A compact optical prototype that captures two views on a single CMOS sensor produced measurable mechanical contrast in laboratory tests, including fresh mouse pancreatic tissue and freshly excised human breast and liver tissue. Across the mouse cohort, the reported average stress-contrast ratio was 1.9. Its average contrast-to-noise ratio, or CNR, was 3.7; here, that metric was used to describe how clearly a feature could be detected. The reported global stress-contrast ratios were 1.4 for the breast specimen and 2.9 for the liver specimen.
The work is aimed at high-contrast elastography, a way of mapping mechanical differences, and cancerous-tissue delineation without the overhead of two cameras. The authors say the current prototype still needs further miniaturization and real-time processing, and that larger clinical trials are required to establish diagnostic sensitivity and specificity.
One sensor, two views
The core design captures both stereoscopic views on one CMOS sensor. The authors report an overall instrument-volume reduction of more than a factor of ten. That architecture is part of the study's effort to support cancerous-tissue delineation without dual-camera overhead.
Basic optical tests broadly matched the design. The measured focal distance was 4.1 millimetres, compared with a designed 4 millimetres. The focal-spot width was 2.35 micrometres, compared with 2.24 micrometres in simulation, and focusing efficiency was 70.24 per cent.
From two images to a stress map
To reconstruct the mechanical signal, the analysis estimated disparity, the small positional shift between matching features in the two views. It used Ncorr v1.2, a circular tracking subset 500 micrometres across and a convergence threshold of 10^-4 pixels. The disparity output was then used to build the study's mechanical maps.
The first test was a silicone phantom containing an inclusion. At 10 per cent strain, the inclusion was detectable with a CNR of 5.2. At 30 per cent preload, the maximum reported CNR reached 7.1. The paper compares these values with the Rose criterion, a benchmark for feature detectability against background noise.
Signals in mouse tissue
The animal test used five fresh pancreatic tissue specimens excised from a cancer model. Across that cohort, the reported average stress-contrast ratio was 1.9, comparing the tumour-region signal with the surrounding benign-tissue signal. The average CNR was 3.7, which the authors report as sufficient for automated tumour detection in the tested specimens.
The reported mouse figures are cohort averages, and the analysis gives no confidence interval or formal statistical test for them. Larger clinical studies are still needed to establish diagnostic sensitivity and specificity.
A first look at human specimens
Human validation used freshly excised breast and liver tissue from surgical patients, with informed consent and organ-specific ethics approvals. In the reported specimens, the global stress-contrast ratio was 1.4 for breast tissue and 2.9 for liver tissue.
The liver specimen produced a tissue-wide CNR of 4.8, above the study's stated Rose threshold. In the breast specimen, global CNR reached 3.9. Local boundary-relative region pairs exceeded the same detectability benchmark at 1 millimetre from the automated boundary.
The human evidence is narrow: the analysis included one breast specimen and one liver specimen. The authors call for larger clinical trials to determine diagnostic sensitivity and specificity.
The gap to clinical testing
The study's results support the feasibility of mechanical-contrast mapping in the tested phantom and tissue specimens, but they do not establish diagnostic performance. The authors say the prototype needs further miniaturization and real-time processing, followed by larger clinical trials.
Disclosures and data
The paper reports that B.F.K. and C.M.S. have financial interests in OncoRes Medical Pty Ltd. The company did not support the work, and the other authors report no competing interests.
The acknowledgements list fellowship, institutional award, Australian Research Council and chair support for named investigators. Main supporting data are available in the paper and Supplementary Information; larger additional data are available from the corresponding authors on reasonable request.
Paper data and sources
Original title: Metalens stereoscopic optical palpation for imaging cancer mechanics
Authors: Haoyi Yu, Rhys Jones, Chi Li et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-25
DOI: Not available
Original paper · Full text