Imageomics is a new scientific field leveraging the strengths of machine learning and computer vision to address biological questions at varying scales globally. By amalgamating high-resolution images of living organisms with advanced computer-aided analysis, Imageomics opens up innovative avenues to explore biological processes inherent to Earth.
The brain behind the concept, Wei-Lun Chao, an investigator at The Ohio State University's Imageomics Institute and an esteemed Assistant Professor of Engineering Excellence in Computer Science and Engineering, recently shed light on the latest advancements in the field at the annual meeting of the American Association for the Advancement of Science.
Elaborated Chao, along with his fellow presenters, imageomics has the power to transform our grasp over the biological and ecological world by converting investigational queries into computable problems. He emphasized imageomics' potential application for micro to macro level problems.
"If we make appropriate use of rapid advancements in machine learning and computer vision techniques, we can significantly aid scientists in solving critical but laborious problems," stated Chao.
Traditionally labor-intensive research problems that could take years to unravel manually, with the support of machine and computer vision techniques - like pattern recognition and multi-modal alignment - could witness a drastic escalation in the rate and efficiency of next-gen scientific breakthroughs.
"By incorporating the biological knowledge gathered over decades or centuries into machine learning techniques, we can perk up their capabilities regarding interpretability and scientific discovery," added Chao.
An instrumental part of this endeavour for Chao and his team is formulating foundation models in imageomics that will harness data from various sources to facilitate assorted tasks. Another aspect is developing machine learning models apt in identifying and even discovering traits to simplify object recognition in images via computer processing.
"Traditional methods for image classification with trait detection demand an enormous volume of human annotation. However, with our newly developed algorithm inspired by how biologists and ecologists identify traits distinguishing different species, the need for human intervention is greatly minimized," explained Chao.
The team's proactive approach instructs the algorithm to actively search for object specific traits like color and patterns while analyzing the image. Consequently, this new technique provides a more exhaustive account of the contents of an image and aids in quick, accurate visual analysis. Significantly, this method has demonstrated proficiency in handling intricate recognition tasks like identifying fine-grained species, notorious for their minute differences.
Chao believes the ability to easily utilize the algorithm potentially allows imageomics to be incorporated into diverse fields, including climate research and material science.
However, Chao points out that integrating different slices of scientific culture to collect adequate data and form novel scientific hypotheses poses a big challenge. That's where the importance of cross-disciplinary collaboration comes in, helping the field of imageomics grow and evolve. With imageomics, Chao envisions a future where AI strongly integrates with scientific knowledge, fostering a never-before-seen understanding of the natural world.
Disclaimer: The above article was written with the assistance of AI. The original sources can be found on ScienceDaily.
