⇧ [VIDÉO] You might also like this partner content (after ad)
Pancreatic cancer remains one of the most difficult to treat. In fact, the diagnosis of pancreatic duct adenocarcinoma is delayed by the absence of symptoms and specific diagnostic biomarkers. Developed by Cedars-Sinai researchers, an artificial intelligence tool was trained on CT images taken months or even years before the diagnosis of pancreatic cancer. It was thus able to predict with 86% accuracy which people would develop cancer based on the appearance of the images.
Pancreatic duct adenocarcinoma is a malignant tumor that accounts for more than 90% of pancreatic cancer cases. Currently the fourth leading cause of cancer death, less than 10% of patients live more than five years after being diagnosed or after starting treatment. However, it is possible to increase the survival rate up to 50% when the diagnosis is early and when the complete removal of the tumor is possible, according to recent studies.
The main problem is that there is no easy way to detect pancreatic duct adenocarcinoma at an early stage, as the disease is asymptomatic to begin with. Therefore, 80% of patients are already at an advanced stage of cancer when it is diagnosed.
Early diagnosis of pancreatic duct adenocarcinoma is very rare
Although people with this type of cancer may experience abdominal pain or unexplained weight loss, these symptoms do not alert them to the onset of cancer, as they are common in other diseases. ” There are no single symptoms for an early diagnosis of pancreatic duct adenocarcinoma “, Explain in a press release Stephen J. Pandol, director of basic and translational pancreatic research at Cedars-Sinai Medical Center (Los Angeles) and co-author of the study.
” Most patients with digestive disorders undergo abdominal CT imaging where they are deemed ‘negative’ by a radiologist’s assessment – although some of these patients eventually develop pancreatic cancer write the American researchers. Indeed, it is very difficult to identify “with the naked eye” anomalies of the pancreas on these images. ” Artificial intelligence is the first choice to carry out the modeling of the prediction of several cancers “, add the authors.
AI training can detect early signs of cancer
By reviewing patients’ electronic medical records, the team selected 36 people who met the criteria they were looking for: having been diagnosed with adenocarcinoma of the pancreatic duct within the last 15 years, and having had a CT scan (or computed tomography) with the result ” negative” from six months to three years before diagnosis (prediagnosis). CT images of 36 people who did not develop cancer were also used as a control.
The artificial intelligence (AI) tool developed by the researchers was trained in detecting variations on the surface of the pancreas in pre-diagnostic CT images, compared to control images. The researchers used the naive Bayes classifier algorithm to automatically rank the CT scan images according to the probability of cancer occurrence, knowing that they were particularly targeting the high risk. In the end, the tool achieved an average classification accuracy of 86% across the data set.
” This AI tool was able to capture and quantify very subtle early signs of pancreatic duct adenocarcinoma on CT scans, years before disease onset “, welcomes Debiao Li, professor of biomedical sciences and imaging at Cedars-Sinai and lead author of the study.
On the other hand, the researchers regret the small amount of data, due to the fact that pre-diagnosis scans are rarely available. They hope that AI will reduce the time to diagnosis and thus promote the complete removal of the tumor by surgery. With these encouraging results, they are now working to replicate the model on a larger amount of data, in order to validate it.
Source : Cancer Biomarkers
We would love to say thanks to the author of this write-up for this outstanding content
Pancreatic cancer: artificial intelligence can detect the first signs of the disease
You can view our social media profiles here , as well as other pages on related topics here.https://www.ai-magazine.com/related-pages/