Identifying the risk of emotional issues is of great significance to the development of adolescents. Electroencephalography (EEG) signals can reflect brain activities, and imply the emotional status, therefore can be used to identify emotional issues. In this study, we designed an experimental paradigm for high school adolescents by displaying emotion-inducing pictures as stimuli and recorded their EEG signals simultaneously. The EEG signals was preprocessed and analyzed. In this paper we applied a convolutional network EEGNet to classify four emotional issues in adolescents: depression symptom, manic symptom, anxiety symptom and the control group. The results showed that the classification accuracy of the four groups can reach 94.24%. In addition, this paper explored using different types of picture stimuli on the classification and reached a result above chance level. This work extended the previous work on the classification of emotional issues to four categories, and achieved a good classification accuracy. © 2022 IEEE.