How AI and Deep Learning Will Address the Challenges of Pathologists
According to the National Cancer Institute, approximately 40% of men and women will be diagnosed with cancer at some point during their lifetimes (reference below). Luckily, we are getting better and better at treating cancer and, in many instances, cancer is now regarded more as a chronic disease than deadly, resulting in regular check-ups and even more pathology tests. This is just a couple of the many reasons for the steady growth in the volume of pathology tests conducted globally over the past few years. As the number of pathology samples increases, so too does the burden in pathology labs.
One of the most significant challenges to accommodating the steady growth in test volume is a simple lack of pathologists. In its December 2012 report, the U.S. Department of Health and Human Services projects a need of 23 percent more pathologists in 2020 than the baseline number of pathologists practicing. In addition, the pressure of delivering accurate interpretations of complex pathology samples at faster rates is ever-growing.
However, pathology (and healthcare as a whole) is at the forefront of significant technological change. With technology like artificial intelligence (AI), computers are becoming smarter and have already reached an intelligence level that enables them to help with many of the mundane, day-to-day tasks.
ContextVision, a medical technology company specializing in image analysis and artificial intelligence, has developed world-class skills within deep learning and AI. The company is currently investing heavily into the task of developing new decision support tools that could save time in the diagnosis process, as well as alleviate variations pathologists face daily.
For instance, these new tools will automate many of the time-consuming, routine tasks like counting mitoses or screening lymph nodes for metastasis in breast tissue, and allow pathologists to save time and focus on more advanced tasks. In prostate, approximately two-thirds of biopsies contain only benign tissue. If the software can safely exclude samples with no cancer, the pathologist can easily sort out benign tissue from malignant tissue, drastically cutting time.
In addition, with regards to prostate cancer, additional immunostaining may be needed as the morphological analysis of H&E stained tissue is not sufficient to rule out cancer. Now, their new software aims at automatically suggest Gleason patterns in accordance with ISUP 2014 new classification, cutting out the uncertainly and subjectivity of evaluation. With the use of deep learning algorithms, the expertise of many pathologists and objective data can be combined and conserved. By digitizing pathology, workflow for the individual pathologist will be more efficient and streamlined. Information and slides are readily available, allowing pathologists to easily switch between, share, store or compare slides side-by-side.
Decision support tools can work simultaneously in the background and are flexible enough to integrate with current diagnostic processes and will be fully compatible with most scanners or viewers. As a result, the introduction of decision support tools would save pathologist´s time and in addition ensure that more objective and consistent evaluations are performed.
Although new technologies often have a learning curve, the future of artificial intelligence – and digital pathology – is full of possibilities and the benefits are endless. We can be certain it will be life changing, in the medical field and beyond.
To stay up-to-date on ContextVision’s product-related news, please visit www.contextvision.com.
Reference:
Based on 2010-2012 data : https://www.cancer.gov/about-cancer/understanding/statistics
Source: ContextVision
