Method for data analysis in different institutions: Example of image guidance of prostate cancer patients - Abstract

Multi-institutional collaborations allow for more information to be analyzed but the data from different sources may vary in the subgroup sizes and/or conditions of measuring.

Rigorous statistical analysis is required for pooling the data in a larger set. Careful comparison of all the components of the data acquisition is indispensable: identical conditions allow for enlargement of the database with improved statistical analysis, clearly defined differences provide opportunity for establishing a better practice. The optimal sequence of required normality, asymptotic normality, and independence tests is proposed. An example of analysis of six subgroups of position corrections in three directions obtained during image guidance procedures for 216 prostate cancer patients from two institutions is presented.

Written by:
Piotrowski T, Rodrigues G, Bajon T, Yartsev S.   Are you the author?
Department of Medical Physics, Greater Poland Cancer Centre, Poznan, Poland; Department of Electroradiology, University of Medical Sciences, Poznan, Poland.

Reference: Phys Med. 2013 May 28. pii: S1120-1797(13)00057-4.
doi: 10.1016/j.ejmp.2013.05.001

PubMed Abstract
PMID: 23721940 Prostate Cancer Section