Deep Learning is Being Used by Researchers to Fight Pancreatic Cancer


While there are some forms of cancer that can easily be treated and have a high rate of success (I was diagnosed with stage 1a Hodgkin’s disease roughly three decades ago and treatment was a successful), other forms are not so lucky.  Pancreatic cancer has a much lower rate of success as early detection is vital for a patient’s treatment.  While this has been an ongoing issue, researchers from Johns Hopkins say that using deep learning detection methods on nearly a third of these cases could be identified four to twelve months sooner.


Researchers Use Artificial Intelligence to Combat Pancreatic Cancer

The American Cancer Society states that seven percent of patients after being diagnosed with pancreatic cancer live for five years.  Researcher and radiologist Elliot K. Fishman, MD, at Johns Hopkins is on the front line in attempting to increase this statistic and his weapon of choice is artificial intelligence.

Recently, Bill Siwicki of Healthcare IT News posted an article explaining why using AI in detecting individuals with pancreatic cancer can be accomplished sooner than by a human and expressing that early diagnosis is key to treatment.  Fishman’s goal is to detect pancreatic cancers much sooner than a human can by themselves by adding GPU-accelerated deep learning artificial intelligence to the equation.   

Apparently, Johns Hopkins is well equipped to create a deep learning system as it holds huge amounts of information regarding pancreatic cancer that is needed to instruct a computer to find this type of cancer in a CT scan.  Also, hospital researchers have access to NIVDIA’s DGX-AI Supercomputer.

According to Fishman, detecting minute textural changes within the tissue of an individual’s pancreas and organs that are nearby usually are the first signs of cancer; so, he is assisting to train deep learning algorithms to detect these changes.  This type of detection methods could lead to a diagnosis earlier and roughly a third of the cases he has seen might have been detected sooner by four to twelve months.


Why Early Detection is Important to Treatment

Although many forms of cancer by now has seen patients being cured or a high rate of survival, this is not the case with pancreatic cancer.  Since it can be difficult to early detect the disease, especially in minimally symptomatic patients, this is one of the struggles of dealing with pancreatic cancer; therefore, early detection becomes the goal.

Fishman, who is professor of urology, radiology, oncology and surgery at Johns Hopkins hospital, said that the major treatment for pancreatic cancer for cure is surgery, but unfortunately, because of late detection, no more than 15-20 percent of patients at the time of presentation are surgical candidates.  A minimal goal would be to increase the number of eligible patients for surgical attempt at cure. This is the first challenge. 

Unfortunately, up to thirty percent of physicians who, after reviewing a patient’s previous scans who were eventually diagnosed with the disease, in retrospect can now see that a tumor was present.  Fishman explained that sometimes it is indeed simply by a retrospectoscope that we see the findings and sometimes it was simply just not seen by the initial interpreter.  The goal of using GPU-accelerated deep learning would be to optimize lesion detection so that you can detect every lesion that is present at the earliest time. 

Fishman added that he strongly feels that utilizing deep learning artificial intelligence by Johns Hopkins will teach the computer not only to be on the same level as the best radiologist but to surpass that individual.  Also, he said that to learn and detect just subtle alterations based on pattern and texture instead of simply masses will be one area that Johns Hopkins will focus on with the supercomputers.


Fishman stated that our group has two NVIDIA DGX-1 supercomputers.  The DGX-1 is the state of the art in AI and is necessary for our work. It allows us to review and study the results of hundreds of cases simultaneously and to be able to change parameters, develop algorithms that would be otherwise impossible.  It is these algorithms that allow us to optimize the detection of tumor and to eventually be able to define specificity of tumor type and hopefully also allow us to determine better management strategies.

Yet, one of the many challenges has to do with in cases that it is not suspected that pancreatic cancer exists; therefore, the reader of the scans can easily not pay attention to a subtle tumor.  He said that I think one of the important things to recognize is that the detection of a tumor of the pancreas especially when the primary requisition is not ‘rule out pancreatic mass’ is often overlooked.  One of the things we are relying on the computer to do is do the optimal examination in every abdominal CT and look carefully at the pancreas whether or not this is the zone of suspected pathology.

One of the goals of the project is the ability to optimize the analysis and optimize the interpretation of each study; the researchers believe they can succeed.  Fishman stated that in an era of increasing study volumes where the radiologist spends less time on an individual case, there is no doubt that error rates are increasing.  We believe that we can decrease error, increase detection and change the outcome for many of our patients. The goal of this project is nothing less than changing the trajectory of treatment of pancreatic cancer and ultimate patient survival.