The Future of Image Analysis
�+ ImageJ/Fiji
Curtis Rueden – CZI Imaging Software Fellow
Eliceiri/LOCI Lab – University of Wisconsin-Madison
CZI Imaging 2023 – July 19, 2023
https://github.com/ctrueden
https://loci.wisc.edu
https://archive.org/details/doctor-fun/page/n1911/mode/2up
My team did not create ImageJ
We also did not create Fiji
(but we have developed + maintained it since 2015)
We did create ImageJ2
and SciJava
and SciJava and SCIFIO
and SciJava and SCIFIO and VisBio, Bio-Formats,�ImageJ Ops, ImageJ-MATLAB, ImageJ-OMERO, imagej-node, jgo, scyjava, PyImageJ, napari-imagej… 😅
Project highlight: napari-imagej
https://napari-imagej.readthedocs.io/
Gabe Selzer
Ed Evans III
Mark Hiner
The future of image analysis
👇 The future, as envisioned by craiyon.com
According to AI
What's in store for the field of scientific imaging and image analysis in the next few years?
1. High-Resolution Imaging
2. Multi-dimensional and Multi-modal Imaging
3. Machine Learning and Artificial Intelligence
4. Quantitative and Computational Analysis
5. Integration of Big Data and Cloud Computing
6. Image Registration and Fusion
7. Visualization and Data Presentation
8. Open Data and Reproducible Research
These are some broad areas where scientific imaging and image analysis are likely to see advancements in the coming years.
The future of image analysis according to stable-diffusion-1.5
According to AI
What's in store for the field of scientific imaging and image analysis in the next few years?
1. High-Resolution Imaging
2. Multi-dimensional and Multi-modal Imaging
3. Machine Learning and Artificial Intelligence
4. Quantitative and Computational Analysis
5. Integration of Big Data and Cloud Computing
6. Image Registration and Fusion
7. Visualization and Data Presentation
8. Open Data and Reproducible Research
These are some broad areas where scientific imaging and image analysis are likely to see advancements in the coming years.
The future of image analysis according to stable-diffusion-1.5
According to experts in the field�(Nature Methods special issue)
https://forum.image.sc/t/83592
“A dialog between man and machine”
—Loic Royer
We will “develop foundation models capable of adapting to various imaging modalities and tackling complex segmentation tasks”
–Jun Ma & Bo Wang
“major technical obstacles are not likely”
—Anne Carpenter, Beth Cimini & Kevin Eliceiri
“We envision improved validation practices to be a key aspect of quantitative image analysis in the coming years”
—Jianxu Chen, Matheus P. Viana & Susanne M. Rafelski
According to experts in the field�(Nature Methods special issue)
https://forum.image.sc/t/83592
“the scientific community must…strive to create more incentives and reward structures.”
—Damian Dalle Nogare et al.
We must “work as a community to develop the educational pathways to disseminating this knowledge effectively to both the biology and the computer vision communities”
—Jianxu Chen, Matheus P. Viana & Susanne M. Rafelski
“We encourage continual communication between method developers and application scientists that emphasizes likely pitfalls”
— Talley Lambert & Jennifer Waters
Trends: What is changing
All uses of AI are valid 👉
Trends: What is not
* FAIR principles: Findable, Accessible, Interoperable, Reusable
https://www.go-fair.org/fair-principles/
Reproducibility�is getting harder
Cloud Computing
Software as a Service (SaaS) → Service goes away
👎
_
Can reduce heterogeneity of computing environments
👍
Reproducibility�is getting harder
Machine Learning
Model of GPU
→ supported CUDA version range
→ supported library versions (e.g. TensorFlow)
→ out-of-range unusable models!
Source: https://forum.image.sc/t/81736/5
Reproducibility�is getting harder
Machine Learning
A high-end GPU from 2022 cannot run a TensorFlow 1 model from 5 years ago
Model of GPU
→ supported CUDA version range
→ supported library versions (e.g. TensorFlow)
→ out-of-range unusable models!
How much reproducibility is enough?
Does software still need to work 10 years later?
Does it actually need to run on any machine?�
Perhaps not…
Motivations & incentives are unchanged
Publish or Perish
Time to Market
Novelty Bias
Not Invented Here
Imposter Syndrome
Tragedy of the Commons
Fear of the Unknown
“That which is common to the greatest number gets the least amount of care. Men pay most attention to what is their own: they care less for what is common.”
�—Aristotle. Politics, Book II, Chapter 3
Information Overload
Choice Paralysis
“We all slide back to what we are comfortable with.”
�—Mark Tsuchida
Our community produces a staggering amount of software!
Every line of code is a liability.�
—A favorite saying of� software maintainers
“If I’d had more time, I would have written a shorter letter.”
�—Various
“Less is more.”
�—19th century proverb
What might help
Image.sc Forum Community Partners�
https://forum.image.sc/
At least these horrible AI-generated approximations of human ingenuity are well organized now!
How to fund this?
THANK YOU
“Pick Two”
Source: https://commons.wikimedia.org/wiki/File:Project_management_triangle_-_fast_cheap_good.pdf