOfOvarianCancer

In this paper and technical report, it is described how to apply the Aggregating Algorithm to diagnose ovarian cancer using the level of the standard biomarker CA125 in conjunction with information provided by mass-spectrometry. A new data set collected over a period of 7 years is analyzed. All the samples from healthy and not-healthy individuals are combined into triplets (following Gammerman et. al, 2008), where in each triplet one sample belongs to the diseased patient, and two others belong to healthy patients. The algorithm gives probability prediction for each of the samples in each triplet, taking into account the knowledge that exactly one of the samples in a triplet is to be detected with cancer. The method developed for Probability Forecasting under the Brier loss function is applied to solve the problem.

To estimate classification accuracy the probability predictions are converted into strict predictions. The algorithm makes fewer errors than almost any linear combination (see Devetyarov et. al, 2009) of the CA125 level and one peak's intensity (taken on the log scale). To check the power of the algorithm, the following hypothesis is tested. The hypothesis is that CA125 and the peaks do not contain useful information for the prediction of the disease at a particular time before the diagnosis. The algorithm produces p-values that are better than those produced by the algorithm that has been previously applied to this data set. The conclusion is that the proposed algorithm is more reliable for prediction on new data.

Retrieved from ?n=Main.OnlinePredictionOfOvarianCancer
Page last modified on September 07, 2010, at 06:56 PM