Deep Learning Aids Diagnosis of Chronic Pancreatitis
A deep learning (DL)–based model achieved a high accuracy in pancreas segmentation for patients with chronic pancreatitis (CP) and healthy individuals, a new study finds. The model showed robust performance across diverse scanning protocols and anatomic variations, although its accuracy was affected by visceral fat area and pancreas volume.
METHODOLOGY:
Researchers developed a DL-based tool using the neural network U-Net (nnU-Net) architecture for the automated segmentation of retrospectively collected CT scans of the pancreas of healthy individuals and of patients with CP.
Scans were obtained from one hospital each in Aalborg (n = 373; 223 patients with CP and 150 healthy individuals) and Bergen (n = 97 patients with CP), along with an online dataset from the National Institutes of Health (NIH; n = 80 healthy individuals).
The tool was validated and tested using internal and external datasets, and its performance was compared with manual processing done by radiologists using the Sørensen-Dice index.
The tool's performance was examined for potential correlation with factors including visceral fat area at the third lumbar level, pancreas volume, and CT scan parameters.
TAKEAWAY:
The tool demonstrated strong performance with mean Sørensen-Dice scores of 0.85 for the Aalborg test dataset, 0.79 for the Bergen dataset, and 0.79 for the NIH dataset.
Sørensen-Dice scores were positively correlated with visceral fat area across datasets (correlation coefficient [r], 0.45; P < .0001) and with pancreas volume in the Aalborg test dataset (r, 0.53; P = .0002).
< .0001) and with pancreas volume in the Aalborg test dataset (r, 0.53; = .0002). CT scan parameters had no significant effect on model performance.
The tool maintained accuracy across diverse anatomic variations, except in cases with severe pancreatic fat infiltration.
IN PRACTICE:
"This study presents a novel AI [artificial intelligence]–based pancreas segmentation model trained on both healthy individuals and CP [chronic pancreatitis] patients, demonstrating consistent and robust performance across internal and external test datasets that vary in patient characteristics and scanner parameters. The model has the potential to significantly enhance the efficiency and accuracy of pancreas segmentation in clinical practice and research, particularly for CP patients with complex anatomical features," the authors wrote.
SOURCE:
This study was led by Surenth Nalliah, Radiology Research Center, Department of Radiology, Aalborg University Hospital, Aalborg, Denmark. It was published online on May 14 in European Journal of Radiology .
LIMITATIONS:
Comprehensive hyperparameter optimisation was not performed due to computational constraints. Additionally, architectures beyond nnU-Net and other segmentation methods were not explored. Post hoc visualisation methods were not studied. Small sizes of datasets could have hindered model performance, and cases of severe pancreatic fat infiltration were not included.
DISCLOSURES:
Funding information was not provided for this study. One author reported receiving financial support from Health Hub, founded by the Spar Nord Foundation.
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Yahoo
2 hours ago
- Yahoo
From droughts to downpours: How AI is revolutionizing extreme weather forecasting
The Brief A new groundbreaking study uncovers ways to use A.I. to prepare for extreme weather. A.I. could even help predict wildfires. Issues remain about the ethical use of this technology, and trust in the data. MILWAUKEE - There is a lot of talk these days about artificial intelligence or A.I., and that is also true in the world of weather forecasting. A.I. is rapidly transforming how we model, detect, and respond to extreme weather and climate events. In a 2025 review published in Nature Communications, a team of researchers explored the full potential of A.I. and its ability to forecast floods, droughts, heatwaves, and even wildfires. The goal was to uncover ways to help communities prepare before disaster strikes. FREE DOWNLOAD: Get breaking news alerts in the FOX LOCAL Mobile app for iOS or Android What we know Machine learning and deep learning A.I. models were found to outperform some traditional forecasting techniques and seemed to excel at identifying and predicting extreme events. That accuracy went up when multiple data sources were integrated into the models, like satellite imagery, climate data, and ground sensors. The study went beyond basic forecasting and used A.I. to explore the "why," "what if," scenarios of different extreme weather events and even asked A.I. "how confident" it was in those predictions. Droughts: Hybrid models were able to help predict impacts on agriculture and forest health. Heatwaves: Specific models were able to give improved forecasts of regional temperature anomalies. Wildfires: Deep learning A.I. was able to provide early detection of dangerous fire events. Floods: A.I. was seen to enhance early warning systems and real-time alerts. The other side Not everything in the study was a net positive. The authors clearly stressed the need for transparent, ethical, and localized AI systems, particularly in high-stakes scenarios where false alarms can erode public trust or mislead disaster responses. In short, A.I. isn't ready to be handed the reins and will still need significant human oversight, but it could be an amazing tool for forecasters in the very near future. There are other significant hurdles as well, such as: Data Limitations: Extreme events, by their very nature, are rare, which makes it hard to train A.I. models. There's also a shortage of data that reflects the diverse geographic and socio-economic realities of our country and our planet. Integration Issues: AI models, in some cases, under-performed traditional methods. Real-world applications came up short due to high uncertainty, and incomplete data. SIGN UP TODAY: Get daily headlines, breaking news emails from FOX6 News What's next To unlock A.I.'s full potential, the study's authors suggest we develop better datasets that are tailored for each type of extreme weather event. They call for greater collaboration across A.I., climate science, and policy disciplines and again stress the need for ethical safeguards, especially for vulnerable communities. The Source Camps-Valls, G., Fernández-Torres, M.Á., Cohrs, K.-H., Höhl, A., Castelletti, A., Pacal, A., et al. (2025). Artificial intelligence for modeling and understanding extreme weather and climate events. Nature Communications, 16, 1919.
Forbes
4 hours ago
- Forbes
COVID-19 Disease Burden Remains High, Study Shows
A recent study has shown that COVID-19 remains a significant threat. Our public health response ... More needs to reflect this continued disease burden. Photographer: Graham Hughes/Bloomberg Researchers in Denmark conducted a large-scale evaluation of the disease burdens associated with COVID-19 and influenza. Their conclusion? COVID-19 has a greater impact than influenza, resulting in more hospitalizations and deaths. These findings, they note, emphasize the continued need for a strong public health response. To compare the disease burden of COVID-19 and influenza, the researchers evaluated data available from Danish health registries from May 16, 2022 to June 7, 2024. The results were published in The Lancet Infectious Diseases. A commentary about the findings was published in the same issue of the journal. During this period, the risk of hospitalization due to COVID-19 was approximately two times higher than the risk of hospitalization due to influenza. The risk of death was approximately three times higher. The researchers note that these increased risks were greater during the first year of the study period but still were evident during the second year. The risk of mortality, for example, was five times greater for COVID-19 during the first year of the study. That increased risk dropped during the second year, but COVID-19 still had a two and a half times greater risk of mortality than influenza. Infectious disease experts often cite the case fatality rate of an infectious agent as one measure of its disease burden. Simply, the CFR represents the proportion of people diagnosed with an infectious disease who die from the disease. Five years since the beginning of the COVID-19 pandemic, roughly 780 million confirmed cases of COVID-19 have been reported worldwide, resulting in over 7 million deaths, according to data provided by the World Health Organization. The ratio in the U.S. is similar. About 103 million cases and 1.22 million deaths have been reported. These numbers indicate a CFR of approximately 1%. This rate is significantly lower than the rates associated with other recent coronavirus outbreaks like SARS (CFR = 11%) and MERS (CFR = 34%). However, the CFR of COVID-19 far exceeds the CFR of seasonal influenza. The disease burden of COVID-19 certainly has dropped over the course of the pandemic. Various factors most likely have contributed to this change. Most people today possess some degree of natural or vaccine-induced immunity. As a result, rates of severe disease, hospitalization, and death should be lower. But the disease burden remains significant. We must continue our efforts to prevent new infections and develop effective treatments for infected individuals. In their article, Bager et al. note, 'These results highlight the continued need for attention and public health efforts to mitigate the impact of SARS-CoV-2.' In the accompanying commentary, Dr. Yan Xie, a researcher at the Washington University School of Medicine, noted that there exists, 'a profound disconnect between empirical evidence and public health policy, particularly in vaccine prioritisation strategies.' He added, 'Governments must prioritize robust surveillance and monitoring systems, data collection, analysis, and clear communication about the state of infectious threats, including COVID-19's ongoing impacts.' When asked to elaborate on these points, Dr. Xie commented, 'The disconnect arises primarily from the notion that COVID-19 has become less severe and in the eyes of many an inconsequential infection.' The empirical evidence, however, suggests otherwise. COVID-19 remains a significant public health problem. Yet, our efforts to address the disease are waning. In recent weeks, the Food and Drug Administration commissioner Dr. Martin A. Makary wrote that future formulations of the COVID-19 vaccine would need to undergo extensive clinical trials before being approved for use in healthy people between 6 months and 64 years and age. And Robert F. Kennedy Jr. announced that the COVID-19 vaccine would be removed from the CDC immunization schedule for healthy children and healthy pregnant women. These changes undoubtedly will reduce access to the vaccines and thus increase people's risk of becoming infected. COVID-19 has not gone away. We must remain vigilante and committed to addressing this ongoing threat. Now is not the time to relax our efforts. As Dr. Xie remarked, 'Aligning policies with reality is crucial to effectively protect public health. Wishful thinking is just that - wishful thinking.'
Digital Trends
4 hours ago
- Digital Trends
A sticky notes app for Safari transformed how I get work done on macOS
Just a few days ago, the Mozilla Foundation announced that Pocket was shutting down. One of the most popular bookmarking and webpage saving tools out there, especially among journalists and researchers, Pocket leaves a gap that will be hard to fill. The absence will be felt deeply because there's no viable alternative that can offer it all in a polished package. To users tied to the Apple Mac ecosystem, they have even fewer choices for a few reasons. The most notable among them all? Safari's save later and bookmarking system. Recommended Videos Why is Safari a laggard? Safari's lack of a rewarding tool, one where you can organize your ideas and save-worthy content, is quite puzzling. All you have are reading lists and bookmarks. It almost feels like a relic in the face of competitors, and even more so when compared to the solutions you find out there from no-name developers and the open-source community. That's one of the core reasons I stick with Collections in Edge and Pinboards in Opera. They are well-designed, offer plenty of organization tricks, and can even be shared. But they are still hidden behind a UI, not something you can have in front of your eyes at all times. An annotated element, highlighted segments, or personal notes you can fix just at the right spot on a webpage is a solution that no utility has managed to offer in a usable fashion. You can find plenty of sticky notes apps that put stuff on your desktop screen, but not on the web destinations where you get work done. This is where Sticky Notes for Safari comes into the picture. Technically, it's a Mac app that lives as an extension in the Safari browser and costs less than a cup of black coffee. It's light, barebones in just the right way, and captures the true spirit of sticky notes positioned on a scrolling digital canvas. How does it work? Sticky Notes for Safari is pretty straightforward. You install it from the Mac App Store, pay the $3 fee, and give it permission to run in Safari. That's it, and you're good to go. The next time you open Safari, it will appear as a tiny sticky notes option to the left of the URL bar. Let's say you are reading a web page, and want to add a sticky note at a particular spot. All you need to do is click on the extension's icon, and you will have a tiny colored box where you can type your word. The sticky notes are freely resizable, so you can position them to your exact liking. Alternatively, you can right-click on any webpage and click on 'Add Sticky Notes' in the action box. Now, when a sticky first appears on a webpage, it's set to a certain text size and paper color. You can, thankfully, change the default for each. You can also change the paper color on the fly and pick from six options. I prefer this simplicity, instead of having something too fancy, such as a color dial or a massive drop-down of color-coded boxes. As for the sticky notes, they are freely movable, and resizable, so you have that flexibility. It solves a realistic problem I believe the best app is the one that solves a tangible problem and doesn't try to cram more features than users actually need. Sticky Notes for Safari takes the former route. You can pin a note at any spot on a webpage, and it stays there. What happens when I close the browser tab? The next time you visit the webpage, the sticky note will be there to greet you. How about closing the browser itself? Not a problem. You see, the note attaches itself to the specific webpage. And even if you clear the browser cache, the color notes stay. What I love most about the app is that it keeps things simple. Beyond the task of creating sticky notes and pinning them to any spot, you also get a neat catalog where you can search through the entire notes history. First, it serves as a neatly organized place where you can find all the sticky notes you have created so far. There's a neat Search feature at the top where you can look through the notes saved on a page using keywords. I've created an unwritten rule for my sticky notes. I prefer red for more pressing or critical pieces of note. Green represents my own personal opinions, while blue is reserved for notes where the words are destined for my journalistic duties (such as sending media queries) tied to the blue DigitalTrends brand logo. Second, there's also a view all section where you find all your notes arranged in chronological order. When you tap on any of the cards, it takes you straight to the web page where it is saved. To save me the chore of scrolling too much, I simply Search (Command + F) on this page, land on the sticky note I was looking for, and with a single click, go straight to the webpage where it lives. In a nutshell, Sticky Notes for Safari combines the idea of bookmarks and reading lists in one go. In fact, I no longer have any other third-party installed on my Mac except this neat tool for my Safari-based workflow. Boiling it to the core Of course, it's not perfect. For example, when you enable Reader Mode in Safari, these sticky notes away are no longer visible. They appear as you return to the regular viewing mode, though. I wish there were a few more text formatting options for text clusters and the text, but personally, I find it to be a fundamental blunder. The point of sticky notes is just quickly writing what's on your mind, and revisiting it at a time of your convenience. For some reason, you can't trigger the system-level font styling and spell check tools, something you can easily do in other apps such as Notes. By default, the app saves everything as plaintext, even if you save stylized text from another app, it will be stripped of all that formatting in the sticky notes. In hindsight, you don't have to worry about copy-pasting heavily formatted content, as the app will do that for you. If you trying to copy an image, it would be pasted as the image's URL link on the sticky note. Finally, there's this little functional overlap with shortcuts. When you hit Command+T in Apple Notes, it opens the font styles. In Safari, by default, that shortcut opens a browser tab. I wish it could borrow some of the shortcuts and UI customization ideas from Antinote, which offers the best note-taking experience I've ever used in an app. But then, there's only so much you can do with a web extension compared to a full-fledged app. But for the sum of its parts, Sticky Notes for Safari does more than it's intended to. It's minimalist, solves a practical browser problem, and then doubles as a beautiful hub that ends the reliance on dedicated or built-in bookmarking tools. I'd say that's a task well done.
