Acute pancreatitis may be the first presentation of pancreatic carcinoma (PaCa). The present study was designed to identify clinical findings suggestive of PaCa in patients with nonalcoholic ...nongallstone-related (NANG) acute pancreatitis and evaluate accuracy of endoscopic ultrasound for diagnosing PaCa in this setting.
This is a retrospective analysis of 332 consecutive patients who underwent endoscopic ultrasound-fine-needle aspiration after acute pancreatitis. Patients with gallstones or common bile duct stones, who were heavy or binge alcohol drinkers, or who had post-endoscopic retrograde cholangiopancreatography pancreatitis were excluded.
Among 218 patients with NANG acute pancreatitis, 38 patients had PaCa. Age more than 50 years (P = 0.008), history of smoking (P < 0.001), weight loss of 10 lb or greater (P = 0.003), serum bilirubin levels of higher than 2 mg/dL (P = 0.035) or serum alkaline phosphatase level of higher than 165 U/mL (in patients with normal serum bilirubin levels) (P = 0.003), and radiological findings of an identifiable pancreatic mass (P = 0.001) or distal pancreatic atrophy (P = 0.006) had significant association with an underlying PaCa on multivariate analysis. Of the 38 patients with PaCa in this cohort, 37 had 2 or more of these findings. Endoscopic ultrasound-fine-needle aspiration had 99.5% accuracy (98.6, 100%) for diagnosing carcinoma in this clinical setting.
The clinical criteria defined previously potentially can help select patients with NANG acute pancreatitis with a higher likelihood of an underlying pancreatic neoplasm for further imaging.
Pancreatic duct (PD) dilation proximal to a solid focal pancreatic lesion on computed tomography (CT) scan is considered highly suggestive of pancreatic adenocarcinoma. There is, however, no ...published data on the differential diagnosis of focal non-cystic pancreatic lesions with and without PD dilation. We assessed the diagnostic utility of this radiologic finding.
This is a retrospective analysis of a prospectively maintained database of university-based clinical practice. A total of 445 non-jaundiced patients who underwent endoscopic ultrasound (EUS) (2002-2010) for evaluation of solid pancreatic lesions noted on CT scan were included. Final diagnosis was based on surgical pathology or definitive cytology with supporting clinical follow-up of ≥12 months. Main outcome measurements included (1) differential diagnoses and (2) performance characteristics of EUS-fine needle aspiration (FNA) for diagnosing neoplasm in patients with non-cystic pancreatic lesions with and without PD dilation.
A neoplasm was finally diagnosed in 152 of 187 patients with and 87 of 258 patients without PD dilation on CT scan. Chronic pancreatitis (diffuse and focal) was the predominant non-malignant diagnosis in patients with PD dilation. In patients without PD dilation, malignant lesions included neuroendocrine tumor, adenocarcinoma, metastasis, PEComa (perivascular epitheloid cell tumor), and lymphoma; and the non-neoplastic diagnosis included chronic pancreatitis, intrapancreatic lymph nodes, and infected pancreatic fluid collection. EUS-FNA had 97.6% accuracy for diagnosing a neoplasm in these patients.
Dilation PD proximal to a focal solid pancreatic lesion increases the likelihood of malignancy but the performance characteristics of this radiologic finding are probably inadequate to guide clinical management. Neoplasms without dilated PD often require immunostaining for a definitive diagnosis.
Most recently transformer models became the state of the art in various medical image segmentation tasks and challenges, outperforming most of the conventional deep learning approaches. Picking up on ...that trend, this study aims at applying various transformer models to the highly challenging task of colorectal cancer (CRC) segmentation in CT imaging and assessing how they hold up to the current state-of-the-art convolutional neural network (CNN), the nnUnet. Furthermore, we wanted to investigate the impact of the network size on the resulting accuracies, since transformer models tend to be significantly larger than conventional network architectures.PURPOSEMost recently transformer models became the state of the art in various medical image segmentation tasks and challenges, outperforming most of the conventional deep learning approaches. Picking up on that trend, this study aims at applying various transformer models to the highly challenging task of colorectal cancer (CRC) segmentation in CT imaging and assessing how they hold up to the current state-of-the-art convolutional neural network (CNN), the nnUnet. Furthermore, we wanted to investigate the impact of the network size on the resulting accuracies, since transformer models tend to be significantly larger than conventional network architectures.For this purpose, six different transformer models, with specific architectural advancements and network sizes were implemented alongside the aforementioned nnUnet and were applied to the CRC segmentation task of the medical segmentation decathlon.METHODSFor this purpose, six different transformer models, with specific architectural advancements and network sizes were implemented alongside the aforementioned nnUnet and were applied to the CRC segmentation task of the medical segmentation decathlon.The best results were achieved with the Swin-UNETR, D-Former, and VT-Unet, each transformer models, with a Dice similarity coefficient (DSC) of 0.60, 0.59 and 0.59, respectively. Therefore, the current state-of-the-art CNN, the nnUnet could be outperformed by transformer architectures regarding this task. Furthermore, a comparison with the inter-observer variability (IOV) of approx. 0.64 DSC indicates almost expert-level accuracy. The comparatively low IOV emphasizes the complexity and challenge of CRC segmentation, as well as indicating limitations regarding the achievable segmentation accuracy.RESULTSThe best results were achieved with the Swin-UNETR, D-Former, and VT-Unet, each transformer models, with a Dice similarity coefficient (DSC) of 0.60, 0.59 and 0.59, respectively. Therefore, the current state-of-the-art CNN, the nnUnet could be outperformed by transformer architectures regarding this task. Furthermore, a comparison with the inter-observer variability (IOV) of approx. 0.64 DSC indicates almost expert-level accuracy. The comparatively low IOV emphasizes the complexity and challenge of CRC segmentation, as well as indicating limitations regarding the achievable segmentation accuracy.As a result of this study, transformer models underline their current upward trend in producing state-of-the-art results also for the challenging task of CRC segmentation. However, with ever smaller advances in total accuracies, as demonstrated in this study by the on par performances of multiple network variants, other advantages like efficiency, low computation demands, or ease of adaption to new tasks become more and more relevant.CONCLUSIONAs a result of this study, transformer models underline their current upward trend in producing state-of-the-art results also for the challenging task of CRC segmentation. However, with ever smaller advances in total accuracies, as demonstrated in this study by the on par performances of multiple network variants, other advantages like efficiency, low computation demands, or ease of adaption to new tasks become more and more relevant.
•Joint liver and hepatic lesion segmentation using a hybrid CNN with transformer layers.•Expert-like and state-of-the-art segmentation accuracy on liver MRI and CT data.•Extensive evaluation on MRI ...and LiTS CT data, including ablation and correlation studies.•Besides the proposed network, ten additional state-of-the-art methods were implemented, trained and tested on the same data to ensure direct comparability.
Backgound and Objective: Deep learning-based segmentation of the liver and hepatic lesions therein steadily gains relevance in clinical practice due to the increasing incidence of liver cancer each year. Whereas various network variants with overall promising results in the field of medical image segmentation have been successfully developed over the last years, almost all of them struggle with the challenge of accurately segmenting hepatic lesions in magnetic resonance imaging (MRI). This led to the idea of combining elements of convolutional and transformer-based architectures to overcome the existing limitations. Methods: This work presents a hybrid network called SWTR-Unet, consisting of a pretrained ResNet, transformer blocks as well as a common Unet-style decoder path. This network was primarily applied to single-modality non-contrast-enhanced liver MRI and additionally to the publicly available computed tomography (CT) data of the liver tumor segmentation (LiTS) challenge to verify the applicability on other modalities. For a broader evaluation, multiple state-of-the-art networks were implemented and applied, ensuring direct comparability. Furthermore, correlation analysis and an ablation study were carried out, to investigate various influencing factors on the segmentation accuracy of the presented method. Results: With Dice similarity scores of averaged 98±2% for liver and 81±28% lesion segmentation on the MRI dataset and 97±2% and 79±25%, respectively on the CT dataset, the proposed SWTR-Unet proved to be a precise approach for liver and hepatic lesion segmentation with state-of-the-art results for MRI and competing accuracy in CT imaging. Conclusion: The achieved segmentation accuracy was found to be on par with manually performed expert segmentations as indicated by inter-observer variabilities for liver lesion segmentation. In conclusion, the presented method could save valuable time and resources in clinical practice.
BACKGROUND:In patients with obstructive jaundice and biliary stricture, the role of endoscopic ultrasound-guided fine-needle aspiration (EUS-FNA) is debated for fear of missing a potentially ...resectable pancreatobiliary malignancy (PBM). We evaluated the prevalence of (1) PBM; (2) lesions that do not require a potentially curative cancer surgery; and (3) potentially resectable PBMs in patients with false-negative diagnosis by EUS-FNA.
PATIENTS AND METHODS:This is a retrospective analysis of 342 patients who underwent EUS/EUS-FNA from 2002 to 2009 after presenting with obstructive jaundice and a biliary stricture. Of these, 170 patients had no definitive mass on computed tomography and 172 patients had definitive mass on computed tomography without evidence of unresectability. Final diagnosis was based on surgical pathology or definitive cytology and clinical follow-up of ≥12 months.
RESULTS:The mean age of patients (176 male) was 68.0±12.5 years. A final diagnosis of malignancy was made in only 248 patients (72.5%; 95% confidence interval, 67.7, 77.2). The overall accuracy of EUS-FNA for diagnosing malignancy was 92.4% (89.0, 94.8), with 91.5% sensitivity (87.1, 94.5) and 80.9% negative predictive value (72.0, 87.5). Among 21 patients with false-negative diagnosis, 8 had cholangiocarcinoma (2 resectable), 13 had pancreatic cancer (5 resectable). EUS-FNA provided information to potentially modify surgical management in 116 patients (33.9%; 95% confidence interval, 29.1, 39.0)89 patients diagnosed as true negatives, 24 with distant malignant lymphadenopathy, and 3 with malignant lymphoma.
CONCLUSIONS:In above-defined patient subset, the risk of missing resectable tumors by EUS-FNA has been exaggerated because of artifactually low negative predictive value resulting from a high pretest probability of PBM. The actual miss rate for resectable PBM by EUS-FNA is rather small and was 2% in present cohort. Information from EUS-FNA can potentially modify surgical management in up to one third of patients.
Pancreatic cancer (PaCa) is the fourth leading cause of cancer-related death in the United States. The median size of pancreatic adenocarcinoma at the time of diagnosis is about 31 mm and has not ...changed significantly in last three decades despite major advances in imaging technology that can help diagnose increasingly smaller tumors. This is largely because patients are asymptomatic till late in course of pancreatic cancer or have nonspecific symptoms. Increased awareness of pancreatic cancer amongst the clinicians and knowledge of the available imaging modalities and their optimal use in evaluation of patients suspected to have pancreatic cancer can potentially help in diagnosing more early stage tumors. Another major challenge in the management of patients with pancreatic cancer involves reliable determination of resectability. Only about 10% of pancreatic adenocarcinomas are resectable at the time of diagnosis and would potentially benefit from a R0 surgical resection. The final determination of resectability cannot be made until late during surgical resection. Failure to identify unresectable tumor pre-operatively can result in considerable morbidity and mortality due to an unnecessary surgery. In this review, we review the relative advantages and shortcomings of imaging modalities available for evaluation of patients with suspected pancreatic cancer and for preoperative determination of resectability.
Expert interpretation of anatomical images of the human brain is the central part of neuroradiology. Several machine learning-based techniques have been proposed to assist in the analysis process. ...However, the ML models typically need to be trained to perform a specific task, e.g., brain tumour segmentation or classification. Not only do the corresponding training data require laborious manual annotations, but a wide variety of abnormalities can be present in a human brain MRI — even more than one simultaneously, which renders a representation of all possible anomalies very challenging. Hence, a possible solution is an unsupervised anomaly detection (UAD) system that can learn a data distribution from an unlabelled dataset of healthy subjects and then be applied to detect out-of-distribution samples. Such a technique can then be used to detect anomalies — lesions or abnormalities, for example, brain tumours, without explicitly training the model for that specific pathology. Several Variational Autoencoder (VAE) based techniques have been proposed in the past for this task. Even though they perform very well on controlled artificially simulated anomalies, many of them perform poorly while detecting anomalies in clinical data. This research proposes a compact version of the “context-encoding” VAE (ceVAE) model, combined with pre and post-processing steps, creating a UAD pipeline (StRegA), which is more robust on clinical data and shows its applicability in detecting anomalies such as tumours in brain MRIs. The proposed pipeline achieved a Dice score of 0.642 ± 0.101 while detecting tumours in T2w images of the BraTS dataset and 0.859 ± 0.112 while detecting artificially induced anomalies, while the best performing baseline achieved 0.522 ± 0.135 and 0.783 ± 0.111, respectively.
•Proposes unsupervised anomaly detection pipeline StRegA and compact ceVAE model.•The model is combined the proposed pre- and post-processing steps to form StRegA.•Trained on anomaly-free brain MRI datasets and evaluated for the task of brain tumour detection.•Proposed pipeline achieved a Dice score of 0.642 ± 0.101 while detecting tumours in T2w images of the BraTS dataset.•Achieved a Dice score of 0.859 ± 0.112 while detecting artificially induced anomalies.