Improving psychiatric screening with artificial intelligence – PORTRAIT

The aim of PORTRAIT project is to develop an adaptive test method that makes it possible to adapt the questions of a test according to the previous answers of the subject. For this, the project aims to extend recent advances in recommender systems and reinforcement learning methods to adapt tests. The project's case study focuses on psychiatric tests, an important public health issue. The current use of psychiatric tests generates significant costs that will be reduced by the approach developed in the project while maintaining their reliability. The major obstacle of adaptive test is to optimize its administration so that it can evaluate several psychiatric dimensions in a short period of time. Three objectives are targeted in the project:
1. The main objective is to propose a reinforcement learning algorithm trained on the data collected during psychiatric tests in order to determine, after each patient answer, the next question or the next cognitive test to be carried out taking into account the previous answers, so as to reduce the handover time.
2. The second objective is to evaluate this method on real psychiatric data, with validation by psychiatrists. Simpler approaches will also be used for comparison purpose.
3. The third objective is to provide an interpretation of the results in order to be able to explain and justify the evaluation of a patient on the psychiatric dimensions. The questions not answered will have to be justified by the system in the light of the other answered questions and some statistics obtained during the learning phase of the model.