Key points are not available for this paper at this time.
In this article a method for schizophrenia disorders detection is presented, which is based on Rorschach Inkblot Test and an eye-tracker system. The method extracts and evaluates the overall time period in defined regions as well as the path an image is scanned through by an individual. It is done by expressing the picture scanning process by Markov chain which is a well know, easy to implement, and mathematical tractable model. Key features used for further processing are vectors of final probabilities and transition matrices. In the next stage the extracted features are classified into positive (schizophrenia disorder) and negative (a healthy individual) classes. For the classification process a non-parametric KNN method was deployed using several strategies and a fine tuning of free parameters. KNN was chosen to eliminate prior model assumptions which are unknown. In our experiments we used a dataset consisting of 44 individuals (22 patients, and 22 healthy individuals). Depending on features and settings the detection accuracy was in the interval of 62% to 75%. Transition matrices achieved better results than the vector of final probabilities either for the best settings or on average.
Kačur et al. (Sun,) studied this question.