Abstract
Background and aims
Accurate evaluation of intraductal papillary mucinous neoplasm (IPMN) is necessary to inform clinical decision-making. But it is still difficult to distinguish benign and malignant IPMN preoperatively. This study aims to evaluate the utility of EUS to predict the pathology of IPMN.
Methods
Patients with IPMN who underwent endoscopic ultrasound within 3 months before surgery were collected from six centers. Logistic regression model and random forest model were used to determine risk factors associated with malignant IPMN. In both models, 70% and 30% of patients were randomly assigned to the exploratory group and validation group, respectively. Sensitivity, specificity, and ROC were used in model assessment.
Results
Of the 115 patients, 56 (48.7%) had low-grade dysplasia (LGD), 25 (21.7%) had high-grade dysplasia (HGD), and 34 (29.6%) had invasive cancer (IC). Smoking history (OR = 6.95, 95%CI: 1.98–24.44, p = 0.002), lymphadenopathy (OR = 7.91, 95%CI: 1.60–39.07, p = 0.011), MPD > 7 mm (OR = 4.75, 95%CI: 1.56–14.47, p = 0.006) and mural nodules > 5 mm (OR = 8.79, 95%CI: 2.40–32.24, p = 0.001) were independent risk factors predicting malignant IPMN according to the logistic regression model. The sensitivity, specificity, and AUC were 0.895, 0.571, and 0.795 in the validation group. In the random forest model, the sensitivity, specificity, and AUC were 0.722, 0.823, and 0.773, respectively. In patients with mural nodules, random forest model could reach a sensitivity of 0.905 and a specificity of 0.900.
Conclusions
Using random forest model based on EUS data is effective to differentiate benign and malignant IPMN in this cohort, especially in patients with mural nodules.
Similar content being viewed by others
Explore related subjects
Discover the latest articles, news and stories from top researchers in related subjects.Avoid common mistakes on your manuscript.
Despite decades of research, pancreatic ductal adenocarcinoma (PDAC) remains the most aggressive solid malignancy with a 5-year survival of only 9% [1]. Intraductal papillary mucinous neoplasm (IPMN) is a well-documented precancerous lesion of PDAC. Probably due to the growing use of imaging examinations, patients with pancreatic cysts are increasingly detected, and IPMN accounted for 25%–38% of those cysts [2,3,4,5]. Histopathologically, IPMN was classified into low-grade dysplasia (LGD), high-grade dysplasia (HGD), and invasive carcinoma (IC) according to the Baltimore Consensus [6]. Different histological types of IPMN differ dramatically in prognosis and require different clinical management. It is reported that IPMN has an overall risk of developing PDAC of 2.8%, but high-risk IPMN (with dilated main pancreatic duct (MPD) or mural nodules) had a 5-year malignancy risk of 9.77% [7, 8]. Some clinical features and cysts features are associated with malignant IPMN, but it remains challenging to accurately categorize IPMN before surgery [9].
The revised Fukuoka Guideline recommended that IPMN with “high-risk stigmata” were indicated for surgical resection, while IPMN with “worrisome features” should undergo further evaluations, especially endoscopic ultrasonography (EUS), to determine the optimal therapeutic regimen [9]. Similarly, the European Guideline and American Gastroenterological Association (AGA) Guideline recommended surgery for those patients with positive cytology on endoscopic ultrasound-guided fine-needle aspiration (EUS-FNA), a solid component (mural nodule) or a dilated MPD [10, 11]. All three abovementioned guidelines emphasized the pivotal role of EUS in the evaluation of IPMN. The sensitivity and specificity of EUS ± FNA to diagnose malignant IPMN was 67.0%–75.6% and 70.0%–94.1%, respectively [12,13,14].
However, the preoperative detection of malignant IPMN and the surgical indication are still controversial. Most recommendations from current guidelines are based on low-quality evidence and expert’s consensus, and disagreement is common. For instance, MPD ≥ 10 mm is recommended in the 2012 International Guideline as an indication for surgery, but in the 2013 European Guideline the cutoff value of MPD width is 6 mm [9, 15]. The cancer risk in IPMN patients with an MPD diameter of 5-10 mm remained controversial until Hackert et al. reported that it did bear a significant risk of malignancy, and he suggested that surgical treatment should be considered when MPD > 5 mm [16]. Plus, a recent systematic review by Wu et al. suggested that MPD ≥ 5 mm in IPMN could be a sign of malignancy, and pancreatectomy is indicated for some patients [17]. Most existing studies in this field were single-center cohort without a validation group. Therefore, we conducted this multicenter study to evaluate the effectiveness of preoperative EUS in predicting malignant IPMN and determine the proper indication for surgical resection.
Methods
Patients population
In this retrospective multicenter study, we enrolled 115 patients from 6 medical centers who were diagnosed with IPMN by post-surgical pathology from January 2008 to October 2019. All the patients underwent EUS within 3 months before surgical treatment. Patients who had a pancreatic operation history were excluded in our study. Surgical pathology reports were reviewed by pathologists specialized in pancreatic diseases from each center who were blinded to clinical features and EUS findings. All the pathologists were uniformly trained and the pathological diagnosis was based on the Baltimore Consensus [6]. LGD was defined as benign IPMN, HGD and IC were defined as malignant IPMN in the present study. The Ethics Committee of Peking Union Medical College Hospital approved this study (ID: S-K937).
Endoscopy techniques
Endoscopists with more than five years of experience performed EUS using the radial array echoendoscope technique (GF-UE260, 6/7.5 MHz, Olympus, Tokyo, Japan) that was connected to an endoscopic ultrasonic observation unit (EU-M2000, Olympus; EU-ME2 Premier plus, Olympus; Prosound F75, Aloka). De-aerated water was instilled to improve transmission of the US beam. During procedures, a radial endosonoscope was used to observe lesions, with endosono-staging starting from the default frequency of 7.5 MHz until satisfied imaging is obtained. A standard set of EUS images with pathological diagnosis were used to train and test endoscopists from each center. After passing the test, endoscopists from each center reviewed all EUS images and reports to extract valuable data and disagreements were solved by discussion with the senior authors (WX and YAM). All the endoscopists were blinded to the pathological classifications of the lesions. Description of pancreas morphology, cyst lesions, MPD, mural nodules as well as peripheral lymph nodes was recorded. The width of MPD was measured at the most dilated part of the main duct and size of cysts and mural nodules were defined as the largest diameter of the lesions.
Statistical analysis
Continuous variables are exhibited as means and ranges and were compared using the Independent-Samples T-Test. Categorical variables were exhibited as frequency and proportion and compared using the chi-square or Fisher’s exact probability tests. To determine the optimal cutoff value of the width of MPD and the size of mural nodules, Youden index was calculated every variation of 0.5 mm. Univariate analysis and multivariate binary logistic regression models were used to explore the factors that may help distinguish the benign and malignant IPMN. We used simple random sampling provided by SPSS version 25.0 to choose 70% of our patients as an exploratory group to build up a predictive model for malignant IPMN. We draw the ROC curve and calculated the sensitivity and specificity. Then we used the remaining 30% data to validate the efficacy of our model. Data of the validation group was substituted into the logistic model to test our predictive model. To further improve the prediction significance of our model, we used random forest method in our analysis. Age, gender, smoking history, CA19-9, size of cysts, width of MPD, size of mural nodules, pancreatic atrophy, and lymphadenopathy were included in the model. Like the logistic regression model, we randomly chose 70% and 30% of the patients into the exploratory and validation group. Mean decrease accuracy and mean decrease Gini were calculated and dot chart was plotted to exhibit the importance rank of different variables. All reported P values were 2-sided with a value of 0.05. All the statistical analyses in our study were performed using SPSS version 25.0 and RStudio version 1.2.5033 for Windows.
Results
Patients’ characteristics
A total of 115 patients were enrolled in our study. Table 1 shows their demographic characteristics and clinical information. The 115 patients (including 78 males) had a mean age of 59.9 years (range 16–82 years). Abdominal pain was the most common chief complaint, which presented in 63 (54.8%) patients. Only a minority of patients had elevated pancreatic enzymes or tumor markers. Most lesions were MD-IPMN and located in the head of the pancreas. Therefore, the Whipple procedure was mostly used and other procedures included duodenum-preserving pancreatic head resection and subtotal resection of the pancreas. All patients survived surgical treatment but 36 (31.3%) patients developed major post-operative complications including pancreatic fistula (13, 11.3%), intra-abdominal infection (11, 9.6%), delayed gastric emptying (10, 8.7%), postpancreatectomy hemorrhage (8, 7.0%), biliary fistula (6, 5.2%), chyle leak (6, 5.2%), hospital-acquired pneumonia (5, 4.3%), acute kidney injury (4, 3.5%), and deep venous thrombosis (2, 1.7%). Patients with LGD, HGD, or IC were 56 (48.7%), 25 (21.7%), and 34 (29.6%), respectively. Figure 1 shows the typical EUS images of IPMN with different pathological types.
EUS images of IPMN with different pathology type. A Low-grade dysplasia. A cyst without mural nodules communicates with a non-dilated MPD. B High-grade dysplasia. A cystic communicates with dilated MPD without mural nodules. CBD is also dilated. C Invasive carcinoma. A mural nodule inside a dilated MPD. EUS endoscopic ultrasound, IPMN intraductal papillary mucinous neoplasm, MPD main pancreatic duct, CBD common bile duct
Predictors of malignant IPMN based on EUS findings
First, we found that MPD > 7 mm and mural nodules > 5 mm showed the best statistical differentiating performance with a Youden index of 0.320 and 0.307. Then, univariate analysis was performed to find the factors that might contribute to the diagnosis of malignant IPMN (Table 2). We found that proportions of male (p = 0.027), smoking history (p = 0.029), pancreatic atrophy (p = 0.027), lymphadenopathy (p = 0.026), MPD > 7 mm (p = 0.008), and mural nodules > 5 mm (p = 0.011) were significantly different between benign and malignant IPMN. Obstructive jaundice, elevated CA 19–9, and size of cysts did not show any significance. Gender, age, and other candidates screened by univariate analysis were involved in the binary logistic regression models (Table 3). Among them, we found that smoking history (OR = 6.95, 95%CI: 1.98–24.44, p = 0.002), lymphadenopathy (OR = 7.91, 95%CI: 1.60–39.07, p = 0.011), MPD > 7 mm (OR = 4.75, 95%CI: 1.56–14.47, p = 0.006) and mural nodules > 5 mm (OR = 8.79, 95%CI: 2.40–32.24, p = 0.001) were independently related with malignant IPMN. To better verify our results, we draw a ROC curve to evaluate the efficacy of EUS (Fig. 2). The sensitivity and specificity were 0.825 and 0.762, respectively, with an area under the curve (AUC) of 0.841. The remaining 30% data were also tested and we draw another ROC curve to compare the results of the exploratory and validation group (Fig. 3). We found that based on our model, the AUC of the validation group was 0.795. The sensitivity and specificity were 0.895 and 0.571, respectively.
Random forest model
To further evaluate the predictive value of EUS, random forest model was used to distinguish benign and malignant IPMN. The sensitivity and specificity of the validation group were 0.722 and 0.823, respectively, with an AUC of 0.773. The importance rank of variables is shown in Fig. 4. Then, the same model was used in patients with mural nodules (N = 72) and acquired a sensitivity of 0.905 and a specificity of 0.900.
Discussion
In the present multicenter study, we tried to identify the predicting factors associated with malignant IPMN. Several clinical and imaging findings including smoking history, lymphadenopathy, widened MPD and large mural nodules were identified as risk factors for malignant IPMN. Random forest model analysis showed that EUS could accurately diagnose malignant IPMN in patients with mural nodules.
According to our findings, MPD > 7 mm was the best cutoff value for distinguishing benign and malignant IPMN. European guidelines summarized previous studies concerning the relationship between the dilatation of MPD and the risk of malignant IPMN [10]. The cutoff value of the width of MPD ranged from 5 to 8 mm, and a recent study recommended a cutoff value of 6 mm [18]. Different cutoff values could be explained by different endoscope instruments and measurement error. Interestingly, we found that smoking history was one of the risk factors for HGD or IC. Nakagawa et al. demonstrated that in patients with IPMN, current smokers, but not former smokers had a greater chance of having PDAC concomitant with IPMN compared with non-smokers (OR = 4.9, 95% CI: 1.21–23.1, p = 0.03) [19]. Carr et al. found that smokers had a higher risk of early emergence of invasive IPMN, which indicated that cigarette smoking might be an accelerator in IPMN malignant progression [20]. In any case, cigarette quitting should be strongly recommended in patients with IPMN considering its carcinogenic effects. Large prospective epidemiological studies were still needed to verify the relationship between smoking and IPMN. Based on our results, mural nodules > 5 mm showed significance in differentiating benign and malignant IPMN. Some researchers suggested a mural nodule of 10 mm or larger was a predictor for malignancy and should undergo surgical resection [21, 22]. However, most studies supported the 5 mm cutoff value, which was accepted as one of the “high-risk stigmata” in the Fukuoka guidelines [9]. The appearance of lymphadenopathy was also considered a “worrisome feature” for IPMN, which was consistent with our result. Several previous reported “worrisome features” on EUS including thickened cyst walls, pancreatic atrophy and cysts > 3 cm were not significant in our study.
As most researchers have agreed, EUS seems to be a method with high sensitivity but low specificity [9, 23,24,25,26]. In the validation group of our logistic regression model, the sensitivity was high (0.895) but the specificity was low (0.571). Therefore, we tried another statistical method called random forest to optimize our model. Random forest is a machine learning method that can achieve maximum accuracy by systematically constructing multiple decision trees [27]. A recent study indicated that random forest showed better accuracy than logistic regression in most binary classification settings, especially for prediction [28]. In our study, random forest model achieved a modest sensitivity (0.722) and higher specificity (0.823). As known, mural nodules could be found in about 90% HGD and nearly all IC in the resected lesions [9]. However, the detection of mural nodules on images was not robust enough to distinguish benign and malignant IPMN, partially because small mural nodules were easily confused with mucus in the cyst [29]. So, we assumed that the diagnostic efficacy might increase if we combine the high sensitivity of mural nodules and the high specificity of our model. As expected, the sensitivity and specificity were both reached 0.900 in patients with mural nodules. From a clinical perspective, our random forest model can help to predict the pathology of IPMN with mural nodules preoperatively.
Several new methods have been applied to further evaluate IPMN. EUS-FNA might be a reasonable choice given its high specificity [30, 31]. One meta-analysis reported the sensitivity and specificity of EUS-FNA were 0.648(95%CI: 0.44–0.82) and 0.906 (95%CI: 0.81–0.96) [32]. But it was technically demanding to obtain enough tissue. Complications associated with EUS-FNA should also be a concern for clinicians [33]. Contrast-enhanced endoscopic ultrasonography was more sensitive in detecting mural nodules and could distinguish tissues from mucus, but it wasn’t widely used in some centers [34, 35]. A series of other methods such as pancreatoscopy, cyst juice analysis, and detection of k-ras mutation have been recommended in the updated European Guideline [10]. Machine learning and artificial intelligence could also be used in the diagnosis of IPMN. Recently, Kuwahara et al. reported that artificial intelligence via deep learning algorithms reached an accuracy of 0.940 in predicting malignant IPMN in a small group of patients, much higher than human diagnosis [36]. We believe that a more individualized and comprehensive evaluation of IPMN will become the mainstream for preoperative evaluation of IPMN.
Our study has some limitations. First, this is a retrospective study and certain selection bias is inevitable. But we collected data from multiple centers to increase the generalizability of our results. Second, EUS was performed by different endoscopists in different centers, and they were not blinded to the previous examination results. However, all endoscopists have qualified skill, and EUS images were reviewed in each center to minimize the bias. Third, our sample size was relatively limited, especially in the validation group. The predictive model needed to be tested in a separate prospective cohort in future studies. Fourth, branch duct IPMN (BD-IPMN), bearing a much different natural history than main duct IPMN (MD-IPMN) or mixed type IPMN (MT-IPMN), has a much lower risk to develop invasive cancer than the other two types. However, although including BD-IPMN may add unnecessary heterogeneity to the study, a very recent multicenter study enrolling 837 patients with BD-IPMN demonstrated that 168 patients (20%) developed worrisome features/high-risk stigmata, out of which 18 patients (11%) were proved to have high-grade dysplasia or invasive cancer from surgical resection samples [37]. This study indicates that the risk of progression in BD-IPMN is low but not negligible. Besides, the European Guideline and the AGA Guideline differ with regard to the optimal surveillance strategy of BD-IPMN, hinting lack of robust evidence in this field. Therefore, at this juncture we believe that it makes sense to include BD-IPMN patients in this study and will consider restricting the crowds to MD-IPMN or MT-IPMN to produce more homogenous evidence in our future studies. Last but not least, contrast-enhanced EUS has improved efficacy to observe IPMN, particularly mural nodules, and we will evaluate it in future studies.
In conclusion, EUS helped to distinguish benign and malignant IPMN. Random forest predictive model showed high accuracy in IPMN with mural nodules. Novel techniques and statistical method will help clinicians to manage patients with IPMN.
References
Siegel RL, Miller KD, Jemal A (2019) Cancer statistics, 2019. CA Cancer J Clin 69:7–34
Andrejevic-Blant S, Kosmahl M, Sipos B, Kloppel G (2007) Pancreatic intraductal papillary-mucinous neoplasms: a new and evolving entity. Virchows Arch 451:863–869
Sakorafas GH, Smyrniotis V, Reid-Lombardo KM, Sarr MG (2011) Primary pancreatic cystic neoplasms revisited. part III intraductal papillary mucinous neoplasms. Surg Oncol 20:e109-118
Sohn TA, Yeo CJ, Cameron JL, Iacobuzio-Donahue CA, Hruban RH, Lillemoe KD (2001) Intraductal papillary mucinous neoplasms of the pancreas: an increasingly recognized clinicopathologic entity. Ann Surg 234:313–321
Valsangkar NP, Morales-Oyarvide V, Thayer SP, Ferrone CR, Wargo JA, Warshaw AL, Fernandez-del Castillo C (2012) 851 resected cystic tumors of the pancreas: a 33-year experience at the massachusetts general hospital. Surgery 152:S4-12
Basturk O, Hong SM, Wood LD, Adsay NV, Albores-Saavedra J, Biankin AV, Brosens LA, Fukushima N, Goggins M, Hruban RH, Kato Y, Klimstra DS, Kloppel G, Krasinskas A, Longnecker DS, Matthaei H, Offerhaus GJ, Shimizu M, Takaori K, Terris B, Yachida S, Esposito I, Furukawa T, Baltimore Consensus M (2015) A revised classification system and recommendations from the baltimore consensus meeting for neoplastic precursor lesions in the pancreas. Am J Surg Pathol 39:1730–1741
Choi SH, Park SH, Kim KW, Lee JY, Lee SS (2017) Progression of unresected intraductal papillary mucinous neoplasms of the pancreas to cancer: a systematic review and meta-analysis. Clin Gastroenterol Hepatol 15(1509–1520):e1504
Scheiman JM, Hwang JH, Moayyedi P (2015) American gastroenterological association technical review on the diagnosis and management of asymptomatic neoplastic pancreatic cysts. Gastroenterology 148(824–848):e822
Tanaka M, Fernandez-Del Castillo C, Kamisawa T, Jang JY, Levy P, Ohtsuka T, Salvia R, Shimizu Y, Tada M, Wolfgang CL (2017) Revisions of international consensus fukuoka guidelines for the management of IPMN of the pancreas. Pancreatology 17:738–753
European Study Group on Cystic Tumours of the P (2018) European evidence-based guidelines on pancreatic cystic neoplasms. Gut 67:789–804
Singhi AD, Zeh HJ, Brand RE, Nikiforova MN, Chennat JS, Fasanella KE, Khalid A, Papachristou GI, Slivka A, Hogg M, Lee KK, Tsung A, Zureikat AH, McGrath K (2016) American gastroenterological association guidelines are inaccurate in detecting pancreatic cysts with advanced neoplasia: a clinicopathologic study of 225 patients with supporting molecular data. Gastrointest Endosc 83(1107–1117):e1102
Choi SY, Kim JH, Yu MH, Eun HW, Lee HK, Han JK (2017) Diagnostic performance and imaging features for predicting the malignant potential of intraductal papillary mucinous neoplasm of the pancreas: a comparison of EUS, contrast-enhanced CT and MRI. Abdom Radiol (NY) 42:1449–1458
Khashab MA, Kim K, Lennon AM, Shin EJ, Tignor AS, Amateau SK, Singh VK, Wolfgang CL, Hruban RH, Canto MI (2013) Should we do EUS/FNA on patients with pancreatic cysts? the incremental diagnostic yield of EUS over CT/MRI for prediction of cystic neoplasms. Pancreas 42:717–721
Lee KH, Lee SJ, Lee JK, Ryu JK, Kim EY, Kim TH, Moon JH, Lee WJ, Cho YK, Kim JJ (2014) Prediction of malignancy with endoscopic ultrasonography in patients with branch duct-type intraductal papillary mucinous neoplasm. Pancreas 43:1306–1311
Del Chiaro M, Verbeke C, Salvia R, Kloppel G, Werner J, McKay C, Friess H, Manfredi R, Van Cutsem E, Lohr M, Segersvard R, European Study Group on Cystic Tumours of the P (2013) European experts consensus statement on cystic tumours of the pancreas. Dig Liver Dis 45:703–711
Hackert T, Fritz S, Klauss M, Bergmann F, Hinz U, Strobel O, Schneider L, Buchler MW (2015) Main-duct intraductal papillary mucinous neoplasm: high cancer risk in duct diameter of 5 to 9 mm. Ann Surg 262:875–880
Wu YHA, Oba A, Beaty L, Colborn KL, Rodriguez Franco S, Harnke B, Meguid C, Negrini D, Valente R, Ahrendt S, Schulick RD, Del Chiaro M (2021) Ductal dilatation of ≥ 5 mm in intraductal papillary mucinous neoplasm should trigger the consideration for pancreatectomy: a meta-analysis and systematic review of resected cases. Cancers (Basel) 13(9):2031
Ateeb Z, Valente R, Pozzi-Mucelli RM, Malgerud L, Schlieper Y, Rangelova E, Fernandez-Moro C, Lohr JM, Arnelo U, Del Chiaro M (2019) Main pancreatic duct dilation greater than 6 mm is associated with an increased risk of high-grade dysplasia and cancer in IPMN patients. Langenbecks Arch Surg. https://doi.org/10.1007/s00423-018-1740-8
Nakagawa T, Masuda A, Toyama H, Shiomi H, Zen Y, Sofue K, Takenaka M, Kobayashi T, Yagi Y, Yamanaka K, Yoshida M, Arisaka Y, Okabe Y, Kutsumi H, Fukumoto T, Ku Y, Azuma T (2017) Smoking status and the incidence of pancreatic cancer concomitant with intraductal papillary mucinous neoplasm. Pancreas 46:582–588
Carr RA, Roch AM, Shaffer K, Aboudi S, Schmidt CM 2nd, DeWitt J, Ceppa EP, House MG, Zyromski NJ, Nakeeb A, Schmidt CM (2017) Smoking and IPMN malignant progression. Am J Surg 213:494–497
Kawada N, Uehara H, Nagata S, Tsuchishima M, Tsutsumi M, Tomita Y (2016) Mural nodule of 10 mm or larger as predictor of malignancy for intraductal papillary mucinous neoplasm of the pancreas: pathological and radiological evaluations. Pancreatology 16:441–448
Uehara H, Ishikawa O, Katayama K, Kawada N, Ikezawa K, Fukutake N, Takakura R, Takano Y, Tanaka S, Takenaka A (2011) Size of mural nodule as an indicator of surgery for branch duct intraductal papillary mucinous neoplasm of the pancreas during follow-up. J Gastroenterol 46:657–663
Efthymiou A, Podas T, Zacharakis E (2014) Endoscopic ultrasound in the diagnosis of pancreatic intraductal papillary mucinous neoplasms. World J Gastroenterol 20:7785–7793
Jang JY, Park T, Lee S, Kang MJ, Lee SY, Lee KB, Chang YR, Kim SW (2014) Validation of international consensus guidelines for the resection of branch duct-type intraductal papillary mucinous neoplasms. Br J Surg 101:686–692
Kim JH, Eun HW, Park HJ, Hong SS, Kim YJ (2012) Diagnostic performance of MRI and EUS in the differentiation of benign from malignant pancreatic cyst and cyst communication with the main duct. Eur J Radiol 81:2927–2935
Vege SS, Ziring B, Jain R, Moayyedi P, Clinical Guidelines C, American Gastroenterology A (2015) American gastroenterological association institute guideline on the diagnosis and management of asymptomatic neoplastic pancreatic cysts. Gastroenterology 148:819–822
Ho TK (1998) The random subspace method for constructing decision forests. IEEE Trans Pattern Anal Mach Intell 20:832–844
Couronne R, Probst P, Boulesteix AL (2018) Random forest versus logistic regression: a large-scale benchmark experiment. BMC Bioinform 19:270
Muthusamy VR, Chandrasekhara V, Acosta RD, Bruining DH, Chathadi KV, Eloubeidi MA, Faulx AL, Fonkalsrud L, Gurudu SR, Khashab MA, Kothari S, Lightdale JR, Pasha SF, Saltzman JR, Shaukat A, Wang A, Yang J, Cash BD, DeWitt JM (2016) The role of endoscopy in the diagnosis and treatment of cystic pancreatic neoplasms. Gastrointest Endosc 84:1–9
Emerson RE, Randolph ML, Cramer HM (2006) Endoscopic ultrasound-guided fine-needle aspiration cytology diagnosis of intraductal papillary mucinous neoplasm of the pancreas is highly predictive of pancreatic neoplasia. Diagn Cytopathol 34:457–462
Rogart JN, Loren DE, Singu BS, Kowalski TE (2011) Cyst wall puncture and aspiration during EUS-guided fine needle aspiration may increase the diagnostic yield of mucinous cysts of the pancreas. J Clin Gastroenterol 45:164–169
Suzuki R, Thosani N, Annangi S, Guha S, Bhutani MS (2014) Diagnostic yield of EUS-FNA-based cytology distinguishing malignant and benign IPMNs: a systematic review and meta-analysis. Pancreatology 14:380–384
Siddiqui AA, Shahid H, Shah A, Khurana T, Huntington W, Ghumman SS, Loren DE, Kowalski TE, Laique S, Hayat U, Eloubeidi MA (2015) High risk of acute pancreatitis after endoscopic ultrasound-guided fine needle aspiration of side branch intraductal papillary mucinous neoplasms. Endosc Ultrasound 4:109–114
Kitano M, Yamashita Y (2017) New imaging techniques for endoscopic ultrasonography: contrast-enhanced endoscopic ultrasonography. Gastrointest Endosc Clin N Am 27:569–583
Ohno E, Itoh A, Kawashima H, Ishikawa T, Matsubara H, Itoh Y, Nakamura Y, Hiramatsu T, Nakamura M, Miyahara R, Ohmiya N, Ishigami M, Katano Y, Goto H, Hirooka Y (2012) Malignant transformation of branch duct-type intraductal papillary mucinous neoplasms of the pancreas based on contrast-enhanced endoscopic ultrasonography morphological changes: focus on malignant transformation of intraductal papillary mucinous neoplasm itself. Pancreas 41:855–862
Kuwahara T, Hara K, Mizuno N, Okuno N, Matsumoto S, Obata M, Kurita Y, Koda H, Toriyama K, Onishi S, Ishihara M, Tanaka T, Tajika M, Niwa Y (2019) Usefulness of deep learning analysis for the diagnosis of malignancy in intraductal papillary mucinous neoplasms of the pancreas. Clin Transl Gastroenterol 10:1–8
Tamburrino D, de Pretis N, Perez-Cuadrado-Robles E, Uribarri-Gonzalez L, Ateeb Z, Belfiori G, Maisonneuve P, Capurso G, Vanella G, Petrone MC, Arcidiacono PG, Vaalavuo Y, Frulloni L, Dominguez-Munoz JE, Deprez PH, Falconi M, Del Chiaro M, Crippa S, Laukkarinen J (2022) Identification of patients with branch-duct intraductal papillary mucinous neoplasm and very low risk of cancer: multicentre study. Br J Surg 109:617–622
Funding
This study was supported by National Key R&D Program of China (2020YFC2002702), Beijing Natural Science Foundation (No. 7192162), and Chinese Academy of Medical Sciences (2019XK320036).
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Disclosure
Wu Dong, Ding Zhen, Wang Xiaoyan, Cheng Bin, Wang Ruifeng, Qin Shanyu, Li Zhuoran, Song Kai, Wu Wenming, Yang Aiming, and Wu Xi have no conflicts of interest or financial ties to disclose.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Dong, W., Zhen, D., Xiaoyan, W. et al. The effectiveness of endoscopic ultrasonography findings to distinguish benign and malignant intraductal papillary mucinous neoplasm. Surg Endosc 37, 4681–4688 (2023). https://doi.org/10.1007/s00464-022-09752-3
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00464-022-09752-3