1
|
Chen T, Zhang D, Chen S, Lu J, Guo Q, Cai S, Yang H, Wang R, Hu Z, Chen Y. Machine learning for differentiating between pancreatobiliary-type and intestinal-type periampullary carcinomas based on CT imaging and clinical findings. Abdom Radiol (NY) 2024; 49:748-761. [PMID: 38236405 PMCID: PMC10909762 DOI: 10.1007/s00261-023-04151-1] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 08/16/2023] [Revised: 12/03/2023] [Accepted: 12/05/2023] [Indexed: 01/19/2024]
Abstract
PURPOSE To develop a diagnostic model for distinguishing pancreatobiliary-type and intestinal-type periampullary adenocarcinomas using preoperative contrast-enhanced computed tomography (CT) findings combined with clinical characteristics. METHODS This retrospective study included 140 patients with periampullary adenocarcinoma who underwent preoperative enhanced CT, including pancreaticobiliary (N = 100) and intestinal (N = 40) types. They were randomly assigned to the training or internal validation set in an 8:2 ratio. Additionally, an independent external cohort of 28 patients was enrolled. Various CT features of the periampullary region were evaluated and data from clinical and laboratory tests were collected. Five machine learning classifiers were developed to identify the histologic type of periampullary adenocarcinoma, including logistic regression, random forest, multi-layer perceptron, light gradient boosting, and eXtreme gradient boosting (XGBoost). RESULTS All machine learning classifiers except multi-layer perceptron used achieved good performance in distinguishing pancreatobiliary-type and intestinal-type adenocarcinomas, with the area under the curve (AUC) ranging from 0.75 to 0.98. The AUC values of the XGBoost classifier in the training set, internal validation set and external validation set are 0.98, 0.89 and 0.84 respectively. The enhancement degree of tumor, the growth pattern of tumor, and carbohydrate antigen 19-9 were the most important factors in the model. CONCLUSION Machine learning models combining CT with clinical features can serve as a noninvasive tool to differentiate the histological subtypes of periampullary adenocarcinoma, in particular using the XGBoost classifier.
Collapse
Affiliation(s)
- Tao Chen
- Department of Radiology, The First Affiliated Hospital, Zhejiang University School of Medicine, 79 Qingchun Road, Hangzhou, 310003, Zhejiang, China
| | - Danbin Zhang
- Department of Radiology, The First Affiliated Hospital, Zhejiang University School of Medicine, 79 Qingchun Road, Hangzhou, 310003, Zhejiang, China
| | - Shaoqing Chen
- Department of Radiology, The Second Affiliated Hospital and Yuying Children's Hospital of Wenzhou Medical University, 109 Xueyuan West Road, Wenzhou, 325027, Zhejiang, China
| | - Juan Lu
- Department of Computer Science and Software Engineering, The University of Western Australia, Crawley, WA, 6009, Australia
- School of Medicine, The University of Western Australia, Crawley, WA, 6009, Australia
- Harry Perkins Institute of Medical Research, Murdoch, WA, 6150, Australia
| | - Qinger Guo
- Department of Radiology, The First Affiliated Hospital, Zhejiang University School of Medicine, 79 Qingchun Road, Hangzhou, 310003, Zhejiang, China
| | - Shuyang Cai
- Department of Radiology, The First Affiliated Hospital, Zhejiang University School of Medicine, 79 Qingchun Road, Hangzhou, 310003, Zhejiang, China
| | - Hong Yang
- Department of Radiology, The First Affiliated Hospital, Zhejiang University School of Medicine, 79 Qingchun Road, Hangzhou, 310003, Zhejiang, China
| | - Ruixuan Wang
- School of Electronics and Computer Science, University of Liverpool, Brownlow Hill, Liverpool, Merseyside, L69 3BX, UK
| | - Ziyao Hu
- School of Electronics and Computer Science, University of Liverpool, Brownlow Hill, Liverpool, Merseyside, L69 3BX, UK
| | - Yang Chen
- Liverpool Centre for Cardiovascular Science at University of Liverpool, Liverpool John Moores University and Liverpool Heart and Chest Hospital, William Henry Duncan Building, 6 West Derby St, Liverpool, Merseyside, L7 8TX, UK.
| |
Collapse
|
2
|
Nalbant MO, Oner O, Akinci O, Hocaoglu E, Inci E. Analysis of Pancreatobiliary and Intestinal Type Periampullary Carcinomas Using Volumetric Apparent Diffusion Coefficient Histograms. Acad Radiol 2023; 30 Suppl 1:S238-S245. [PMID: 37211479 DOI: 10.1016/j.acra.2023.04.031] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 04/02/2023] [Revised: 04/21/2023] [Accepted: 04/22/2023] [Indexed: 05/23/2023]
Abstract
RATIONALE AND OBJECTIVES Magnetic resonance imaging plays an important role in the evaluation of patients with known or suspected periampullary masses. The utilization of volumetric apparent diffusion coefficient (ADC) histogram evaluation for the entire lesion eradicates the potential for subjectivity in the region of interest placement, thus guaranteeing the accuracy of computation and repeatability. PURPOSE To investigate the value of volumetric ADC histogram analysis in the differentiation of intestinal-type (IPAC) and pancreatobiliary-type periampullary adenocarcinomas (PPAC). MATERIALS AND METHODS This retrospective study included 69 patients with histopathologically confirmed periampullary adenocarcinoma (54 PPAC and 15 IPAC). Diffusion-weighted imaging was obtained at b values of 1000 mm²/s. The histogram parameters of ADC values, comprising the mean, minimum, maximum, 5th, 10th, 25th, 50th, 75th, 90th, and 95th percentiles, as well as skewness, kurtosis, and variance, were calculated independently by two radiologists. Using the interclass correlation coefficient, the interobserver agreement was evaluated. RESULTS The ADC parameters for the PPAC group were all lower than those of the IPAC group. The PPAC group had higher variance, skewness, and kurtosis than the IPAC group. However, the difference between the kurtosis (P = .003), the 5th (P = .032), 10th (P = .043), and 25th (P = .037) percentiles of ADC values was statistically significant. The area under the curve (AUC) of the kurtosis was the highest (AUC=0.752; cut-off value=-0.235; sensitivity=61.1%; specificity=80.0%). CONCLUSION Volumetric ADC histogram analysis with b values of 1000 mm²/s can discriminate subtypes noninvasively before surgery.
Collapse
Affiliation(s)
- Mustafa Orhan Nalbant
- University of Health Sciences, Bakirkoy Dr. Sadi Konuk Training and Research Hospital, Radiology Department, Tevfik Saglam Cad. No: 11, Zuhuratbaba, 34147 Bakırkoy, Istanbul, Turkey.
| | - Ozkan Oner
- University of Health Sciences, Bakirkoy Dr. Sadi Konuk Training and Research Hospital, Radiology Department, Tevfik Saglam Cad. No: 11, Zuhuratbaba, 34147 Bakırkoy, Istanbul, Turkey
| | - Ozlem Akinci
- University of Health Sciences, Bakirkoy Dr. Sadi Konuk Training and Research Hospital, Radiology Department, Tevfik Saglam Cad. No: 11, Zuhuratbaba, 34147 Bakırkoy, Istanbul, Turkey
| | - Elif Hocaoglu
- University of Health Sciences, Bakirkoy Dr. Sadi Konuk Training and Research Hospital, Radiology Department, Tevfik Saglam Cad. No: 11, Zuhuratbaba, 34147 Bakırkoy, Istanbul, Turkey
| | - Ercan Inci
- University of Health Sciences, Bakirkoy Dr. Sadi Konuk Training and Research Hospital, Radiology Department, Tevfik Saglam Cad. No: 11, Zuhuratbaba, 34147 Bakırkoy, Istanbul, Turkey
| |
Collapse
|