Test oracles determine whether a particular execution of a Program under Test (PUT) is correct or not. In automated testing scenarios, test oracles represent an essential mechanism for testing productivity and effectiveness. However, depending on the PUT's output domain, proper oracle automation turns into a challenge, making these systems hard to test. In the literature, these systems are widely known as complex-output systems. Systems with graphical/audio outputs, three-dimensional objects, complex graphical user interfaces, and some Web applications are common examples of contemporary complex-output systems. A large portion of complex-output systems addresses to support the medical area, and, generally, automated testing strategies to them are severely limited by lack of test oracles. The lack of automated test oracles leads to the application of manual testing activities, in which the tester plays the role of the test oracle, making the test informal, ad-hoc, and nonproductive. A possible contribution to alleviate testers' efforts is the implementation of test oracles based on features extracted from the PUT's outputs. Results of previous studies show that this technique contributes to increase testing productivity, mitigating and complementing manual efforts. This project aims at setting and evaluating automated test oracles for systems whose outputs consist of synthetic three-dimensional medical images. Nowadays, synthetic three-dimensional objects are powerful resources to support the medical diagnosis in different fields of medicine. Our strategy is to explore the framework O-FIm/CO (Oracle for Images and Complex Outputs) that uses concepts of Content-Based Image Retrieval concepts (CBIR) as an effective strategy to automate test oracles. Besides some extensions in the framework, we are intended to develop a set of feature extractors of three-dimensional medical images of blood vessels. Initially, the oracle automation processes will be applied in three-dimensional synthetic images of vascular networks. Through experiments with three-dimensional images produced by real-world systems, we expected to measure the strengths and drawbacks of the technique. It is expected that this project contributes reducing time and efforts demanded by manual approaches (human oracle) during the quality assessment of generating three dimensional medical imaging systems.
News published in Agência FAPESP Newsletter about the scholarship: