Smartphone 'Selfies' May Help Diagnose Pancreatic Cancer

Roxanne Nelson, BSN, RN

August 30, 2017

Selfies taken with smartphones appear throughout social media, but they may have a usefulness that extends beyond vanity.

Researchers from the University of Washington, in Seattle, are in the process of developing a smartphone app that may be able to transform selfies into a method of catching pancreatic cancer in its very early stages, which in turn could greatly improve outcomes.

Jaundice is an early symptom of pancreatic cancer. It causes yellowing of the skin and sclera, owing to a build-up of bilirubin in the blood. It only becomes readily visible after the disease is advanced.

This is where the new app, dubbed BiliScreen, could make a difference: the smartphone app is able to capture pictures of the sclera in a patient's eye and produce an estimate of the bilirubin level, before jaundice is visible.

A clinical study that included 70 healthy volunteers compared findings using the BiliScreen app with results from the blood test that is normally used to assess serum bilirubin. In the study, the app was used in conjunction with a 3-D printed box that controls the eye's exposure to light.

The researchers found that the BiliScreen could identify "cases of concern" with a sensitivity of 89.7% and a specificity of 96.8% when used with the box accessory.

Lead author Alex Mariakakis, a doctoral student at the Paul G. Allen School of Computer Science and Engineering, University of Washington, cautioned that despite the encouraging results, the device is not yet ready for prime time.

"We are not yet at the point where we can distribute the app for diagnostic/screening purposes," he told Medscape Medical News. "We're planning a larger study sometime within the next academic year, with a far larger sample size.

"Our algorithm is machine-learning based, which means that the more examples of different bilirubin levels we have in our dataset, the better," said Mariakakis. "Assuming that goes well, we would then start working with the FDA to figure out how to get their approval ― a process that the FDA is currently going through for the first time with some of our lab's older projects."

He emphasized that, when measuring bilirubin levels, the app uses jaundice as a marker. Although jaundice is is a symptom of pancreatic cancer, it is not pathognomonic for that disease.

"Jaundice also appears in people with hepatitis, Gilbert's syndrome, and alcoholism," explained Mariakakis. "We are particularly interested in pancreatic cancer because we have friends and family who have been affected by the disease, which is one of the reasons for that focus."

The research was published in the Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies and will be presented September 13 at Ubicomp 2017, the Association for Computing Machinery's International Joint Conference on Pervasive and Ubiquitous Computing.

Eyes as a Gateway

"The eyes are a really interesting gateway into the body ― tears can tell you how much glucose you have, sclera can tell you how much bilirubin is in your blood," commented senior author Shwetak Patel, PhD, the Washington Research Foundation Entrepreneurship Endowed Professor in Computer Science and Engineering and Electrical Engineering, in a statement. "Our question was, Could we capture some of these changes that might lead to earlier detection with a selfie?"

The current gold standard for measuring bilirubin is with a blood test to assess total serum bilirubin (TSB) levels. Noninvasive methods of measuring bilirubin, such as use of a transcutaneous bilirubinometer (TcB), have been investigated. However, note the authors, the underlying computations for TcBs are designed for newborns and are not readily applicable to adults. This is partly due to the fact that normal concentrations of bilirubin are much higher in neonates, in whom jaundice is a fairly common occurrence in comparison with adults.

Earlier, the team at the Ubiquitous Computing Lab of the University of Washington developed a smartphone app called the BiliCam, which was designed to detect neonatal jaundice by taking pictures of the skin.

A study published this month in Pediatrics that included 530 infants found that the BiliCam provided accurate estimates of TSB values.

The researchers point out that the sclerae are more sensitive than the skin to changes in bilirubin levels, owing to the fact that elastin in the sclerae has a high affinity for bilirubin. Thus, the researchers explored its potential for use in a noninvasive screening tool that could detect disease at early stages.

This type of device would need to be sufficiently sensitive to measure the range of bilirubin levels exhibited by adults. Jaundice, however, is generally not observed by even a trained eye until bilirubin reaches a level of about 3.0 mg/dL. Levels higher than 1.3 mg/dL are a cause for clinical concern. Therefore, the researchers note, there is a "detection gap" ― bilirubin levels of 1.3 to 3.0 mg/dL will be missed unless the TSB level is assessed, and that is rarely done without sufficient reason.

Capturing Trends and Lighting

Another important consideration is that bilirubin levels vary over time. The authors point out the bilirubin level might exceed normal values at one measurement but then subsequently return to normal, owing to normal variation.

Values that continue on an upward trend are more indicative of an underlying pathologic condition, such as blockage of the common bile duct by a tumor. In addition to diagnosis, trends in levels are also important for gauging treatment efficacy. But trends in bilirubin levels are difficult to capture, the authors point out, because of the need for repeated blood draws, which can be uncomfortable and inconvenient, especially in the outpatient setting.

The BiliScreen app uses the smartphone's built-in camera to collect pictures of the eyes. Representations of the sclerae are then extracted from the image using computer vision. The app calculates the color information from the sclerae, which is based on the wavelengths of light that are being reflected and absorbed, and then correlates it with bilirubin levels using machine-learning algorithms.

Because changes in color can occur through variations in lighting, the Biliscreen was evaluated with two different accessories that help to account for differences in ambient lighting conditions. The first accessory is a head-worn box that simultaneously blocks out ambient lighting and provides controlled internal lighting via the flash from the camera. The second is a pair of paper eyeglasses that are printed with colored squares that facilitate calibration. Slightly better results were achieved when the app was used in conjunction with the box accessory.

The BiliScreen has tighter precision requirements than the Bilicam, which is intended for newborns. Unlike skin color, which varies considerably across different populations, the changes in the sclera are more consistent across all races and ethnicities.

Target Population

"To make this more usable, we are looking at removing the need for an accessory ― glasses or box ― altogether," said Mariakakis. "Our thought on this is that if we can somehow figure out the precise color of an object that is already in the picture, we could use it as a color reference like we do with the glasses right now."

He noted that probably the most reasonable approach for use of the app would be that the app be used at the recommendation of a clinician for patients who are members of a high-risk group.

"That way, there is someone to interpret the results in a more reasonable manner," he said. "We are very aware that this app could go down the slippery slope of worrying hypochondriacs even more about their health than before. Such an app would give people data to support their worries, which is something we do not take lightly."

He added that another use ― one that could have more immediate impact ― would be with regard to disease management. "People who already have pancreatic cancer receive daily blood draws to monitor their bilirubin, particularly after a treatment, and using BiliScreen to get that data noninvasively could prove useful."

The study was funded by the National Science Foundation and the Coulter Foundation and received endowment funds from the Washington Research Foundation.

Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies. 2017;1:art20. Abstract


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.