My Experience in Organizing a Bioimage Analysis Workshop in Taiwan

During the Christmas break, I had the opportunity to conduct two one-day workshops on bioimage analysis at Academia Sinica, Taiwan. Not only did I learn a great deal about bioimage analysis through this workshop, but I also gained insight into the specific needs and interests of biologists and how to effectively teach these topics. Additionally, the diverse range of participants, including researchers from various fields and institutions, provided an opportunity to gain a comprehensive understanding of the bioimaging community in Taiwan (or north Taiwan, at least).

I am now writing an article to share my reasons for organizing the workshop, the planning process, and the valuable lessons I learned from the experience. Additionally, I will also be sharing some statistics that reflect the current state of bioimaging in Taiwan.

How it started

My partner is a cell-molecular biologist in Taiwan. For her experiments, she had to do lots of imaging. Yet, she processed most images (e.g. counting cells) manually, which takes ages. To the point that she got a pile of data unanalysed, and struggled to generate a report. Bio-image analysis is not included in the educational syllabus for bioscientists in Taiwan, so there is a good reason she didn’t know how to do it properly.

I studied physics and molecular biology at the university, and for many years aside from my formal education I’ve been primarily working on processing bioimages from microscopies, including cryo-EM/ET and super-resolution/light-sheet. I won’t say that I am an expert, but I believe I am “good enough”. So I thought, I could teach her. But wait, I teach coding at the university (as a PhD student at the University of Cambridge, I have some teaching responsibilities) and I know I could teach 15 at once.

In mid-November, I discussed my idea with my partner, and she thought its a good idea too. I emailed my previous PI at Sinica to ask if he would be interested in helping me. Meanwhile, my partner and I made a questionnaire, to learn about if there are needs and what they are, and what people find the most problematic in their approaches to bioimage analysis.

The questionnaire

The questionnaire included questions about their career stages, place of work/study, professional title, coding abilities and language used, and the imaging techniques and image processing software they use. 40+ people responded to my questionnaire, and the responses to these questions provided valuable insights into the bio-imaging community in Taiwan. The results are summarized in the figure below.

Some Stats. Additional points to mention: *1 Where are they from? Most people who responded to my questionnaire were from NTU and Academia Sinica; *2 Field of study: I asked the question “What do you identify yourself as?” There were cell/ developmental/ immune/ cancer/ structural/ synthetic/ plant biologists, bio-chemists/physicists, and a few bioimaging/ biophotonics specialists. There were many that identify themselves as neuroscientists, but from the research methods, they can be subdivided into the aforementioned groups; *3 Microscopy methods: Confocal is overwhelmingly popular, multiphoton interestingly comes second, and most are custom-built. Many use EM and a small minority uses light-sheet, super-resolution, SHG and TIRF (Are they super-res?) Not everyone answered this question so I am not plotting a graph for it

The answer to the question that I found most interesting was, “What do you think is the main challenge/ obstacle of the bioimaging field in Taiwan?” The responses to this question were summarized into three main points:

1. Lack of people to discuss work with.
2. Limited resources.
3. Insufficient time.

I understand that I cannot directly address the issue of limited time as the complexity of biological questions they are trying to solve may not allow for it. However, I can help to alleviate the challenges of limited resources by providing tutorials and directing people to relevant resources. My main focus will be on addressing the main challenge identified, which is the lack of people to discuss work with, probably by encouraging people to discuss their images and analysis methods.

Conversations with Professionals

I consulted with several individuals to gather insights on what and how to teach for the workshop.

During the Bioimage Analysis Symposium at the Francis Crick Institute, I spoke with Robert Haase, who is a developer of many image analysis software and curator of many online bioimage-analysis tutorials. He shared his insight that having participants aware of their fellow attendees and encouraging collaboration with those from similar scientific backgrounds is crucial. He suggested that I could request a slide from each attendee at the beginning of the workshop, with three pictures: one of themselves, one of their science, and one of something they enjoy doing. This way, participants could easily identify and connect with others who share similar interests. I realized that my role in the workshop was to serve as a “match-maker” and facilitate connections between attendees.

I also had a conversation with Chien-Cheng Michael Shih, who works as a bioinformatician at Norvatis and specializes in image analysis. During our discussion, we shared ideas about image analysis, and he gave me an overview of the history of bioimage analysis packages. This was my first time gaining an understanding of image analysis from an industry perspective. Through our conversation, it became clear that I should teach ML-based software that is easy to use and accessible to anyone in the workshop. Based on this insight, I decided to teach Cellpose and ilastik in the workshop. He also provided me with valuable insights that prompted me to reconsider my career as a bioimaging specialist.

I had many more conversations with people from various backgrounds, all of which helped me to plan a workshop.

The schedule for the workshop

Preparing for the workshop…

Since I am just a starting PhD student and I doubted, anyone would want to stay for more than a day with me, I planned my workshop to be a day long.

From conversations, I learnt that most people are doing image processing without knowing the terminologies of what they are doing. This resulted in difficulties when they face problems and they don’t know how to ask/ google, as they don’t know how to verbalise their questions. So I made the first two sections, Sections 0 and 1 of the workshop to cover the image terminologies (e.g. image/pixel/look-up table) and image operators (thresholding/ filtering/ transforms) respectively. Both sections are quite standard as an introductory course for image processing. Additionally, I included some other conceptual information such as the importance to view bioimage analysis as part of the experiment rather than something you do as a side.

For Section 2, I wanted to incorporate machine learning (ML) techniques and decided to demonstrate the use of Cellpose and ilastik. However, I was unsure of what additional content to include. After watching a lecture by Robert Haase on YouTube, I decided to briefly discuss basic ML principles such as what are machine learning, deep learning, and supervised and unsupervised learning, as well as introduce the use of the random forest for pixel and object classification and algorithm validation. The teaching materials for Sections 0, 1, and 2 were primarily based on Robert Haase’s YouTube tutorials and Peter Bankhead’s Bioimage book.

From the questionnaire, many reported that although they had attended image analysis workshops in the past and understood the content, they still struggled to apply the knowledge to their own work. So I decided to have my last section, section 3 be titled “How to Think like a Bioimage Analyst”, in which I will show participants some images, tell them what I want to quantify and ask people to use what they have learnt in the day to discuss in the group to draw a solution pipeline. After each case study, I expanded on more specific image analysis concepts and shared my experiences in the field of bioimage analysis.

In addition, I also arranged two coding sessions (codelabs) between sections for people to “experience” coding and image processing with python. For the teaching material, I copied and modified the Jupyter notebook for bioimage analysis in python made by Jonas Hartmann when he was in EMBL, who I also consult for the workshop.

The overall schedule looks like this:

Workshop schedule. All the workshop materials are available on this Github page

The Workshop

Workshop @ICOB

As mentioned before I reached out to my old PI at the Institute of Biological Chemistry (IBC) and they connected me with the imaging core facility at IBC. We discussed and realized that there would not be a lot of people from biochemistry who would be interested, so we decided to co-organize with the imaging core at the Institute of Cellular and Organismic Biology (ICOB) as well. Additionally, while speaking with Dr Ching-Lung Hsu about an internship over the Christmas break, he suggested I could also do a workshop for Neurosciences Program at Academia Sinica (NPAS) and this is how my second workshop (but happened first) came about.

I communicated with the participants through emails and Slack and collected slides from them prior to the workshops. The first slide consisted of three pictures; one of themselves, one of their sciences, and one of their hobbies (as suggested by Robert). The second slide was optional and featured their own bioimage and what they wanted to quantify. I also made a short python assessment to test people’s familiarity with python (interestingly the score shows a bimodal distribution). I was told that most people in Taiwan are good at reading English but less so at speaking and listening, so I decided to teach in Mandarin Chinese.

The NPAS workshop was attended by approximately 25 participants, mostly from labs affiliated with the NPAS framework. The ICOB workshop, which happened a week after, was open to anyone at Sinica and had a diverse group of attendees, with more than 50 participants from various backgrounds such as plant biology, biochemistry, cell biology, and biophysics. Most attendees are either PhD students or postdocs. There were also a few PIs attending.

I was too excited about the workshop and I did not sleep so well the night before, so it was to my surprise that most parts of both workshops ran smoother than expected. At the beginning of the workshop, I gave an opportunity for everyone to introduce themselves in 10 seconds with the slides they provided, which helped as an ice-breaker. The part that most people got stuck with, was the beginning of codelab 1, I found that there were many people getting problems setting up their software environment, which took them quite a bit of time. In codelab 2 of the ICOB workshop someone asked me to do a live demo on coding. umm… coding with my typing shown on the big screen was great pressure, I should never try again 😅

The part of the workshop that I was most unsure about was the third section, “How to Think like a Bioimage Analyst.” I spent a lot of time thinking about the case studies and trying to find the right balance between something that was challenging but still manageable for the attendees. It was really towards the night before the first workshop that I was able to prepare three case studies: the first one on colocalization (I found that colocalization quantification was what most people want to do via image analysis) and the second on volumetric image analysis, and third case was on 2-photon imaging. I was nervous that the case studies may not be well received, but to my relief, this section went extremely well. The participants were engaged in discussion and trying to come up with ideas. I attribute the success of this section to Dr Ching-Lung’s exceptional skills as a moderator during the first workshop, which made the class discussion after the group discussion more insightful. This experience gave me the confidence to run the second workshop.

“How to Think like a Bioimage Analyst” session, workshop @NPAS

Both workshops finished later than the scheduled finishing time. We had a social at MoS at Sinica. Although there are only a handful of people attending the social, I loved this section the most as I finally got to talk to some people about their sciences! 😉

Attendees’ Feedback and How I will change it if I am to run again

I got mostly positive feedback. Most negative feedbacks were to do with having not enough time to explore the code, or the material being too condensed. Other than that, I found that although the codelab sessions were for people to learn how to do image processing with python, most people did not have a good foundation in programming. Some reflect that there were not enough case studies, or that certain specific topics were not covered.

If I am to run the workshop again, I will definitely make it into a two-day event (or longer). In addition, I might also add a one-day-long optional session a few days before the workshop, to make people confident with scientific python. Maybe another optional half-day introduction to ImageJ too. Also, I would make the codelab longer, during which I will make “breakout rooms” to demonstrate how to use some of the more “specialized” image analysis tools, such as those for cell-tracking, volumetric segmentation, and image stitching, for interested audience. To make all of these happen, I also need to collaborate with other people with different domain knowledge than what I have — I can even ask the attendees to be the presenters for one of the breakout rooms! I believe encouraging people to share ideas is a good way to build up a supportive community. Aside from that, having a number of participants presenting their own data at the beginning of the workshop (2mins each) would be great too.

Another thing I noticed during the workshop is that although there are many people doing image analysis in Taiwan, they are often “isolated.” They are usually the only one or two in labs that are doing the image analysis, as a result, they have no one to discuss. During the workshop, I encountered quite a few cases where someone just happened to have the solution to the solution of someone else — and people are willing to share and discuss. I now have more than 70 people who attended my workshop. I am thinking about what I can do to make a platform for people to discuss sciences, not necessarily image analysis, freely like during the workshop. Still figuring it out… any suggestions are welcome!

It looks like my next workshop will be at National Taiwan University… I am very much looking forward to it! 🤗