AI-based image auto-segmentation with AI-based accuracy assessment with respect to a clinical task: Application in Radiation Therapy; A Virtue Ethics Analysis of Facebook’s Actions Following Implementation of Its Suicide Detection Algorithm in March 2017

My technical and STS projects are connected primarily through their ties to technology in the healthcare field. The technical project aims to build a pipeline with several computational capabilities that can be used during radiation therapy (RT) treatment planning of head and neck cancer patients. The STS project uses virtue ethics and Pritchard’s “Virtues for Morally Responsible Engineers” to analyze a case study that occurred with Facebook in 2017, in which I claim that Facebook’s implementation of a suicide detection algorithm was morally unacceptable because the company lacked three virtues from Pritchard’s list. While my technical and STS projects approach software in healthcare from vastly different avenues, the theme of utilizing code in the field of medicine is consistent for both topics.

The technical project is based on constructing a virtual pipeline that will aid in the RT treatment planning process. The user inputs CT scans of a patient, and the pipeline undergoes the following steps: it automatically segments the organs at risk (OARs) and target volume, it performs a Knowledge-Based Quality Control step to determine the plausibility of the automatic segmentations and reduce gross errors, it conducts dose estimations for each OAR and target, and it uses the dose estimations and standard dose objectives to determine which OARs are “relevant” and need to be further reviewed by the physician. Python was the coding language used, and Jupyter Notebook, Voilà, and many other Python packages were utilized to make the pipeline functional, interactive and easy to use. Another major aspect of the project was to analyze whether the pipeline would be worth using in a clinical setting. The pipeline was considered worth using if an average of at least one OAR was deemed “irrelevant” among many publicly-available datasets. The main current practice for physicians is manual segmentation, which relies on user expertise, and it is long, laborious, and prone to human error (Altman et al., 2015). The project aims to make an automated process in the form of a pipeline to increase quality and decrease time, labor, and errors in the RT treatment planning process.

The STS project is focused on utilizing virtue ethics and Pritchard’s “Virtues for Morally Responsible Engineers” to analyze Facebook’s actions during and after implementation of a suicide detection algorithm in March 2017. The algorithm rates each post based on likelihood of “imminent harm” and Facebook’s human reviewers have the ability to send first responders to the poster’s location (Goggin, 2019). I argue that using this code was morally unacceptable because Facebook lacked the following three virtues: “the ability to communicate clearly and informatively,” “the habit of documenting work thoroughly and clearly,” and “seeing the ‘big picture’ as well as the details of smaller domains” (Pritchard, 2001). Among other sources, I used a newsroom post directly from Facebook, a study on comparisons between Turkish and American suicide notes, and an individual report from The New York Times to support this argument.

By doing both projects together, I gained a greater appreciation of the impact that software developers have on the medical field. I realized that I needed to develop a pipeline that could be trusted by both physicians and patients. The product needs to be simple to use, understand, and update. That way, any potential issues can be easily rectified. Had I done these projects separately, I would have failed to understand how important it is for the engineers behind the software to follow the virtues for morally responsible engineers. I also would not have understood just how much of an impact that software has on the medical field, especially when the lines between technology and healthcare are becoming increasingly blurred. I learned to see the bigger picture when designing a novel product related to health and to understand the character of the person or company taking part in that product’s creation and implementation.

School of Engineering and Applied Science
Bachelor of Science in Biomedical Engineering
Technical Advisors: Dr. Shannon Barker, Dr. Jeffrey Siebers
STS Advisor: Dr. Benjamin Laugelli
Technical Team Member: Aaron Patton