The Value of Deep Learning in Gallbladder Lesion Characterization.

BACKGROUND: The similarity of gallbladder cancer and benign gallbladder lesions brings challenges to diagnosing gallbladder cancer (GBC). This study investigated whether a convolutional neural network (CNN) could adequately differentiate GBC from benign gallbladder diseases, and whether information from adjacent liver parenchyma could improve its performance. METHODS: Consecutive patients referred to our hospital with suspicious gallbladder lesions with histopathological diagnosis confirmation and available contrast-enhanced portal venous phase CT scans were retrospectively selected. A CT-based CNN was trained once on gallbladder only and once on gallbladder including a 2 cm adjacent liver parenchyma. The best-performing classifier was combined with the diagnostic results based on radiological visual analysis. RESULTS: A total of 127 patients were included in the study: 83 patients with benign gallbladder lesions and 44 with gallbladder cancer. The CNN trained on the gallbladder including adjacent liver parenchyma achieved the best performance with an AUC of 0.81 (95% CI 0.71-0.92), being >10% better than the CNN trained on only the gallbladder (p = 0.09). Combining the CNN with radiological visual interpretation did not improve the differentiation between GBC and benign gallbladder diseases. CONCLUSIONS: The CT-based CNN shows promising ability to differentiate gallbladder cancer from benign gallbladder lesions. In addition, the liver parenchyma adjacent to the gallbladder seems to provide additional information, thereby improving the CNN's performance for gallbladder lesion characterization. However, these findings should be confirmed in larger multicenter studies.

Optimal radiological gallbladder lesion characterization by combining visual assessment with CT-based radiomics.

OBJECTIVES: Differentiating benign gallbladder diseases from gallbladder cancer (GBC) remains a radiological challenge because they can appear very similar on imaging. This study aimed at investigating whether CT-based radiomic features of suspicious gallbladder lesions analyzed by machine learning algorithms could adequately discriminate benign gallbladder disease from GBC. In addition, the added value of machine learning models to radiological visual CT-scan interpretation was assessed. METHODS: Patients were retrospectively selected based on confirmed histopathological diagnosis and available contrast-enhanced portal venous phase CT-scan. The radiomic features were extracted from the entire gallbladder, then further analyzed by machine learning classifiers based on Lasso regression, Ridge regression, and XG Boosting. The results of the best-performing classifier were combined with radiological visual CT diagnosis and then compared with radiological visual CT assessment alone. RESULTS: In total, 127 patients were included: 83 patients with benign gallbladder lesions and 44 patients with GBC. Among all machine learning classifiers, XG boosting achieved the best AUC of 0.81 (95% CI 0.72-0.91) and the highest accuracy rate of 73% (95% CI 65-80%). When combining radiological visual interpretation and predictions of the XG boosting classifier, the highest diagnostic performance was achieved with an AUC of 0.98 (95% CI 0.96-1.00), a sensitivity of 91% (95% CI 86-100%), a specificity of 93% (95% CI 90-100%), and an accuracy of 92% (95% CI 90-100%). CONCLUSIONS: Machine learning analysis of CT-based radiomic features shows promising results in discriminating benign from malignant gallbladder disease. Combining CT-based radiomic analysis and radiological visual interpretation provided the most optimal strategy for GBC and benign gallbladder disease differentiation. KEY POINTS: Radiomic-based machine learning algorithms are able to differentiate benign gallbladder disease from gallbladder cancer. Combining machine learning algorithms with a radiological visual interpretation of gallbladder lesions at CT increases the specificity, compared to visual interpretation alone, from 73 to 93% and the accuracy from 85 to 92%. Combined use of machine learning algorithms and radiological visual assessment seems the most optimal strategy for GBC and benign gallbladder disease differentiation.

Intra-observer agreements in multidisciplinary team assessments of pancreatic cancer patients.

BACKGROUND AND METHODS: Treatment strategies for pancreatic cancer patients are made by a multidisciplinary team (MDT) board. We aimed to assess intra-observer variance at MDT boards. Participating units staged, assessed resectability, and made treatment allocations for the same patients as they did two years earlier. We disseminated clinical information and CT images of pancreatic cancer patients judged by one MDT board to have nonmetastatic pancreatic cancer to the participating units. All units were asked to re-assess the TNM stage, resectability, and treatment allocation for each patient. To assess intra-observer variance, we computed %-agreements for each participating unit, defined as low (<50%), moderate (50%-75%), and high (>75%) agreement. RESULTS: Eighteen patients were re-assessed by six MDT boards. The overall agreement was moderate for TNM-stage (ranging from 50%-70%) and resectability assessment (53%) but low for treatment allocation (46%). Agreement on resectability assessments was low to moderate. Findings were similar but more pronounced for treatment allocation. We observed a shift in treatment strategy towards increasing use of neoadjuvant chemotherapy, particularly in patients with borderline resectable and locally advanced tumors. CONCLUSIONS: We found substantial intra-observer agreement variations across six different MDT boards of 18 pancreatic cancer patients with two years between the first and second assessment.

Diagnostic performance of preoperative CT in differentiating between benign and malignant origin of suspicious gallbladder lesions.

PURPOSE: To determine diagnostic performance of preoperative CT in differentiating between benign and malignant suspicious gallbladder lesions and to develop a preoperative risk score. METHOD: All patients referred between January 2007 and September 2018 for suspicion of gallbladder cancer (GBC) or incidentally found GBC were retrospectively analyzed. Patients were excluded when preoperative CT or histopathologic examination was lacking. Two radiologists, blinded to histopathology results, independently reviewed CT images to differentiate benign disease from GBC. Multivariable analysis and internal validation were used to develop a risk score for GBC. Model discrimination, calibration, and diagnostic performance were assessed. RESULTS: In total, 118 patients with 39 malignant (33 %) and 79 benign (67 %) lesions were included. Sensitivity of CT for diagnosing GBC was 90 % (95 % confidence interval [CI]: 76-97). Specificity rates were 61 % (95 % CI: 49-72) and 59 % (95 % CI: 48-70). Three predictors of GBC (irregular lesion aspect, absence of fat stranding, and locoregional lymphadenopathy) were included in the risk score ranging from -1 to 4. Adequate performance was found (AUC: 0.79, calibration slope: 0.89). In patients allocated >0 points, the model showed higher performance in excluding GBC than the radiologists (sensitivity 92 % [95 % CI: 79-98]). Moreover, when allocated >3 points, the risk score was superior in diagnosing GBC (specificity 99 % [95 % CI: 93-100]). CONCLUSIONS: Sensitivity rates of CT for differentiation between benign and malignant gallbladder lesions are high, however specificity rates are relatively low. The proposed risk score may facilitate differentiation between benign and malignant suspicious gallbladder lesions.

Hypothermic Machine Perfusion in Liver Transplantation - A Randomized Trial.

BACKGROUND: Transplantation of livers obtained from donors after circulatory death is associated with an increased risk of nonanastomotic biliary strictures. Hypothermic oxygenated machine perfusion of livers may reduce the incidence of biliary complications, but data from prospective, controlled studies are limited. METHODS: In this multicenter, controlled trial, we randomly assigned patients who were undergoing transplantation of a liver obtained from a donor after circulatory death to receive that liver either after hypothermic oxygenated machine perfusion (machine-perfusion group) or after conventional static cold storage alone (control group). The primary end point was the incidence of nonanastomotic biliary strictures within 6 months after transplantation. Secondary end points included other graft-related and general complications. RESULTS: A total of 160 patients were enrolled, of whom 78 received a machine-perfused liver and 78 received a liver after static cold storage only (4 patients did not receive a liver in this trial). Nonanastomotic biliary strictures occurred in 6% of the patients in the machine-perfusion group and in 18% of those in the control group (risk ratio, 0.36; 95% confidence interval [CI], 0.14 to 0.94; P = 0.03). Postreperfusion syndrome occurred in 12% of the recipients of a machine-perfused liver and in 27% of those in the control group (risk ratio, 0.43; 95% CI, 0.20 to 0.91). Early allograft dysfunction occurred in 26% of the machine-perfused livers, as compared with 40% of control livers (risk ratio, 0.61; 95% CI, 0.39 to 0.96). The cumulative number of treatments for nonanastomotic biliary strictures was lower by a factor of almost 4 after machine perfusion, as compared with control. The incidence of adverse events was similar in the two groups. CONCLUSIONS: Hypothermic oxygenated machine perfusion led to a lower risk of nonanastomotic biliary strictures following the transplantation of livers obtained from donors after circulatory death than conventional static cold storage. (Funded by Fonds NutsOhra; DHOPE-DCD ClinicalTrials.gov number, NCT02584283.).

Liver volumetry plug and play: do it yourself with ImageJ.

BACKGROUND: A small remnant liver volume is an important risk factor for posthepatectomy liver failure and can be predicted accurately by computed tomography (CT) volumetry using radiologic image analysis software. Unfortunately, this software is expensive and usually requires support by a radiologist. ImageJ is a freely downloadable image analysis software package developed by the National Institute of Health (NIH) and brings liver volumetry to the surgeon's desktop. We aimed to assess the accuracy of ImageJ for hepatic CT volumetry. METHODS: ImageJ was downloaded from http://www.rsb.info.nih.gov/ij/ . Preoperative CT scans of 15 patients who underwent liver resection for colorectal cancer liver metastases were retrospectively analyzed. Scans were opened in ImageJ; and the liver, all metastases, and the intended parenchymal transection line were manually outlined on each slice. The area of each selected region, metastasis, resection specimen, and remnant liver was multiplied by the slice thickness to calculate volume. Volumes of virtual liver resection specimens measured with ImageJ were compared with specimen weights and calculated volumes obtained during pathology examination after resection. RESULTS: There was an excellent correlation between the volumes calculated with ImageJ and the actual measured weights of the resection specimens (r(2) = 0.98, p < 0.0001). The weight/volume ratio amounted to 0.88 +/- 0.04 (standard error) and was in agreement with our earlier findings using CT-linked radiologic software. CONCLUSION: ImageJ can be used for accurate hepatic CT volumetry on a personal computer. This application brings CT volumetry to the surgeon's desktop at no expense and is particularly useful in cases of tertiary referred patients, who already have a proper CT scan on CD-ROM from the referring institution. Most likely the discrepancy between volume and weight results from exsanguination of the liver after resection.