Adam M Awe1,2, Michael M Vanden Heuvel3, Tianyuan Yuan3, Victoria R Rendell2, Mingren Shen4, Agrima Kampani3, Shanchao Liang3, Dane D Morgan4, Emily R Winslow5, Meghan G Lubner6. 1. Department of Radiology, University of Wisconsin School of Medicine & Public Health, E3/311 Clinical Sciences Center, 600 Highland Ave, Madison, WI, 53792, USA. 2. Department of Surgery, University of Wisconsin School of Medicine & Public Health, Madison, WI, USA. 3. Department of Computer Sciences, University of Wisconsin - Madison, Madison, WI, USA. 4. Department of Materials Science and Engineering, University of Wisconsin - Madison, Madison, WI, USA. 5. Medstar Georgetown Transplant Institute, Medstar Georgetown University Hospital, Washington, DC, USA. 6. Department of Radiology, University of Wisconsin School of Medicine & Public Health, E3/311 Clinical Sciences Center, 600 Highland Ave, Madison, WI, 53792, USA. mlubner@uwhealth.org.
Abstract
PURPOSE: Current diagnostic and treatment modalities for pancreatic cysts (PCs) are invasive and are associated with patient morbidity. The purpose of this study is to develop and evaluate machine learning algorithms to delineate mucinous from non-mucinous PCs using non-invasive CT-based radiomics. METHODS: A retrospective, single-institution analysis of patients with non-pseudocystic PCs, contrast-enhanced computed tomography scans within 1 year of resection, and available surgical pathology were included. A quantitative imaging software platform was used to extract radiomics. An extreme gradient boosting (XGBoost) machine learning algorithm was used to create mucinous classifiers using texture features only, or radiomic/radiologic and clinical combined models. Classifiers were compared using performance scoring metrics. Shapely additive explanation (SHAP) analyses were conducted to identify variables most important in model construction. RESULTS: Overall, 99 patients and 103 PCs were included in the analyses. Eighty (78%) patients had mucinous PCs on surgical pathology. Using multiple fivefold cross validations, the texture features only and combined XGBoost mucinous classifiers demonstrated an area under the curve of 0.72 ± 0.14 and 0.73 ± 0.14, respectively. By SHAP analysis, root mean square, mean attenuation, and kurtosis were the most predictive features in the texture features only model. Root mean square, cyst location, and mean attenuation were the most predictive features in the combined model. CONCLUSION: Machine learning principles can be applied to PC texture features to create a mucinous phenotype classifier. Model performance did not improve with the combined model. However, specific radiomic, radiologic, and clinical features most predictive in our models can be identified using SHAP analysis.
PURPOSE: Current diagnostic and treatment modalities for pancreatic cysts (PCs) are invasive and are associated with patient morbidity. The purpose of this study is to develop and evaluate machine learning algorithms to delineate mucinous from non-mucinous PCs using non-invasive CT-based radiomics. METHODS: A retrospective, single-institution analysis of patients with non-pseudocystic PCs, contrast-enhanced computed tomography scans within 1 year of resection, and available surgical pathology were included. A quantitative imaging software platform was used to extract radiomics. An extreme gradient boosting (XGBoost) machine learning algorithm was used to create mucinous classifiers using texture features only, or radiomic/radiologic and clinical combined models. Classifiers were compared using performance scoring metrics. Shapely additive explanation (SHAP) analyses were conducted to identify variables most important in model construction. RESULTS: Overall, 99 patients and 103 PCs were included in the analyses. Eighty (78%) patients had mucinous PCs on surgical pathology. Using multiple fivefold cross validations, the texture features only and combined XGBoost mucinous classifiers demonstrated an area under the curve of 0.72 ± 0.14 and 0.73 ± 0.14, respectively. By SHAP analysis, root mean square, mean attenuation, and kurtosis were the most predictive features in the texture features only model. Root mean square, cyst location, and mean attenuation were the most predictive features in the combined model. CONCLUSION: Machine learning principles can be applied to PC texture features to create a mucinous phenotype classifier. Model performance did not improve with the combined model. However, specific radiomic, radiologic, and clinical features most predictive in our models can be identified using SHAP analysis.
Authors: Masao Tanaka; Carlos Fernández-Del Castillo; Terumi Kamisawa; Jin Young Jang; Philippe Levy; Takao Ohtsuka; Roberto Salvia; Yasuhiro Shimizu; Minoru Tada; Christopher L Wolfgang Journal: Pancreatology Date: 2017-07-13 Impact factor: 3.996
Authors: Marc A Attiyeh; Jayasree Chakraborty; Lior Gazit; Liana Langdon-Embry; Mithat Gonen; Vinod P Balachandran; Michael I D'Angelica; Ronald P DeMatteo; William R Jarnagin; T Peter Kingham; Peter J Allen; Richard K Do; Amber L Simpson Journal: HPB (Oxford) Date: 2018-08-07 Impact factor: 3.647
Authors: Jayasree Chakraborty; Abhishek Midya; Lior Gazit; Marc Attiyeh; Liana Langdon-Embry; Peter J Allen; Richard K G Do; Amber L Simpson Journal: Med Phys Date: 2018-09-27 Impact factor: 4.071
Authors: Clifford S Cho; Andrew J Russ; Agnes G Loeffler; Robert J Rettammel; Gregory Oudheusden; Emily R Winslow; Sharon M Weber Journal: Ann Surg Oncol Date: 2013-04-18 Impact factor: 5.344