A novel approach for selecting combination clinical markers of pathology applied to a large retrospective cohort of surgically resected pancreatic cysts

David L. Masica, Marco Dal Molin, Christopher Wolfgang, Tyler Tomita, Mohammad R. Ostovaneh, Amanda Blackford, Robert A. Moran, Joanna K. Law, Thomas Barkley, Michael S Goggins, Marcia Canto, Meredith Pittman, James Eshleman, Syed Z Ali, Elliot K Fishman, Ihab R Kamel, Siva P. Raman, Atif Zaheer, Nita Ahuja, Martin A Makary & 16 others Matthew J Weiss, Kenzo Hirose, John L Cameron, Neda Rezaee, Jin He, Young Joon Ahn, Wenchuan Wu, Yuxuan Wang, Simeon Springer, Luis L. Diaz, Nickolas Papadopoulos, Ralph H Hruban, Kenneth W Kinzler, Bert Vogelstein, Rachel Karchin, Anne Marie O'Broin-Lennon

Research output: Contribution to journalArticle

Abstract

Objective: Our objective was to develop an approach for selecting combinatorial markers of pathology from diverse clinical data types. We demonstrate this approach on the problem of pancreatic cyst classification. Materials and Methods: We analyzed 1026 patients with surgically resected pancreatic cysts, comprising 584 intraductal papillary mucinous neoplasms, 332 serous cystadenomas, 78 mucinous cystic neoplasms, and 42 solid-pseudopapillary neoplasms. To derive optimal markers for cyst classification from the preoperative clinical and radiological data, we developed a statistical approach for combining any number of categorical, dichotomous, or continuous-valued clinical parameters into individual predictors of pathology. The approach is unbiased and statistically rigorous. Millions of feature combinations were tested using 10-fold cross-validation, and the most informative features were validated in an independent cohort of 130 patients with surgically resected pancreatic cysts. Results: We identified combinatorial clinical markers that classified serous cystadenomas with 95% sensitivity and 83% specificity; solid-pseudopapillary neoplasms with 89% sensitivity and 86% specificity; mucinous cystic neoplasms with 91% sensitivity and 83% specificity; and intraductal papillary mucinous neoplasms with 94% sensitivity and 90% specificity. No individual features were as accurate as the combination markers. We further validated these combinatorial markers on an independent cohort of 130 pancreatic cysts, and achieved high and well-balanced accuracies. Overall sensitivity and specificity for identifying patients requiring surgical resection was 84% and 81%, respectively. Conclusions: Our approach identified combinatorial markers for pancreatic cyst classification that had improved performance relative to the individual features they comprise. In principle, this approach can be applied to any clinical dataset comprising dichotomous, categorical, and continuous-valued parameters.

LanguageEnglish (US)
Pages145-152
Number of pages8
JournalJournal of the American Medical Informatics Association
Volume24
Issue number1
DOIs
StatePublished - 2017

Fingerprint

Pancreatic Cyst
Clinical Pathology
Biomarkers
Sensitivity and Specificity
Serous Cystadenoma
Neoplasms
Cysts
Pathology

Keywords

  • Clinical model
  • Combination marker
  • Composite marker
  • IPMN
  • MOCA
  • Mucinous cyst
  • Pancreatic cyst

ASJC Scopus subject areas

  • Health Informatics

Cite this

A novel approach for selecting combination clinical markers of pathology applied to a large retrospective cohort of surgically resected pancreatic cysts. / Masica, David L.; Molin, Marco Dal; Wolfgang, Christopher; Tomita, Tyler; Ostovaneh, Mohammad R.; Blackford, Amanda; Moran, Robert A.; Law, Joanna K.; Barkley, Thomas; Goggins, Michael S; Canto, Marcia; Pittman, Meredith; Eshleman, James; Ali, Syed Z; Fishman, Elliot K; Kamel, Ihab R; Raman, Siva P.; Zaheer, Atif; Ahuja, Nita; Makary, Martin A; Weiss, Matthew J; Hirose, Kenzo; Cameron, John L; Rezaee, Neda; He, Jin; Ahn, Young Joon; Wu, Wenchuan; Wang, Yuxuan; Springer, Simeon; Diaz, Luis L.; Papadopoulos, Nickolas; Hruban, Ralph H; Kinzler, Kenneth W; Vogelstein, Bert; Karchin, Rachel; O'Broin-Lennon, Anne Marie.

In: Journal of the American Medical Informatics Association, Vol. 24, No. 1, 2017, p. 145-152.

Research output: Contribution to journalArticle

Masica, David L. ; Molin, Marco Dal ; Wolfgang, Christopher ; Tomita, Tyler ; Ostovaneh, Mohammad R. ; Blackford, Amanda ; Moran, Robert A. ; Law, Joanna K. ; Barkley, Thomas ; Goggins, Michael S ; Canto, Marcia ; Pittman, Meredith ; Eshleman, James ; Ali, Syed Z ; Fishman, Elliot K ; Kamel, Ihab R ; Raman, Siva P. ; Zaheer, Atif ; Ahuja, Nita ; Makary, Martin A ; Weiss, Matthew J ; Hirose, Kenzo ; Cameron, John L ; Rezaee, Neda ; He, Jin ; Ahn, Young Joon ; Wu, Wenchuan ; Wang, Yuxuan ; Springer, Simeon ; Diaz, Luis L. ; Papadopoulos, Nickolas ; Hruban, Ralph H ; Kinzler, Kenneth W ; Vogelstein, Bert ; Karchin, Rachel ; O'Broin-Lennon, Anne Marie. / A novel approach for selecting combination clinical markers of pathology applied to a large retrospective cohort of surgically resected pancreatic cysts. In: Journal of the American Medical Informatics Association. 2017 ; Vol. 24, No. 1. pp. 145-152.
@article{c56890a0b473419da3e91ffd76183ca7,
title = "A novel approach for selecting combination clinical markers of pathology applied to a large retrospective cohort of surgically resected pancreatic cysts",
abstract = "Objective: Our objective was to develop an approach for selecting combinatorial markers of pathology from diverse clinical data types. We demonstrate this approach on the problem of pancreatic cyst classification. Materials and Methods: We analyzed 1026 patients with surgically resected pancreatic cysts, comprising 584 intraductal papillary mucinous neoplasms, 332 serous cystadenomas, 78 mucinous cystic neoplasms, and 42 solid-pseudopapillary neoplasms. To derive optimal markers for cyst classification from the preoperative clinical and radiological data, we developed a statistical approach for combining any number of categorical, dichotomous, or continuous-valued clinical parameters into individual predictors of pathology. The approach is unbiased and statistically rigorous. Millions of feature combinations were tested using 10-fold cross-validation, and the most informative features were validated in an independent cohort of 130 patients with surgically resected pancreatic cysts. Results: We identified combinatorial clinical markers that classified serous cystadenomas with 95{\%} sensitivity and 83{\%} specificity; solid-pseudopapillary neoplasms with 89{\%} sensitivity and 86{\%} specificity; mucinous cystic neoplasms with 91{\%} sensitivity and 83{\%} specificity; and intraductal papillary mucinous neoplasms with 94{\%} sensitivity and 90{\%} specificity. No individual features were as accurate as the combination markers. We further validated these combinatorial markers on an independent cohort of 130 pancreatic cysts, and achieved high and well-balanced accuracies. Overall sensitivity and specificity for identifying patients requiring surgical resection was 84{\%} and 81{\%}, respectively. Conclusions: Our approach identified combinatorial markers for pancreatic cyst classification that had improved performance relative to the individual features they comprise. In principle, this approach can be applied to any clinical dataset comprising dichotomous, categorical, and continuous-valued parameters.",
keywords = "Clinical model, Combination marker, Composite marker, IPMN, MOCA, Mucinous cyst, Pancreatic cyst",
author = "Masica, {David L.} and Molin, {Marco Dal} and Christopher Wolfgang and Tyler Tomita and Ostovaneh, {Mohammad R.} and Amanda Blackford and Moran, {Robert A.} and Law, {Joanna K.} and Thomas Barkley and Goggins, {Michael S} and Marcia Canto and Meredith Pittman and James Eshleman and Ali, {Syed Z} and Fishman, {Elliot K} and Kamel, {Ihab R} and Raman, {Siva P.} and Atif Zaheer and Nita Ahuja and Makary, {Martin A} and Weiss, {Matthew J} and Kenzo Hirose and Cameron, {John L} and Neda Rezaee and Jin He and Ahn, {Young Joon} and Wenchuan Wu and Yuxuan Wang and Simeon Springer and Diaz, {Luis L.} and Nickolas Papadopoulos and Hruban, {Ralph H} and Kinzler, {Kenneth W} and Bert Vogelstein and Rachel Karchin and O'Broin-Lennon, {Anne Marie}",
year = "2017",
doi = "10.1093/jamia/ocw069",
language = "English (US)",
volume = "24",
pages = "145--152",
journal = "Journal of the American Medical Informatics Association : JAMIA",
issn = "1067-5027",
publisher = "Oxford University Press",
number = "1",

}

TY - JOUR

T1 - A novel approach for selecting combination clinical markers of pathology applied to a large retrospective cohort of surgically resected pancreatic cysts

AU - Masica, David L.

AU - Molin, Marco Dal

AU - Wolfgang, Christopher

AU - Tomita, Tyler

AU - Ostovaneh, Mohammad R.

AU - Blackford, Amanda

AU - Moran, Robert A.

AU - Law, Joanna K.

AU - Barkley, Thomas

AU - Goggins, Michael S

AU - Canto, Marcia

AU - Pittman, Meredith

AU - Eshleman, James

AU - Ali, Syed Z

AU - Fishman, Elliot K

AU - Kamel, Ihab R

AU - Raman, Siva P.

AU - Zaheer, Atif

AU - Ahuja, Nita

AU - Makary, Martin A

AU - Weiss, Matthew J

AU - Hirose, Kenzo

AU - Cameron, John L

AU - Rezaee, Neda

AU - He, Jin

AU - Ahn, Young Joon

AU - Wu, Wenchuan

AU - Wang, Yuxuan

AU - Springer, Simeon

AU - Diaz, Luis L.

AU - Papadopoulos, Nickolas

AU - Hruban, Ralph H

AU - Kinzler, Kenneth W

AU - Vogelstein, Bert

AU - Karchin, Rachel

AU - O'Broin-Lennon, Anne Marie

PY - 2017

Y1 - 2017

N2 - Objective: Our objective was to develop an approach for selecting combinatorial markers of pathology from diverse clinical data types. We demonstrate this approach on the problem of pancreatic cyst classification. Materials and Methods: We analyzed 1026 patients with surgically resected pancreatic cysts, comprising 584 intraductal papillary mucinous neoplasms, 332 serous cystadenomas, 78 mucinous cystic neoplasms, and 42 solid-pseudopapillary neoplasms. To derive optimal markers for cyst classification from the preoperative clinical and radiological data, we developed a statistical approach for combining any number of categorical, dichotomous, or continuous-valued clinical parameters into individual predictors of pathology. The approach is unbiased and statistically rigorous. Millions of feature combinations were tested using 10-fold cross-validation, and the most informative features were validated in an independent cohort of 130 patients with surgically resected pancreatic cysts. Results: We identified combinatorial clinical markers that classified serous cystadenomas with 95% sensitivity and 83% specificity; solid-pseudopapillary neoplasms with 89% sensitivity and 86% specificity; mucinous cystic neoplasms with 91% sensitivity and 83% specificity; and intraductal papillary mucinous neoplasms with 94% sensitivity and 90% specificity. No individual features were as accurate as the combination markers. We further validated these combinatorial markers on an independent cohort of 130 pancreatic cysts, and achieved high and well-balanced accuracies. Overall sensitivity and specificity for identifying patients requiring surgical resection was 84% and 81%, respectively. Conclusions: Our approach identified combinatorial markers for pancreatic cyst classification that had improved performance relative to the individual features they comprise. In principle, this approach can be applied to any clinical dataset comprising dichotomous, categorical, and continuous-valued parameters.

AB - Objective: Our objective was to develop an approach for selecting combinatorial markers of pathology from diverse clinical data types. We demonstrate this approach on the problem of pancreatic cyst classification. Materials and Methods: We analyzed 1026 patients with surgically resected pancreatic cysts, comprising 584 intraductal papillary mucinous neoplasms, 332 serous cystadenomas, 78 mucinous cystic neoplasms, and 42 solid-pseudopapillary neoplasms. To derive optimal markers for cyst classification from the preoperative clinical and radiological data, we developed a statistical approach for combining any number of categorical, dichotomous, or continuous-valued clinical parameters into individual predictors of pathology. The approach is unbiased and statistically rigorous. Millions of feature combinations were tested using 10-fold cross-validation, and the most informative features were validated in an independent cohort of 130 patients with surgically resected pancreatic cysts. Results: We identified combinatorial clinical markers that classified serous cystadenomas with 95% sensitivity and 83% specificity; solid-pseudopapillary neoplasms with 89% sensitivity and 86% specificity; mucinous cystic neoplasms with 91% sensitivity and 83% specificity; and intraductal papillary mucinous neoplasms with 94% sensitivity and 90% specificity. No individual features were as accurate as the combination markers. We further validated these combinatorial markers on an independent cohort of 130 pancreatic cysts, and achieved high and well-balanced accuracies. Overall sensitivity and specificity for identifying patients requiring surgical resection was 84% and 81%, respectively. Conclusions: Our approach identified combinatorial markers for pancreatic cyst classification that had improved performance relative to the individual features they comprise. In principle, this approach can be applied to any clinical dataset comprising dichotomous, categorical, and continuous-valued parameters.

KW - Clinical model

KW - Combination marker

KW - Composite marker

KW - IPMN

KW - MOCA

KW - Mucinous cyst

KW - Pancreatic cyst

UR - http://www.scopus.com/inward/record.url?scp=85014759428&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85014759428&partnerID=8YFLogxK

U2 - 10.1093/jamia/ocw069

DO - 10.1093/jamia/ocw069

M3 - Article

VL - 24

SP - 145

EP - 152

JO - Journal of the American Medical Informatics Association : JAMIA

T2 - Journal of the American Medical Informatics Association : JAMIA

JF - Journal of the American Medical Informatics Association : JAMIA

SN - 1067-5027

IS - 1

ER -