User Experience Research in the Imaging Field
What We Learned by Listening to Researchers, Imaging Scientists, and Bioimage Analysts
CZI’s imaging team wants to help all researchers measure, visualize, and quantify the biological processes underlying health and disease. To reach this goal, we work closely with the scientific community to understand what holds them back and collaborate to improve current systems. CZI supports imaging scientists that increase collaboration between biologists and technology experts, software fellows that maintain critical software tools, biologists who use imaging, and disruptive imaging technologies that will improve our mechanistic understanding of health and disease.
We heard from the scientific community about challenges using bioimaging software — tools that help researchers draw insights from the images their microscopes capture. As we considered what solutions might be possible, we wanted to better understand the experiences of the bioimaging user community: How do these individuals work together and how do their perspectives align? How might we build software for their short-term needs and long-term goals?
By focusing on people first, we aim to stay close to the challenges facing the imaging field and work towards enduring human-centered solutions.
With the help of CZI’s user experience (UX) research team, we interviewed 34 members of the imaging community using the UX research method called persona mapping. Personas are representative profiles of users that help product teams distinguish between different groups, build shared language when talking about groups, and frame understanding of the challenges with a user-centric (rather than product-centric) lens. Before jumping to “should we build feature X or Y,” we want to get to know the researchers who might benefit from such features. When interviewing individuals, we looked for common trends that sparked actionable insights that will keep our users’ needs front and center as we move forward with our strategy.
Imaging Community Personas: Who We’re Building For
After our interviews, we organized participants into three groups based on their skills, tasks, goals, and relationship to image analysis software. Let’s get to know them:
👩🏽🔬 Research Biologist Trainee: Uses Imaging Tools
Research Biologist Trainees are early-to-mid career academics working in a research laboratory. They design, implement, and analyze experiments to test hypotheses about biological questions under the mentorship of a lab’s Principal Investigator. Research Biologist Trainees implement the scientific method and advance understanding of biological processes by sharing their experimental findings through journals and conferences.
👩🏼💼 Imaging Core Scientist: Teaches Imaging Tools
Imaging Core Scientists are imaging lifecycle specialists, code-dabblers, and lovers of microscopes who work in an imaging core facility. They train research biologists in the end-to-end imaging process and manage shared facility equipment. Imaging Core Scientists train researchers to be more efficient and accurate imaging practitioners by offering centralized resources and education on the latest imaging techniques.
👨🏿💻 Community Bioimage Analyst: Builds Imaging Tools
Community Bioimage Analysts are computational experts that work at the intersection of biology and mathematics. They build machine learning methods and package them as workflows that others can use for their image analysis. Community Bioimage Analysts harness modern analytics for scientific discovery by making their work as accessible as possible to the largest number of researchers.
From our interviews, we found that each group has different goals, needs, and challenges to overcome. Here are four key learnings:
1. Research Biologist Trainees are stymied by the limitations of existing bioimage analysis platforms, relying on communities of practice to discover, learn, and troubleshoot tools
To consider building new software, we needed to know what imaging tools are currently used and why. Several platforms exist — they range from commercial to open source, span multiple programming languages, and take different approaches to interface and workflow.
To achieve the level of required analysis, Research Biologists often have to use multiple platforms to perform a single experiment and hunt for specific algorithmic methods via several social networks. Research Biologists acknowledge that each platform has drawbacks, and the absence of a curated algorithm ecosystem makes finding a trusted solution cumbersome, especially when they look to apply a certain tool — such as a machine-learning algorithm to segment cells — only to discover it is hard to install or doesn’t accurately fit their experiment. Biologists yearn for centralized, standardized, and intuitive ways to discover, evaluate, and implement bioimage analysis on large datasets — something our team endeavors to build.
2. Imaging Core Scientists and Community Bioimage Analysts encounter difficulty understanding and meeting the needs of Research Biologist Trainees
In speaking with Imaging Core Scientists and Community Bioimage Analysts, we found a common goal: help biologists succeed with their experiments; and a common challenge: difficulty connecting effectively with biologists. Imaging Core Scientists (who see themselves as educators and efficiency-makers for biologists) were frustrated by documentation and disparate dissemination of tools, making them harder to teach and recommend, which causes bottlenecks in simple workflow processes. If these simple tasks were easier to teach and share, biologists could be more autonomous, freeing up time for Imaging Core Scientists to assist with complex issues.
Bioimage Analysts expressed that they often “don’t speak the same language” as their biologist collaborators. One participant noted that the term “segmentation,” a common imaging task for biologists, means something quite different to a software engineer. Another analyst asks biologist colleagues to read through documentation before release in order to flag anything that might not make sense to a non-programmer. Bioimage Analysts struggle to build tools that are user-friendly; they acknowledge user interface design is not their expertise (or interest for that matter; they’d prefer to focus on complex computational algorithms over software interface and packaging!). Both of these challenges impede users’ goals and remind us how interdependent roles are within this ecosystem.
“If you write the most amazing code and nobody knows how to use it, what was the point?” — Bioimage Analyst
3. Everyone has long-term goals to level up their computational abilities
A Research Biologist Trainee should not have to be a master programmer to harness the power of image analysis. Biologists can spend several hours manually annotating the thousands of images they can now acquire from a microscope. Drawing the boundaries of cells or other regions of interest is a task that machine learning and computer programing is working to automate, but currently involves some coding skills by biologists. We assumed coding was outside Research Biologist Trainees’ comfort zone. Instead, we found biologists’ interest in coding — specifically Python for its machine learning capabilities — has been piqued after seeing successful use cases. Interview participants were excited by the prospect of reducing annotation time from hours to minutes by using machine learning models to segment cells based on reliable training models. Biologists who have already used Python in their workflows expressed a need for better tools to support image analysis in Python, which has good support for machine learning.
“As people become more advanced, not only will they be able to use more advanced tools; you’re going to inspire more people to come up with their own advanced tools.” — Imaging Core Scientist
Our other personas are also leveling up computational abilities. Imaging Core Scientists want to automate tasks that their facility users repeatedly request, and Bioimage Analysts contribute to low-level algorithmic code because a more mature programming ecosystem makes it easier for them to build plugin tools at scale. Unfortunately, these long-term goals are impeded by the challenges our users face with their immediate needs to use, teach, and build tools.
4. Great progress has been made, with expectations of more to come
While participants expressed frustrations with current field challenges, they also recognized just how much technological innovation has improved imaging in recent years. Participants shared stories of belabored manual work and life before machine learning, and many expressed appreciation for the advances in both commercial and open source analysis software. One participant described the field as being in a transitional state and expected more changes to disrupt the status quo.
“I’ve seen improvements from incredibly tedious annotations of 10,000 images to a tool so good anyone could use it…I assume it will continue this way. Most of these problems can be solved; someone just has to focus on them.” — Research Biologist Trainee
There is excitement for what machine and deep learning can do for the industry, and a willingness to adopt new tools, as long as the tools don’t sacrifice accuracy, transparency, and bias data. For new technology to succeed, users need to trust the process behind it, easily understand how to find and use it, and adapt it to their specific experiment or tool-building specifications.
After interviewing different types of people who interact with analysis tools and organizing the trends into personas, our product team has gained a stronger understanding of who we’re building software for, and more confidence that our strategies and next steps are driven by the real challenges community members face. Guided by this codification of needs, we now discuss ideas with a shared understanding of the path ahead. The process also generated a whole new set of UX questions we want to explore with Research Biologist Trainees, Imaging Core Scientists, and Community Bioimage Analysts that will help us shape our software solutions.
These findings provide a snapshot of how people are working right now, and we plan to continue learning and building alongside users to help drive imaging solutions. Our team has begun supporting the work of napari, a multi-dimensional image viewer platform for Python, and started an initiative focused on the developer services needed to build robust tools. You can learn more about developer services strategies at napari.dev. We know this is just the beginning, and we look forward to sharing progress with the imaging community, fostering transparency and engagement along the way.
To learn more about our work in science and to stay updated on funding opportunities, visit our website and sign up for our mailing list. You can also follow us on Twitter.
Lucy Obus, Senior User Experience Researcher for Science
Lucy Obus is a Senior UX Researcher on CZI’s Imaging team, where she collaborates with a cross-functional team to put modern technology in the hands of biologists who can change the world. Prior to CZI, she worked as IBM’s first Design Researcher on their flagship cloud-native integrated services platform. She holds undergraduate and graduate degrees from Georgetown University in Washington, DC.