CLIP-DSA: Textual Knowledge-Guided Cerebrovascular Diseases Recognition in Multi-view Digital Subtraction Angiography

Qihang Xie; Dan Zhang; Mengting Liu; Jianwei Zhang; Ruisheng Su; Caifeng Shan; Jiong Zhang

doi:10.1007/978-3-032-04978-0_7

CLIP-DSA: Textual Knowledge-Guided Cerebrovascular Diseases Recognition in Multi-view Digital Subtraction Angiography

Qihang Xie
, Dan Zhang (Corresponding author-nrf)
, Mengting Liu
, Jianwei Zhang
, Ruisheng Su
, Caifeng Shan
, Jiong Zhang (Corresponding author-nrf)

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

Abstract

Digital Subtraction Angiography (DSA) sequences are the gold standard for diagnosing most Cerebrovascular diseases (CVDs). Rapid and accurate recognition of CVDs in DSA sequences helps clinicians make the right decisions, which is important in clinical practice. However, the pathological characteristics of CVDs are numerous and complex, and the spatiotemporal complexity of DSA sequences is high, making the diagnosis of CVDs challenging. Therefore, in this paper, we propose a novel CVDs classification framework CLIP-DSA based on CLIP, a pre-trained vision language model. We aim to utilize textual knowledge to guide the robust classification of common CVDs in multi-view DSA sequences. Specifically, our CLIP-DSA comprises a dual-branch vision encoder and a text encoder. The vision encoder is used to extract features from multi-view sequences, while the text encoder is used to obtain textual knowledge. To optimally harness the temporal information in DSA sequences, we introduce a temporal pooling module that dynamically compresses image features in the time dimension. Additionally, we design a multi-view contrastive loss to enhance the network’s image-text representation ability by constraining the image features between two views. In a large dataset with 2,026 patients, the proposed CLIP-DSA achieved an AUC of 90.8% in the CVDs classification. The code is available at this website (https://github.com/jiongzhang-john/CLIP-DSA).

Original language	English
Title of host publication	Medical Image Computing and Computer Assisted Intervention, MICCAI 2025
Subtitle of host publication	28th International Conference, Daejeon, South Korea, September 23–27, 2025, Proceedings
Editors	James C. Gee, Daniel C. Alexander, Jaesung Hong, Juan Eugenio Iglesias, Carole H. Sudre, Archana Venkataraman, Polina Golland, Jong Hyo Kim, Jinah Park
Place of Publication	Cham
Publisher	Springer
Pages	68-77
Number of pages	10
Volume	VI
ISBN (Electronic)	978-3-032-04978-0
ISBN (Print)	978-3-032-04977-3
DOIs	https://doi.org/10.1007/978-3-032-04978-0_7
Publication status	Published - 19 Sept 2025
Event	28th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2025 - Daejeon, Korea, Republic of Duration: 23 Sept 2025 → 27 Sept 2025

Publication series

Name	Lecture Notes in Computer Science (LNCS)
Volume	15965
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	28th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2025
Country/Territory	Korea, Republic of
City	Daejeon
Period	23/09/25 → 27/09/25

Bibliographical note

Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.

Keywords

Cerebrovascular Diseases
Digital Subtraction Angiography
Image-Text
Vision Language Model

Access to Document

10.1007/978-3-032-04978-0_7

Cite this

Xie, Q., Zhang, D., Liu, M., Zhang, J., Su, R., Shan, C., & Zhang, J. (2025). CLIP-DSA: Textual Knowledge-Guided Cerebrovascular Diseases Recognition in Multi-view Digital Subtraction Angiography. In J. C. Gee, D. C. Alexander, J. Hong, J. E. Iglesias, C. H. Sudre, A. Venkataraman, P. Golland, J. H. Kim, & J. Park (Eds.), Medical Image Computing and Computer Assisted Intervention, MICCAI 2025: 28th International Conference, Daejeon, South Korea, September 23–27, 2025, Proceedings (Vol. VI, pp. 68-77). (Lecture Notes in Computer Science (LNCS); Vol. 15965). Springer. https://doi.org/10.1007/978-3-032-04978-0_7

Xie, Qihang ; Zhang, Dan ; Liu, Mengting et al. / CLIP-DSA : Textual Knowledge-Guided Cerebrovascular Diseases Recognition in Multi-view Digital Subtraction Angiography. Medical Image Computing and Computer Assisted Intervention, MICCAI 2025: 28th International Conference, Daejeon, South Korea, September 23–27, 2025, Proceedings. editor / James C. Gee ; Daniel C. Alexander ; Jaesung Hong ; Juan Eugenio Iglesias ; Carole H. Sudre ; Archana Venkataraman ; Polina Golland ; Jong Hyo Kim ; Jinah Park. Vol. VI Cham : Springer, 2025. pp. 68-77 (Lecture Notes in Computer Science (LNCS)).

@inproceedings{275c9ea4b8b1467f9efe261d105dfea6,

title = "CLIP-DSA: Textual Knowledge-Guided Cerebrovascular Diseases Recognition in Multi-view Digital Subtraction Angiography",

abstract = "Digital Subtraction Angiography (DSA) sequences are the gold standard for diagnosing most Cerebrovascular diseases (CVDs). Rapid and accurate recognition of CVDs in DSA sequences helps clinicians make the right decisions, which is important in clinical practice. However, the pathological characteristics of CVDs are numerous and complex, and the spatiotemporal complexity of DSA sequences is high, making the diagnosis of CVDs challenging. Therefore, in this paper, we propose a novel CVDs classification framework CLIP-DSA based on CLIP, a pre-trained vision language model. We aim to utilize textual knowledge to guide the robust classification of common CVDs in multi-view DSA sequences. Specifically, our CLIP-DSA comprises a dual-branch vision encoder and a text encoder. The vision encoder is used to extract features from multi-view sequences, while the text encoder is used to obtain textual knowledge. To optimally harness the temporal information in DSA sequences, we introduce a temporal pooling module that dynamically compresses image features in the time dimension. Additionally, we design a multi-view contrastive loss to enhance the network{\textquoteright}s image-text representation ability by constraining the image features between two views. In a large dataset with 2,026 patients, the proposed CLIP-DSA achieved an AUC of 90.8\% in the CVDs classification. The code is available at this website (https://github.com/jiongzhang-john/CLIP-DSA).",

keywords = "Cerebrovascular Diseases, Digital Subtraction Angiography, Image-Text, Vision Language Model",

author = "Qihang Xie and Dan Zhang and Mengting Liu and Jianwei Zhang and Ruisheng Su and Caifeng Shan and Jiong Zhang",

note = "Publisher Copyright: {\textcopyright} The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.; 28th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2025 ; Conference date: 23-09-2025 Through 27-09-2025",

year = "2025",

month = sep,

day = "19",

doi = "10.1007/978-3-032-04978-0\_7",

language = "English",

isbn = "978-3-032-04977-3",

volume = "VI",

series = "Lecture Notes in Computer Science (LNCS)",

publisher = "Springer",

pages = "68--77",

editor = "Gee, \{James C.\} and Alexander, \{Daniel C.\} and Jaesung Hong and Iglesias, \{Juan Eugenio\} and Sudre, \{Carole H.\} and Archana Venkataraman and Polina Golland and Kim, \{Jong Hyo\} and Jinah Park",

booktitle = "Medical Image Computing and Computer Assisted Intervention, MICCAI 2025",

address = "Germany",

}

Xie, Q, Zhang, D, Liu, M, Zhang, J, Su, R, Shan, C & Zhang, J 2025, CLIP-DSA: Textual Knowledge-Guided Cerebrovascular Diseases Recognition in Multi-view Digital Subtraction Angiography. in JC Gee, DC Alexander, J Hong, JE Iglesias, CH Sudre, A Venkataraman, P Golland, JH Kim & J Park (eds), Medical Image Computing and Computer Assisted Intervention, MICCAI 2025: 28th International Conference, Daejeon, South Korea, September 23–27, 2025, Proceedings. vol. VI, Lecture Notes in Computer Science (LNCS), vol. 15965, Springer, Cham, pp. 68-77, 28th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2025, Daejeon, Korea, Republic of, 23/09/25. https://doi.org/10.1007/978-3-032-04978-0_7

CLIP-DSA: Textual Knowledge-Guided Cerebrovascular Diseases Recognition in Multi-view Digital Subtraction Angiography. / Xie, Qihang; Zhang, Dan (Corresponding author-nrf); Liu, Mengting et al.
Medical Image Computing and Computer Assisted Intervention, MICCAI 2025: 28th International Conference, Daejeon, South Korea, September 23–27, 2025, Proceedings. ed. / James C. Gee; Daniel C. Alexander; Jaesung Hong; Juan Eugenio Iglesias; Carole H. Sudre; Archana Venkataraman; Polina Golland; Jong Hyo Kim; Jinah Park. Vol. VI Cham: Springer, 2025. p. 68-77 (Lecture Notes in Computer Science (LNCS); Vol. 15965).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

TY - GEN

T1 - CLIP-DSA

T2 - 28th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2025

AU - Xie, Qihang

AU - Liu, Mengting

AU - Zhang, Jianwei

AU - Su, Ruisheng

AU - Shan, Caifeng

A2 - Zhang, Dan

A2 - Zhang, Jiong

A2 - Gee, James C.

A2 - Alexander, Daniel C.

A2 - Hong, Jaesung

A2 - Iglesias, Juan Eugenio

A2 - Sudre, Carole H.

A2 - Venkataraman, Archana

A2 - Golland, Polina

A2 - Kim, Jong Hyo

A2 - Park, Jinah

PY - 2025/9/19

Y1 - 2025/9/19

N2 - Digital Subtraction Angiography (DSA) sequences are the gold standard for diagnosing most Cerebrovascular diseases (CVDs). Rapid and accurate recognition of CVDs in DSA sequences helps clinicians make the right decisions, which is important in clinical practice. However, the pathological characteristics of CVDs are numerous and complex, and the spatiotemporal complexity of DSA sequences is high, making the diagnosis of CVDs challenging. Therefore, in this paper, we propose a novel CVDs classification framework CLIP-DSA based on CLIP, a pre-trained vision language model. We aim to utilize textual knowledge to guide the robust classification of common CVDs in multi-view DSA sequences. Specifically, our CLIP-DSA comprises a dual-branch vision encoder and a text encoder. The vision encoder is used to extract features from multi-view sequences, while the text encoder is used to obtain textual knowledge. To optimally harness the temporal information in DSA sequences, we introduce a temporal pooling module that dynamically compresses image features in the time dimension. Additionally, we design a multi-view contrastive loss to enhance the network’s image-text representation ability by constraining the image features between two views. In a large dataset with 2,026 patients, the proposed CLIP-DSA achieved an AUC of 90.8% in the CVDs classification. The code is available at this website (https://github.com/jiongzhang-john/CLIP-DSA).

AB - Digital Subtraction Angiography (DSA) sequences are the gold standard for diagnosing most Cerebrovascular diseases (CVDs). Rapid and accurate recognition of CVDs in DSA sequences helps clinicians make the right decisions, which is important in clinical practice. However, the pathological characteristics of CVDs are numerous and complex, and the spatiotemporal complexity of DSA sequences is high, making the diagnosis of CVDs challenging. Therefore, in this paper, we propose a novel CVDs classification framework CLIP-DSA based on CLIP, a pre-trained vision language model. We aim to utilize textual knowledge to guide the robust classification of common CVDs in multi-view DSA sequences. Specifically, our CLIP-DSA comprises a dual-branch vision encoder and a text encoder. The vision encoder is used to extract features from multi-view sequences, while the text encoder is used to obtain textual knowledge. To optimally harness the temporal information in DSA sequences, we introduce a temporal pooling module that dynamically compresses image features in the time dimension. Additionally, we design a multi-view contrastive loss to enhance the network’s image-text representation ability by constraining the image features between two views. In a large dataset with 2,026 patients, the proposed CLIP-DSA achieved an AUC of 90.8% in the CVDs classification. The code is available at this website (https://github.com/jiongzhang-john/CLIP-DSA).

KW - Cerebrovascular Diseases

KW - Digital Subtraction Angiography

KW - Image-Text

KW - Vision Language Model

UR - https://www.scopus.com/pages/publications/105017847357

U2 - 10.1007/978-3-032-04978-0_7

DO - 10.1007/978-3-032-04978-0_7

M3 - Conference contribution

AN - SCOPUS:105017847357

SN - 978-3-032-04977-3

VL - VI

T3 - Lecture Notes in Computer Science (LNCS)

SP - 68

EP - 77

BT - Medical Image Computing and Computer Assisted Intervention, MICCAI 2025

PB - Springer

CY - Cham

Y2 - 23 September 2025 through 27 September 2025

ER -

Xie Q, Zhang D, Liu M, Zhang J, Su R, Shan C et al. CLIP-DSA: Textual Knowledge-Guided Cerebrovascular Diseases Recognition in Multi-view Digital Subtraction Angiography. In Gee JC, Alexander DC, Hong J, Iglesias JE, Sudre CH, Venkataraman A, Golland P, Kim JH, Park J, editors, Medical Image Computing and Computer Assisted Intervention, MICCAI 2025: 28th International Conference, Daejeon, South Korea, September 23–27, 2025, Proceedings. Vol. VI. Cham: Springer. 2025. p. 68-77. (Lecture Notes in Computer Science (LNCS)). doi: 10.1007/978-3-032-04978-0_7