Smartphone app helps screen for pancreatic cancer

Alex Mariakakis
 

Researchers at University of Washington have developed a smartphone app designed to detect jaundice, a common early symptom of pancreatic cancer.

Investigators assessed the use of a novel app called BiliScreen, intended to capture pictures of a person’s eye and then estimate the person’s bilirubin levels.

“Jaundice is only recognizable to the naked eye in severe stages, but a ubiquitous test using computer vision and machine learning can detect milder forms of jaundice,” Alex Mariakakis, doctoral student at Paul G. Allen School of Computer Science and Engineering at University of Washington, and colleagues wrote.

Mariakakis and colleagues accounted for various lighting conditions by using BiliScreen with two different accessories: a 3-D printed box that blocks ambient lighting, and paper glasses printed with colored squares to help calibrate color.

In a cohort of 70 people, BiliScreen — when used with the 3-D box accessory — achieved a Pearson correlation coefficient of 0.89 and a mean error of -0.09 ± 2.76 mg/dL for predicting bilirubin level. Use of the app with the 3-D box accessory resulted in an 89.7% sensitivity and 96.8% specificity.

HemOnc Today spoke with Mariakakis about the app, the early efficacy it has demonstrated, and the potential benefits it may offer, and what must confirmed or improved before it becomes more widely used.

Question: How did this app come about?

Answer: The idea for this app was inspired by another project our lab is working on for an app that detects neonatal jaundice. The observation was that nurses and clinicians were essentially using their eye as a sensor, and then learning an algorithm that basically took the color of the baby’s skin and mapped it to the result of the blood draw that they would do later. The question became, if nurses can do this with some level of accuracy, can we get a smartphone to do this so we can perform this test anywhere with greater accuracy? This project has been going on for about 4 years. It led to the idea for this app for adults, which has been in development for about 2.5 years. It has been suggested that the collagen in the eyes are supposed to be more susceptible to bilirubin, so the thought was that we would be able to see jaundice in adults through the eyes.

 

Q: Can you describe the efficacy of the app so far?

A: We conducted a trial that included 70 adults, 31 of whom were healthy individuals. The other 39 had conditions of the pancreas and liver, but not necessarily pancreatic cancer. These 39 patients had bilirubin levels ranging from borderline levels — 1.3 mg/dL to 3 mg/dL — to very severe levels of 30 mg/dL. We wanted to quantify jaundice as a symptom of different conditions, as not every patient with jaundice has pancreatic cancer. The app was associated with 90% accuracy.

 

Q: What needs to be confirmed or improved before this app becomes more widely used?

A: There obviously is more that needs to be confirmed in future research. We are speaking with other groups to look at the bigger population, more borderline values and repeated measures, and things of that nature. There also is the matter of using a separate accessory — the box and the glasses — to account for various lighting conditions. We are exploring ways of performing calibration with everyday objects instead, but this is further down the line.

 

Q: What are the potential benefits of this app if it is adopted into clinical practice?

A: One of the initial dream goals was that this would be the kind of thing that anyone can download and use. There is this thought that we can make a serendipitous diagnosis that a physician would normally miss. At the same time, we have to be careful. Such an app could lead people to infer that they may have cancer, but they could just have jaundice due to some other medical condition. I also want to emphasize that we are not trying to replace physicians with this app. We want to work with them. Maybe the app would work like a prescription — the physician wants to follow up with a patient but cannot see him or her daily in the office, so the app could be prescribed. They could use the app occasionally and monitor bilirubin levels. Another potential use is for disease management. If a patient has pancreatic cancer, the app could help monitor their condition to and ensure treatment is working. This could take the place of blood draws for these patients.

 

Q: Can you offer some insight into the overall potential of smartphone technology for cancer detection and diagnosis?

A: Doctors will not order a blood draw out of the blue; however, anyone can download an app and get tested. Most people already have smartphones, and there is a convenience factor that makes testing much more pervasive. Pancreatic cancer has a very low survival rate because it often is detected very late. Being able to make screening possibly more pervasive and accessible could be powerful. However, the issue of how people interpret the data is a concern.

 

Q: Is there anything else that you would like to mention ?

A: It is important to emphasize that not everyone who has jaundice has pancreatic cancer, and vice versa. This app does not detect all cancers; it is simply detecting a symptom of one type of cancer.

 

Reference:

Mariakakis A, et al. Proc ACM Interact Mob Wearable Ubiquitous Technol. 2017;doi:10.1145/3090085.

 

For more information:

Alex Mariakakis can be reached at University of Washington, Computer Science and Engineering, Room AC101, Paul G. Allen Center for Computer Science and Engineering, Box 352350, 185 Stevens Way, Seattle, WA, 98195; email: atm15@cs.washington.edu.

 

Disclosure: Mariakakis reports no relevant financial disclosures.

Alex Mariakakis
 

Researchers at University of Washington have developed a smartphone app designed to detect jaundice, a common early symptom of pancreatic cancer.

Investigators assessed the use of a novel app called BiliScreen, intended to capture pictures of a person’s eye and then estimate the person’s bilirubin levels.

“Jaundice is only recognizable to the naked eye in severe stages, but a ubiquitous test using computer vision and machine learning can detect milder forms of jaundice,” Alex Mariakakis, doctoral student at Paul G. Allen School of Computer Science and Engineering at University of Washington, and colleagues wrote.

Mariakakis and colleagues accounted for various lighting conditions by using BiliScreen with two different accessories: a 3-D printed box that blocks ambient lighting, and paper glasses printed with colored squares to help calibrate color.

In a cohort of 70 people, BiliScreen — when used with the 3-D box accessory — achieved a Pearson correlation coefficient of 0.89 and a mean error of -0.09 ± 2.76 mg/dL for predicting bilirubin level. Use of the app with the 3-D box accessory resulted in an 89.7% sensitivity and 96.8% specificity.

HemOnc Today spoke with Mariakakis about the app, the early efficacy it has demonstrated, and the potential benefits it may offer, and what must confirmed or improved before it becomes more widely used.

Question: How did this app come about?

Answer: The idea for this app was inspired by another project our lab is working on for an app that detects neonatal jaundice. The observation was that nurses and clinicians were essentially using their eye as a sensor, and then learning an algorithm that basically took the color of the baby’s skin and mapped it to the result of the blood draw that they would do later. The question became, if nurses can do this with some level of accuracy, can we get a smartphone to do this so we can perform this test anywhere with greater accuracy? This project has been going on for about 4 years. It led to the idea for this app for adults, which has been in development for about 2.5 years. It has been suggested that the collagen in the eyes are supposed to be more susceptible to bilirubin, so the thought was that we would be able to see jaundice in adults through the eyes.

 

Q: Can you describe the efficacy of the app so far?

A: We conducted a trial that included 70 adults, 31 of whom were healthy individuals. The other 39 had conditions of the pancreas and liver, but not necessarily pancreatic cancer. These 39 patients had bilirubin levels ranging from borderline levels — 1.3 mg/dL to 3 mg/dL — to very severe levels of 30 mg/dL. We wanted to quantify jaundice as a symptom of different conditions, as not every patient with jaundice has pancreatic cancer. The app was associated with 90% accuracy.

 

PAGE BREAK

Q: What needs to be confirmed or improved before this app becomes more widely used?

A: There obviously is more that needs to be confirmed in future research. We are speaking with other groups to look at the bigger population, more borderline values and repeated measures, and things of that nature. There also is the matter of using a separate accessory — the box and the glasses — to account for various lighting conditions. We are exploring ways of performing calibration with everyday objects instead, but this is further down the line.

 

Q: What are the potential benefits of this app if it is adopted into clinical practice?

A: One of the initial dream goals was that this would be the kind of thing that anyone can download and use. There is this thought that we can make a serendipitous diagnosis that a physician would normally miss. At the same time, we have to be careful. Such an app could lead people to infer that they may have cancer, but they could just have jaundice due to some other medical condition. I also want to emphasize that we are not trying to replace physicians with this app. We want to work with them. Maybe the app would work like a prescription — the physician wants to follow up with a patient but cannot see him or her daily in the office, so the app could be prescribed. They could use the app occasionally and monitor bilirubin levels. Another potential use is for disease management. If a patient has pancreatic cancer, the app could help monitor their condition to and ensure treatment is working. This could take the place of blood draws for these patients.

 

Q: Can you offer some insight into the overall potential of smartphone technology for cancer detection and diagnosis?

A: Doctors will not order a blood draw out of the blue; however, anyone can download an app and get tested. Most people already have smartphones, and there is a convenience factor that makes testing much more pervasive. Pancreatic cancer has a very low survival rate because it often is detected very late. Being able to make screening possibly more pervasive and accessible could be powerful. However, the issue of how people interpret the data is a concern.

 

Q: Is there anything else that you would like to mention ?

A: It is important to emphasize that not everyone who has jaundice has pancreatic cancer, and vice versa. This app does not detect all cancers; it is simply detecting a symptom of one type of cancer.

 

Reference:

Mariakakis A, et al. Proc ACM Interact Mob Wearable Ubiquitous Technol. 2017;doi:10.1145/3090085.

 

For more information:

Alex Mariakakis can be reached at University of Washington, Computer Science and Engineering, Room AC101, Paul G. Allen Center for Computer Science and Engineering, Box 352350, 185 Stevens Way, Seattle, WA, 98195; email: atm15@cs.washington.edu.

 

Disclosure: Mariakakis reports no relevant financial disclosures.