Researchers Develop Framework to Infer 3D Blood Cell Properties from 2D Microscopy Images

It is a training approach that relies on aggregated statistics of a cell population, such as means, instead of per-cell labels, enabling learning when no reference exists for individual cells.

No. It does not claim explicit 3D reconstruction; it infers latent biophysical quantities that are aggregated to match measured population statistics.

Reported Pearson correlations range from 0.86 to 0.98 against a Sysmex analyser, on a development set of 390 specimens and 1,105 acquisitions across six devices.

The results are preliminary and research-stage; independent clinical validation is required before operational use, though the approach could reduce reliance on costly reference analysers.