Training-free interpretation of count-calibrated density maps for label-free cell localization
Japanese Journal of Applied Physics · 16.11.2026 · 65 (21) · 21SP07
DOI: 10.35848/1347-4065/aeb94f ↗Kurz gefasst
Fachartikel in der Zeitschrift Japanese Journal of Applied Physics, online erschienen am 10.11.2026. Verfasst von Ryuto Shinozaki und 1 weiteren Autor:innen. Die Arbeit nennt 35 Literaturangaben.
Abstract (Originalsprache)
Density maps preserve spatial information, and permit object counts to be estimated by integration, but their continuous outputs do not directly provide the coordinates required for cell-by-cell analysis. We developed multi-scale density map interpretation (MSDMI), a training-free decoder that recovers cell-center coordinates from a count-calibrated density map. MSDMI calculates local density mass using two-dimensional moving sums, thresholds the response at each window size, accumulates the resulting binary support maps across window sizes, and returns regional maxima as coordinates. The method introduces no additional learning model and does not prescribe the upstream density-map generator. To determine whether cell-coordinate information retained in continuous density maps could be recovered, we performed a best-attainable recovery analysis across 10 density maps. The median optimal sub-pattern assignment (OSPA) distance was 3.9046 pixels for MSDMI, compared with 19.298 pixels for Gaussian blur followed by watershed and 19.501 pixels for direct binarization. OSPA distances differed significantly among the three methods (Friedman test, p = 4.54 × 10 −5 ), and paired comparisons with Holm correction showed lower distances for MSDMI than for either comparator (adjusted p = 0.0039 for both comparisons). The recovered coordinates also supported fluorescence measurement at the single-cell level. These results demonstrate that MSDMI can recover cell-coordinate information retained in continuous density maps and extend density-map outputs from object counting to localization and functional analysis.
Zitieren
APA
Shinozaki, R., & Numano, R. (2026). Training-free interpretation of count-calibrated density maps for label-free cell localization. Japanese Journal of Applied Physics, 65(21), 21SP07. https://doi.org/10.35848/1347-4065/aeb94f
Vancouver
Shinozaki R, Numano R. Training-free interpretation of count-calibrated density maps for label-free cell localization. Japanese Journal of Applied Physics. 2026;65(21):21SP07. doi:10.35848/1347-4065/aeb94f
BibTeX
@article{Shinozaki2026,
author = {Shinozaki, Ryuto and Numano, Rika},
title = {Training-free interpretation of count-calibrated density maps for label-free cell localization},
journal = {Japanese Journal of Applied Physics},
year = {2026},
volume = {65},
number = {21},
pages = {21SP07},
doi = {10.35848/1347-4065/aeb94f}
}
EUVIMED