After more than two decades shaping the frontier of biomedical imaging at Caltech, Singaporean scientist Changhuei Yang has returned to his homeland to join the National University of Singapore, drawn back by a national program that recognizes the value of its diaspora. His new laboratory will pursue a quietly radical idea: that microscopes need not speak to human eyes at all, but rather to the pattern-recognition depths of artificial intelligence. In an age when a machine has already outpaced expert pathologists in predicting how cancer spreads, Yang's homecoming carries implications that ext
Biomedical imaging pioneer Changhuei Yang joins NUS to develop AI-optimized microscopy
Machines can read with far greater accuracy than human eyes
Mark
So Yang spent over twenty years at Caltech and then just left. What made him decide to move to Singapore now?
Mimi
He's returning through Singapore's Returning Singaporean Scientists Scheme, which actively recruits leading researchers back home. But the real draw seems to be the research direction itself—he wants to build microscopes designed for AI, not for human eyes.
Luke
That's the pitch. But we should be careful: the source doesn't actually quote Yang explaining why he left Caltech or what Singapore offers that Caltech didn't. We know the scheme exists and that he joined, but the motivation is inferred.
Mark
Fair point. But the AI-optimized microscopy idea—that's concrete, right? He's actually going to do that work?
Mimi
Yes. He's leading a new lab at NUS specifically for this. The concept is that AI can detect patterns in medical images that human pathologists miss, so instead of designing microscopes for human vision, you design them to highlight what machines need to see.
Luke
And the evidence for that working comes from the lung cancer study—87 percent accuracy for AI versus 57 percent for expert pathologists. But the source itself says that was a pilot study and larger trials are needed. So we have proof of concept, not proof of clinical utility yet.
Mark
Right, but 87 versus 57 is a significant gap. That's not noise.
Mimi
It is significant. And if it holds up in larger studies, it could change how doctors screen for metastasis. Yang himself suggested it might spare some patients unnecessary aggressive treatment.
Luke
Which is a meaningful potential benefit, but it's still conditional on validation. The source is careful about that, which is good. What we don't know is whether Yang's move to NUS will accelerate that validation or whether he's simply relocating his lab.
Mark
Does it matter where he does the work?
Mimi
Probably not for the science itself. But NUS is Asia's top-ranked university, and Singapore has invested in attracting researchers like him. There may be resources or collaboration opportunities there that matter.
Luke
That's plausible, but it's not stated in the source. We know he moved, we know why the scheme exists, but we don't know what specific advantages NUS offered him over staying at Caltech.
Il Polso
- An AI trained on Yang's imaging technology predicted brain metastasis in lung cancer patients with 87% accuracy, leaving four expert pathologists behind at just 57% — a gap that reframes what medical expertise can mean.
- The tension at the heart of Yang's new lab is architectural: conventional microscopes are built for human perception, but the features AI needs to see may be invisible to any human observer.
- Yang's departure from Caltech after 21 years signals a broader gravitational shift, as Asia's leading universities compete to attract world-class researchers whose work sits at the intersection of biology, optics, and machine intelligence.
- Singapore's National Research Foundation is actively funding this return through its Returning Singaporean Scientists Scheme, treating scientific talent as a strategic national resource.
- The immediate work — designing microscopes optimized for AI rather than human vision — could eventually spare patients from unnecessary chemotherapy or radiation by identifying those at low risk of cancer spreading to the brain.
- With NUS ranked tenth globally and Yang among the top 1% of cited scientists worldwide, the lab launches from a position of institutional and intellectual credibility that will accelerate its path to clinical relevance.
After more than two decades shaping the frontier of biomedical imaging at Caltech, Singaporean scientist Changhuei Yang has returned to his homeland to join the National University of Singapore, drawn back by a national program that recognizes the value of its diaspora. His new laboratory will pursue a quietly radical idea: that microscopes need not speak to human eyes at all, but rather to the pattern-recognition depths of artificial intelligence. In an age when a machine has already outpaced expert pathologists in predicting how cancer spreads, Yang's homecoming carries implications that extend well beyond any single institution or career.
Changhuei Yang spent more than two decades at the California Institute of Technology redefining what light and computation could reveal inside living tissue. In August 2026, he left for the National University of Singapore, returning to his homeland under a national scheme designed to bring leading Singaporean researchers home. The move was both personal and strategic — for Yang, a homecoming; for Singapore, a signal of intent.
At NUS, Yang will lead the newly established Computational Microscopy, AI and Wavefront Shaping Lab, where the guiding premise inverts conventional microscope design entirely. Rather than producing images legible to human pathologists, his team will engineer systems whose output is optimized for artificial intelligence — images that may appear meaningless to the human eye but that machines can read with extraordinary precision. The animating question is simple and profound: what can AI detect in medical images that humans cannot?
Yang's career suggests this is no idle speculation. His 2013 development of Fourier ptychographic microscopy resolved a longstanding trade-off in the field, combining multiple lower-resolution images taken under varying light angles to produce high-detail views across a wide field — something conventional lenses could not achieve simultaneously. His team also found ways to image fluorescent structures deep inside biological tissue by counteracting the scattering of light.
More recently, a 2024 pilot study with Washington University School of Medicine demonstrated that an AI trained on lung cancer biopsy images could predict brain metastasis in individual patients with 87% accuracy, compared to 57% among expert pathologists reviewing the same material. The researchers were careful to frame it as a first step, but Yang has pointed to its potential to identify patients who might safely forgo aggressive preventive treatment.
Yang holds over 90 patents, more than half licensed to companies including Amgen and Leica Microsystems, and was ranked among the world's top 1% of cited scientists in 2025. NUS, now Asia's highest-ranked university, gains in him not only a distinguished researcher but a living argument that the future of medical diagnostics may be written in the language machines alone can read.
Changhuei Yang spent more than two decades at the California Institute of Technology building a career at the frontier of biomedical imaging—the science of seeing inside living tissue with light and computation. In August 2026, he left Caltech for the National University of Singapore, joining its College of Design and Engineering as a professor in electrical and computer engineering and biomedical engineering. The move, announced on August 26, was supported by Singapore's National Research Foundation through its Returning Singaporean Scientists Scheme, a program designed to draw leading researchers back home.
Yang's departure marks a significant shift in his research direction. At NUS, he will lead the newly established Computational Microscopy, AI and Wavefront Shaping Lab, where the fundamental premise of microscope design itself will be inverted. Rather than building instruments to produce images that make sense to human eyes, Yang's team will design systems optimized for artificial intelligence—images that may appear meaningless to a pathologist but that machines can interpret with far greater precision. The question driving this work is direct: what can AI detect in medical images that humans cannot, and how should microscopes be engineered to capture those invisible features?
Yang's track record suggests the question is not merely theoretical. Over his 21 years at Caltech, where he held the Thomas G. Myers Professorship of Electrical Engineering, Bioengineering and Medical Engineering and later served as executive officer of the electrical engineering department, he developed technologies that redefined what microscopy could do. His most influential contribution came in 2013 with Fourier ptychographic microscopy, a computational imaging technique that solved a fundamental trade-off in conventional microscopy: viewing a larger area meant sacrificing fine detail, while capturing extreme resolution limited the field of view. Yang's method takes multiple lower-resolution images captured under different lighting angles and combines them computationally to produce highly detailed images across a wider field of view. His team also pioneered optical time-reversal techniques to counteract light scattering as it passes through biological tissue, successfully imaging fluorescent structures 2.5 millimeters deep inside tissue.
More recently, Yang has turned his attention to the intersection of imaging and artificial intelligence. In a 2024 pilot study conducted with researchers at Washington University School of Medicine in St. Louis, his team trained an AI system on lung cancer biopsy images to predict whether the disease would spread to the brain. The results were striking: the AI correctly predicted brain metastasis in individual patients 87 percent of the time, compared with 57 percent accuracy among four expert pathologists examining the same images. The researchers acknowledged the pilot study as an initial step requiring validation through larger trials, but Yang has suggested the technology could eventually help physicians identify patients at low risk of brain metastasis, potentially sparing some from unnecessary chemotherapy or radiation treatment.
Yang's contributions have earned him recognition at the highest levels of the scientific establishment. He was ranked among the world's top one percent of scientists in the 2025 Stanford/Elsevier citation database and elected a Fellow of the U.S. National Academy of Inventors in 2020, an honor recognizing academic inventors whose work has advanced quality of life, economic development, and society. He holds more than 90 patents, 52 of which have been licensed to startups and major companies including Amgen, Bio-Rad Laboratories, and Leica Microsystems. His journey to this point began in Singapore, where he completed National Service before studying at the Massachusetts Institute of Technology and building his career in the United States.
NUS, where Yang now works, ranks tenth globally in the QS World University Rankings 2027 and stands as Asia's highest-ranked university. His arrival represents both a personal return and an institutional investment in the future of biomedical imaging at a time when the convergence of AI and microscopy is beginning to reshape medical diagnostics. The work ahead will test whether machines can see what humans cannot—and whether that vision can save lives.
Citazioni salienti
Instead of designing microscopes to produce images that make sense to the human eye, we'll be designing them to produce images optimised for AI. Ones that may look completely unrecognisable to us, but that machines can read with far greater accuracy.— Changhuei Yang, in remarks to NUS
Vuoi la storia completa?
Leggi l'originale su VnExpress International ↗
Education