AI Classifies Pancreatic Cancer Subtypes Using Histopathology Images
AI Breakthrough: Deep Learning Model Accurately Classifies Pancreatic Cancer Subtypes from Routine Pathology Slides
New technology could revolutionize personalized treatment for aggressive disease
Pancreatic ductal adenocarcinoma (PDAC), the most common form of pancreatic cancer, is a devastating disease with a grim prognosis. Now, researchers have developed a groundbreaking deep learning model that can accurately classify PDAC subtypes directly from routine pathology slides, offering a faster, more cost-effective alternative to current molecular testing methods. This innovative approach, published in The American Journal of Pathology, holds immense promise for improving personalized treatment strategies and patient outcomes.
PDAC has tragically surpassed breast cancer as the third leading cause of cancer-related deaths in the United States and Canada. While surgery can be curative for approximately 20% of cases detected early, the five-year survival rate remains a dismal 20%. The aggressive nature of PDAC often leads to metastatic disease at diagnosis,with most patients succumbing to the disease within a year.
“More and more potentially actionable subtypes to personalize treatment for pancreatic cancer patients are being discovered,” explains Dr. David Schaeffer, co-lead investigator and a pathologist at the University of British Columbia. “However,the current subtyping relies entirely on genomic methodology using DNA and RNA extracted from tissue. This method is excellent if sufficient tissue is available, but that’s not always the case for PDAC tumors due to the challenging location of the pancreas.”
The new deep learning model addresses this challenge by leveraging the power of artificial intelligence to analyze widely available and cost-effective hematoxylin and eosin (H&E) stained slides.These slides are routinely used in pathology laboratories for diagnostics and prognostication, offering a readily accessible source of information.
The researchers trained their AI models on whole-slide pathology images to identify two key PDAC subtypes: basal-like and classical. The models were trained on 97 slides from The Cancer Genome Atlas (TCGA) and tested on 110 slides from 44 patients in a local cohort.
The results were impressive. The best-performing model achieved an accuracy of 96.19% in identifying the classical and basal subtypes in the TCGA dataset and 83.03% on the local cohort, demonstrating its robustness across different datasets.”The sensitivity and specificity of the model were 85% and 100%, respectively,” notes Dr. Ali Bashashati, co-lead investigator and a biomedical engineer at the University of British Columbia. “This makes this AI tool a highly applicable tool for triaging patients for molecular testing. Importantly, the AI model was able to detect the subtypes from biopsy images, making it a highly useful tool that can be deployed at the time of diagnosis.”
This AI-powered approach offers a significant advancement in pancreatic cancer diagnostics. By enabling rapid and cost-effective identification of key molecular subtypes, it paves the way for more personalized treatment strategies and potentially improved outcomes for patients facing this aggressive disease.
Q&A: AI’s Potential in Classifying Pancreatic Cancer Subtypes
NewsDirectory3.com: Dr. Schaeffer, can you elaborate on the significance of classifying pancreatic ductal adenocarcinoma (PDAC) subtypes and why this new AI model is so groundbreaking?
Dr. David Schaeffer: Classifying PDAC subtypes is crucial becuase it allows us to tailor treatment plans for individual patients. Different subtypes respond differently to therapies. Traditionally, we’ve relied on genomic testing, which can be expensive and time-consuming, especially considering the limited tissue frequently enough available from pancreatic tumors.
This AI model is groundbreaking because it uses readily available H&E stained slides, which are routinely used in pathology labs. This eliminates the need for additional tissue sampling and makes subtyping much faster and more cost-effective.
NewsDirectory3.com: Dr. Bashashati, how accurate is the AI model and can it be reliably used in clinical settings?
Dr. Ali Bashashati: Our model demonstrated impressive accuracy, achieving over 96% in identifying classical and basal subtypes in one dataset and over 83% in another independent dataset. Importantly, it was also able to accurately classify subtypes from biopsy images, signifying its potential for early use during diagnosis.
We believe the model’s high sensitivity and specificity, combined with its ability to analyze standard pathology slides, make it a powerful tool for triaging patients for further molecular testing and informing treatment decisions.
NewsDirectory3.com: What are the next steps for this research, and when might we see this technology available for mainstream use?
Dr. Schaeffer: The next step is validating our findings in larger, multicenter studies that encompass a wider range of PDAC subtypes.We also need to explore how this AI model can be integrated into existing pathology workflows and clinical decision-making processes.
While it’s challenging to predict a precise timeline, we are optimistic that this technology could be available for clinical use within the next few years, potentially revolutionizing how we diagnose, treat, and manage pancreatic cancer.