TY - GEN
T1 - Digital Tumor-Collagen Proximity Signature Predicts Survival in Diffuse Large B-Cell Lymphoma
AU - Qaiser, Talha
AU - Pugh, Matthew
AU - Margielewska, Sandra
AU - Hollows, Robert
AU - Murray, Paul
AU - Rajpoot, Nasir
N1 - Publisher Copyright:
© 2019, Springer Nature Switzerland AG.
PY - 2019
Y1 - 2019
N2 - Diffuse large B-cell lymphoma (DLBCL) is a heterogeneous tumor that originates from normal B-cells. A limited number of studies have investigated the role of acellular stromal microenvironment on outcome in DLBCL. Here, we propose a novel digital proximity signature (DPS) for predicting overall survival (OS) in DLBCL patients. We propose a novel end-to-end multi-task deep learning model for cell detection and classification and investigate the spatial proximity of collagen (type VI) and tumor cells for estimating the DPS. To the best of our knowledge, this is the first study that performs automated analysis of tumor and collagen on DLBCL to identify potential prognostic factors. Experimental results favor our cell classification algorithm over conventional approaches. In addition, our pilot results show that strongly associated tumor-collagen regions are statistically significant (p = 0.03) in predicting OS in DLBCL patients.
AB - Diffuse large B-cell lymphoma (DLBCL) is a heterogeneous tumor that originates from normal B-cells. A limited number of studies have investigated the role of acellular stromal microenvironment on outcome in DLBCL. Here, we propose a novel digital proximity signature (DPS) for predicting overall survival (OS) in DLBCL patients. We propose a novel end-to-end multi-task deep learning model for cell detection and classification and investigate the spatial proximity of collagen (type VI) and tumor cells for estimating the DPS. To the best of our knowledge, this is the first study that performs automated analysis of tumor and collagen on DLBCL to identify potential prognostic factors. Experimental results favor our cell classification algorithm over conventional approaches. In addition, our pilot results show that strongly associated tumor-collagen regions are statistically significant (p = 0.03) in predicting OS in DLBCL patients.
KW - Computational pathology
KW - Deep learning
KW - Survival analysis
UR - http://www.scopus.com/inward/record.url?scp=85069185046&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-23937-4_19
DO - 10.1007/978-3-030-23937-4_19
M3 - Conference contribution
AN - SCOPUS:85069185046
SN - 9783030239367
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 163
EP - 171
BT - Digital Pathology - 15th European Congress, ECDP 2019, Proceedings
A2 - Reyes-Aldasoro, Constantino Carlos
A2 - Janowczyk, Andrew
A2 - Veta, Mitko
A2 - Bankhead, Peter
A2 - Sirinukunwattana, Korsuk
PB - Springer Verlag
T2 - 15th European Congress on Digital Pathology, ECDP 2019
Y2 - 10 April 2019 through 13 April 2019
ER -
