An Effective Transformer-based Solution for RSNA Intracranial Hemorrhage Detection Competition

Fangxin Shang, Siqi Wang, Xiaorong Wang, Yehui Yang
Proceedings of The First AAAI Bridge Program on AI for Medicine and Healthcare, PMLR 281:14-19, 2025.

Abstract

We propose an effective method for the detection of intracranial hemorrhages (IHD) that exceeds the performance of the winning solution in the RSNA-IHD competition (2019)(Anouk Stein et al. 2019). Meanwhile, our model only takes quarter parameters and ten percent FLOPs compared to the winner’s solution. The IHD task must predict the hemorrhage category of each slice for the input brain CT. We review the top five solutions for the IHD competition held by the Radiological Society of North America(RSNA) in 2019. Almost all the top solutions rely on 2D convolutional networks and sequential models (Bidirectional GRU or LSTM) to extract intraslice and interslice features, respectively. All the top solutions improve performance by using the ensemble of models, and the number of models varies from 7 to 31. In the past years, since much progress has been made in the computer vision regime especially Transformer-based models, we introduce the Transformer-based techniques to extract the features in both intra-slice and inter-slice views for IHD tasks. Additionally, a semi-supervised method is embedded into our workflow to further improve the performance. The code is already available online. Code — https: //github.com/PaddlePaddle/Research/tree/master/CV. Datasets — https://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection

Cite this Paper

BibTeX
@InProceedings{pmlr-v281-shang25a, title = {An Effective Transformer-based Solution for RSNA Intracranial Hemorrhage Detection Competition}, author = {Shang, Fangxin and Wang, Siqi and Wang, Xiaorong and Yang, Yehui}, booktitle = {Proceedings of The First AAAI Bridge Program on AI for Medicine and Healthcare}, pages = {14--19}, year = {2025}, editor = {Wu, Junde and Zhu, Jiayuan and Xu, Min and Jin, Yueming}, volume = {281}, series = {Proceedings of Machine Learning Research}, month = {25 Feb}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v281/main/assets/shang25a/shang25a.pdf}, url = {https://proceedings.mlr.press/v281/shang25a.html}, abstract = {We propose an effective method for the detection of intracranial hemorrhages (IHD) that exceeds the performance of the winning solution in the RSNA-IHD competition (2019)(Anouk Stein et al. 2019). Meanwhile, our model only takes quarter parameters and ten percent FLOPs compared to the winner’s solution. The IHD task must predict the hemorrhage category of each slice for the input brain CT. We review the top five solutions for the IHD competition held by the Radiological Society of North America(RSNA) in 2019. Almost all the top solutions rely on 2D convolutional networks and sequential models (Bidirectional GRU or LSTM) to extract intraslice and interslice features, respectively. All the top solutions improve performance by using the ensemble of models, and the number of models varies from 7 to 31. In the past years, since much progress has been made in the computer vision regime especially Transformer-based models, we introduce the Transformer-based techniques to extract the features in both intra-slice and inter-slice views for IHD tasks. Additionally, a semi-supervised method is embedded into our workflow to further improve the performance. The code is already available online. Code — https: //github.com/PaddlePaddle/Research/tree/master/CV. Datasets — https://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection} }
Endnote
%0 Conference Paper %T An Effective Transformer-based Solution for RSNA Intracranial Hemorrhage Detection Competition %A Fangxin Shang %A Siqi Wang %A Xiaorong Wang %A Yehui Yang %B Proceedings of The First AAAI Bridge Program on AI for Medicine and Healthcare %C Proceedings of Machine Learning Research %D 2025 %E Junde Wu %E Jiayuan Zhu %E Min Xu %E Yueming Jin %F pmlr-v281-shang25a %I PMLR %P 14--19 %U https://proceedings.mlr.press/v281/shang25a.html %V 281 %X We propose an effective method for the detection of intracranial hemorrhages (IHD) that exceeds the performance of the winning solution in the RSNA-IHD competition (2019)(Anouk Stein et al. 2019). Meanwhile, our model only takes quarter parameters and ten percent FLOPs compared to the winner’s solution. The IHD task must predict the hemorrhage category of each slice for the input brain CT. We review the top five solutions for the IHD competition held by the Radiological Society of North America(RSNA) in 2019. Almost all the top solutions rely on 2D convolutional networks and sequential models (Bidirectional GRU or LSTM) to extract intraslice and interslice features, respectively. All the top solutions improve performance by using the ensemble of models, and the number of models varies from 7 to 31. In the past years, since much progress has been made in the computer vision regime especially Transformer-based models, we introduce the Transformer-based techniques to extract the features in both intra-slice and inter-slice views for IHD tasks. Additionally, a semi-supervised method is embedded into our workflow to further improve the performance. The code is already available online. Code — https: //github.com/PaddlePaddle/Research/tree/master/CV. Datasets — https://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection
APA
Shang, F., Wang, S., Wang, X. & Yang, Y.. (2025). An Effective Transformer-based Solution for RSNA Intracranial Hemorrhage Detection Competition. Proceedings of The First AAAI Bridge Program on AI for Medicine and Healthcare, in Proceedings of Machine Learning Research 281:14-19 Available from https://proceedings.mlr.press/v281/shang25a.html.

Related Material