KRASFormer: a fully vision transformer-based framework for predicting KRAS gene mutations in histopathological images of colorectal cancer

被引:0
作者
Singh, Vivek Kumar [1 ,2 ]
Makhlouf, Yasmine [1 ]
Sarker, M. Mostafa Kamal [3 ]
Craig, Stephanie [1 ]
Baena, Juvenal [1 ]
Greene, Christine [1 ]
Mason, Lee [1 ]
James, Jacqueline A. [1 ,4 ]
Salto-Tellez, Manuel [1 ,4 ,5 ,6 ]
O'Reilly, Paul [5 ]
Maxwell, Perry [1 ]
机构
[1] Queens Univ Belfast, Precis Med Ctr Excellence, Patrick G Johnston Ctr Canc Res, Hlth Sci Bldg, Belfast BT9 7AE, North Ireland
[2] Queen Mary Univ London, Barts Canc Inst, Ctr Biomarkers & Biotherapeut, London EC1M 6BQ, England
[3] Univ Oxford, Inst Biomed Engn, Oxford OX37DQ, England
[4] Belfast Hlth & Social Care Trust, Reg Mol Diagnost Serv, Belfast BT9 7AE, North Ireland
[5] Sonrai Analyt, Belfast BT9 7AE, North Ireland
[6] Belfast City Hosp, Belfast Hlth & Social Care Trust, Cellular Pathol, Lisburn Rd, Belfast BT9 7AB, North Ireland
关键词
colorectal cancer; vision-transformer; KRAS gene; whole slide image; mutation classification; next generation sequencing;
D O I
10.1088/2057-1976/ad5bed
中图分类号
R8 [特种医学]; R445 [影像诊断学];
学科分类号
1002 ; 100207 ; 1009 ;
摘要
Detecting the Kirsten Rat Sarcoma Virus (KRAS) gene mutation is significant for colorectal cancer (CRC) patients. The KRAS gene encodes a protein involved in the epidermal growth factor receptor (EGFR) signaling pathway, and mutations in this gene can negatively impact the use of monoclonal antibodies in anti-EGFR therapy and affect treatment decisions. Currently, commonly used methods like next-generation sequencing (NGS) identify KRAS mutations but are expensive, time-consuming, and may not be suitable for every cancer patient sample. To address these challenges, we have developed KRASFormer, a novel framework that predicts KRAS gene mutations from Haematoxylin and Eosin (H & E) stained WSIs that are widely available for most CRC patients. KRASFormer consists of two stages: the first stage filters out non-tumor regions and selects only tumour cells using a quality screening mechanism, and the second stage predicts the KRAS gene either wildtype' or mutant' using a Vision Transformer-based XCiT method. The XCiT employs cross-covariance attention to capture clinically meaningful long-range representations of textural patterns in tumour tissue and KRAS mutant cells. We evaluated the performance of the first stage using an independent CRC-5000 dataset, and the second stage included both The Cancer Genome Atlas colon and rectal cancer (TCGA-CRC-DX) and in-house cohorts. The results of our experiments showed that the XCiT outperformed existing state-of-the-art methods, achieving AUCs for ROC curves of 0.691 and 0.653 on TCGA-CRC-DX and in-house datasets, respectively. Our findings emphasize three key consequences: the potential of using H & E-stained tissue slide images for predicting KRAS gene mutations as a cost-effective and time-efficient means for guiding treatment choice with CRC patients; the increase in performance metrics of a Transformer-based model; and the value of the collaboration between pathologists and data scientists in deriving a morphologically meaningful model.
引用
收藏
页数:14
相关论文
共 36 条
[1]   Recent application of artificial intelligence on histopathologic image-based prediction of gene mutation in solid cancers [J].
Alam, Mohammad Rizwan ;
Seo, Kyung Jin ;
Abdul-Ghafar, Jamshid ;
Yim, Kwangil ;
Lee, Sung Hak ;
Jang, Hyun-Jong ;
Jung, Chan Kwon ;
Chong, Yosep .
BRIEFINGS IN BIOINFORMATICS, 2023, 24 (03)
[2]  
Ali A., 2021, Adv. Neural Inf. Process. Syst., V34, DOI DOI 10.48550/ARXIV.2106.09681
[3]   QuPath: Open source software for digital pathology image analysis [J].
Bankhead, Peter ;
Loughrey, Maurice B. ;
Fernandez, Jose A. ;
Dombrowski, Yvonne ;
Mcart, Darragh G. ;
Dunne, Philip D. ;
McQuaid, Stephen ;
Gray, Ronan T. ;
Murray, Liam J. ;
Coleman, Helen G. ;
James, Jacqueline A. ;
Salto-Tellez, Manuel ;
Hamilton, Peter W. .
SCIENTIFIC REPORTS, 2017, 7
[4]  
Bao H., 2021, arXiv
[5]   Development and validation of a weakly supervised deep learning framework to predict the status of molecular pathways and key mutations in colorectal cancer from routine histology images: a retrospective study [J].
Bilal, Mohsin ;
Raza, Shan E. Ahmed ;
Azam, Ayesha ;
Graham, Simon ;
Ilyas, Mohammad ;
Cree, Ian A. ;
Snead, David ;
Minhas, Fayyaz ;
Rajpoot, Nasir M. .
LANCET DIGITAL HEALTH, 2021, 3 (12) :E763-E772
[6]   The Cancer Imaging Archive (TCIA): Maintaining and Operating a Public Information Repository [J].
Clark, Kenneth ;
Vendt, Bruce ;
Smith, Kirk ;
Freymann, John ;
Kirby, Justin ;
Koppel, Paul ;
Moore, Stephen ;
Phillips, Stanley ;
Maffitt, David ;
Pringle, Michael ;
Tarbox, Lawrence ;
Prior, Fred .
JOURNAL OF DIGITAL IMAGING, 2013, 26 (06) :1045-1057
[7]  
Deininger L., 2022, arXiv, DOI DOI 10.48550/ARXIV.2206.00389
[8]  
Deng J, 2009, PROC CVPR IEEE, P248, DOI 10.1109/CVPRW.2009.5206848
[9]  
Dosovitskiy A, 2021, Arxiv, DOI [arXiv:2010.11929, DOI 10.48550/ARXIV.2010.11929]
[10]   An introduction to ROC analysis [J].
Fawcett, Tom .
PATTERN RECOGNITION LETTERS, 2006, 27 (08) :861-874