Automated learning of patient image retrieval knowledge: neural networks versus inductive decision trees

Retrieving a patient's prior examination images that are relevant to the cu rrent ones is a critical component in radiologists' primary examination rea ding services. The important role of such image retrieval support will be g reatly accentuated in digital radiology practice. Radiologists' knowledge o f patient prior image retrievals is rooted in their interpretation and appl ication of the pertinent underlying medical/radiological knowledge as well as in their clinical training and experiences. At the same time, this knowl edge may vary with individual practice preferences and styles, and may dyna mically evolve over time. The complexity and dynamics suggest that patient image retrievals are a promising area for artificial intelligence-based aut omated learning techniques. Automated learning of patient image retrieval k nowledge can provide continuous knowledge repository update support in an i mage retrieval knowledge-based system. However, the implementation of the l earning techniques needs to address several challenges that include missing and noisy data, as well as multiple decision outcomes. Two techniques base d on salient automated learning paradigms, neural network and symbolic lear ning, are investigated. Specifically, we describe the design or extension o f each learning technique to address the unique characteristics of patient image retrieval knowledge and compare the resulting learning performances. The results show that the knowledge derived from the automated learning met hods can achieve effective image retrievals that are comparable to those ba sed on a knowledge-engineer-driven approach. (C) 2000 Elsevier Science B.V. All rights reserved.