Deep Learning to Classify Intraductal Papillary Mucinous Neoplasms Using Magnetic Resonance Imaging

Juan E. Corral, Sarfaraz Hussein, Pujan Kandel, Candice W. Bolan, Ulas Bagci, Michael B Wallace

Research output: Contribution to journalArticle

Abstract

OBJECTIVE: This study aimed to evaluate a deep learning protocol to identify neoplasia in intraductal papillary mucinous neoplasia (IPMN) in comparison to current radiographic criteria. METHODS: A computer-aided framework was designed using convolutional neural networks to classify IPMN. The protocol was applied to magnetic resonance images of the pancreas. Features of IPMN were classified according to American Gastroenterology Association guidelines, Fukuoka guidelines, and the new deep learning protocol. Sensitivity and specificity were calculated using surgically resected cystic lesions or healthy controls. RESULTS: Of 139 cases, 58 (42%) were male; mean (standard deviation) age was 65.3 (11.9) years. Twenty-two percent had normal pancreas; 34%, low-grade dysplasia; 14%, high-grade dysplasia; and 29%, adenocarcinoma. The deep learning protocol sensitivity and specificity to detect dysplasia were 92% and 52%, respectively. Sensitivity and specificity to identify high-grade dysplasia or cancer were 75% and 78%, respectively. Diagnostic performance was similar to radiologic criteria. Areas under the receiver operating curves (95% confidence interval) were 0.76 (0.70-0.84) for American Gastroenterology Association, 0.77 (0.70-0.85) for Fukuoka, and 0.78 (0.71-0.85) for the deep learning protocol (P = 0.90). CONCLUSIONS: The deep learning protocol showed accuracy comparable to current radiographic criteria. Computer-aided frameworks could be implemented as aids for radiologists to identify high-risk IPMN.

Original languageEnglish (US)
Pages (from-to)805-810
Number of pages6
JournalPancreas
Volume48
Issue number6
DOIs
StatePublished - Jul 1 2019
Externally publishedYes

Fingerprint

Magnetic Resonance Imaging
Learning
Neoplasms
Gastroenterology
Sensitivity and Specificity
Pancreas
Guidelines
Adenocarcinoma
Magnetic Resonance Spectroscopy
Confidence Intervals

ASJC Scopus subject areas

  • Internal Medicine
  • Endocrinology, Diabetes and Metabolism
  • Hepatology
  • Endocrinology

Cite this

Corral, J. E., Hussein, S., Kandel, P., Bolan, C. W., Bagci, U., & Wallace, M. B. (2019). Deep Learning to Classify Intraductal Papillary Mucinous Neoplasms Using Magnetic Resonance Imaging. Pancreas, 48(6), 805-810. https://doi.org/10.1097/MPA.0000000000001327

Deep Learning to Classify Intraductal Papillary Mucinous Neoplasms Using Magnetic Resonance Imaging. / Corral, Juan E.; Hussein, Sarfaraz; Kandel, Pujan; Bolan, Candice W.; Bagci, Ulas; Wallace, Michael B.

In: Pancreas, Vol. 48, No. 6, 01.07.2019, p. 805-810.

Research output: Contribution to journalArticle

Corral, JE, Hussein, S, Kandel, P, Bolan, CW, Bagci, U & Wallace, MB 2019, 'Deep Learning to Classify Intraductal Papillary Mucinous Neoplasms Using Magnetic Resonance Imaging', Pancreas, vol. 48, no. 6, pp. 805-810. https://doi.org/10.1097/MPA.0000000000001327
Corral, Juan E. ; Hussein, Sarfaraz ; Kandel, Pujan ; Bolan, Candice W. ; Bagci, Ulas ; Wallace, Michael B. / Deep Learning to Classify Intraductal Papillary Mucinous Neoplasms Using Magnetic Resonance Imaging. In: Pancreas. 2019 ; Vol. 48, No. 6. pp. 805-810.
@article{a0e1e866f60e4c99b30d810bb25e5669,
title = "Deep Learning to Classify Intraductal Papillary Mucinous Neoplasms Using Magnetic Resonance Imaging",
abstract = "OBJECTIVE: This study aimed to evaluate a deep learning protocol to identify neoplasia in intraductal papillary mucinous neoplasia (IPMN) in comparison to current radiographic criteria. METHODS: A computer-aided framework was designed using convolutional neural networks to classify IPMN. The protocol was applied to magnetic resonance images of the pancreas. Features of IPMN were classified according to American Gastroenterology Association guidelines, Fukuoka guidelines, and the new deep learning protocol. Sensitivity and specificity were calculated using surgically resected cystic lesions or healthy controls. RESULTS: Of 139 cases, 58 (42{\%}) were male; mean (standard deviation) age was 65.3 (11.9) years. Twenty-two percent had normal pancreas; 34{\%}, low-grade dysplasia; 14{\%}, high-grade dysplasia; and 29{\%}, adenocarcinoma. The deep learning protocol sensitivity and specificity to detect dysplasia were 92{\%} and 52{\%}, respectively. Sensitivity and specificity to identify high-grade dysplasia or cancer were 75{\%} and 78{\%}, respectively. Diagnostic performance was similar to radiologic criteria. Areas under the receiver operating curves (95{\%} confidence interval) were 0.76 (0.70-0.84) for American Gastroenterology Association, 0.77 (0.70-0.85) for Fukuoka, and 0.78 (0.71-0.85) for the deep learning protocol (P = 0.90). CONCLUSIONS: The deep learning protocol showed accuracy comparable to current radiographic criteria. Computer-aided frameworks could be implemented as aids for radiologists to identify high-risk IPMN.",
author = "Corral, {Juan E.} and Sarfaraz Hussein and Pujan Kandel and Bolan, {Candice W.} and Ulas Bagci and Wallace, {Michael B}",
year = "2019",
month = "7",
day = "1",
doi = "10.1097/MPA.0000000000001327",
language = "English (US)",
volume = "48",
pages = "805--810",
journal = "Pancreas",
issn = "0885-3177",
publisher = "Lippincott Williams and Wilkins",
number = "6",

}

TY - JOUR

T1 - Deep Learning to Classify Intraductal Papillary Mucinous Neoplasms Using Magnetic Resonance Imaging

AU - Corral, Juan E.

AU - Hussein, Sarfaraz

AU - Kandel, Pujan

AU - Bolan, Candice W.

AU - Bagci, Ulas

AU - Wallace, Michael B

PY - 2019/7/1

Y1 - 2019/7/1

N2 - OBJECTIVE: This study aimed to evaluate a deep learning protocol to identify neoplasia in intraductal papillary mucinous neoplasia (IPMN) in comparison to current radiographic criteria. METHODS: A computer-aided framework was designed using convolutional neural networks to classify IPMN. The protocol was applied to magnetic resonance images of the pancreas. Features of IPMN were classified according to American Gastroenterology Association guidelines, Fukuoka guidelines, and the new deep learning protocol. Sensitivity and specificity were calculated using surgically resected cystic lesions or healthy controls. RESULTS: Of 139 cases, 58 (42%) were male; mean (standard deviation) age was 65.3 (11.9) years. Twenty-two percent had normal pancreas; 34%, low-grade dysplasia; 14%, high-grade dysplasia; and 29%, adenocarcinoma. The deep learning protocol sensitivity and specificity to detect dysplasia were 92% and 52%, respectively. Sensitivity and specificity to identify high-grade dysplasia or cancer were 75% and 78%, respectively. Diagnostic performance was similar to radiologic criteria. Areas under the receiver operating curves (95% confidence interval) were 0.76 (0.70-0.84) for American Gastroenterology Association, 0.77 (0.70-0.85) for Fukuoka, and 0.78 (0.71-0.85) for the deep learning protocol (P = 0.90). CONCLUSIONS: The deep learning protocol showed accuracy comparable to current radiographic criteria. Computer-aided frameworks could be implemented as aids for radiologists to identify high-risk IPMN.

AB - OBJECTIVE: This study aimed to evaluate a deep learning protocol to identify neoplasia in intraductal papillary mucinous neoplasia (IPMN) in comparison to current radiographic criteria. METHODS: A computer-aided framework was designed using convolutional neural networks to classify IPMN. The protocol was applied to magnetic resonance images of the pancreas. Features of IPMN were classified according to American Gastroenterology Association guidelines, Fukuoka guidelines, and the new deep learning protocol. Sensitivity and specificity were calculated using surgically resected cystic lesions or healthy controls. RESULTS: Of 139 cases, 58 (42%) were male; mean (standard deviation) age was 65.3 (11.9) years. Twenty-two percent had normal pancreas; 34%, low-grade dysplasia; 14%, high-grade dysplasia; and 29%, adenocarcinoma. The deep learning protocol sensitivity and specificity to detect dysplasia were 92% and 52%, respectively. Sensitivity and specificity to identify high-grade dysplasia or cancer were 75% and 78%, respectively. Diagnostic performance was similar to radiologic criteria. Areas under the receiver operating curves (95% confidence interval) were 0.76 (0.70-0.84) for American Gastroenterology Association, 0.77 (0.70-0.85) for Fukuoka, and 0.78 (0.71-0.85) for the deep learning protocol (P = 0.90). CONCLUSIONS: The deep learning protocol showed accuracy comparable to current radiographic criteria. Computer-aided frameworks could be implemented as aids for radiologists to identify high-risk IPMN.

UR - http://www.scopus.com/inward/record.url?scp=85068420002&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85068420002&partnerID=8YFLogxK

U2 - 10.1097/MPA.0000000000001327

DO - 10.1097/MPA.0000000000001327

M3 - Article

VL - 48

SP - 805

EP - 810

JO - Pancreas

JF - Pancreas

SN - 0885-3177

IS - 6

ER -