Vision Transformer-based Classification Study of Intracranial Hemorrhage

Intracranial hemorrhage is described as bleeding within the skull. It is a serious and heavy craniosynostosis, recognized for its high mortality and lethality rate, and generally requires urgent follow-up diagnosis and determination of the location and subtype of the intracranial hemorrhage lesion in order to improve the probability of successfully rescuing the patient. In this paper, we propose an end-to-end Vision Transformer-based model to address the drawbacks of existing RNN and LSTM-based intracranial hemorrhage detection algorithms, such as poor parallelism and long training time, which greatly reduces the training time of the model while considering the time-series characteristics of cranial CT images. Meanwhile, under the DICOM format, the range values of Hu in the images acquired from CT scans will be varied for different kinds of tissues and various pathologies. On this basis, considering the window functions that radiologists often resort to in clinical diagnosis, this paper studies and compares the brain CT images under different windows and selects the pre-processed data after the best window setting as the input of CNN detection model to improve the accuracy of intracranial hemorrhage identification.