EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Development and Evaluation of a Facial Skin RBX Image Generation System Based on Pix2Pix Image Translation

Patrick Po-Han Huang, Yu-Ting Tu, Chih-Yu Wang

Life · 07.11.2026 · 16 (11) · 1887

DOI: 10.3390/life16111887 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Life, online erschienen am 07.11.2026. Verfasst von Patrick Po-Han Huang und 2 weiteren Autor:innen. Die Arbeit nennt 32 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Science and Technology Council (NSTC), Taiwan.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This study evaluated paired Pix2Pix translation of standard-light and polarized RGB facial photographs into device-derived RBX (Red/Brown/X) Red and Brown images. The dataset comprised 617 matched image sets from 22 subjects, partitioned at the subject level into training (509 sets, 18 subjects), validation (60 sets, two subjects), and testing (48 sets, two subjects). Separate 256 × 256 U-Net and PatchGAN models were trained for each input and target combination. The reported comparison uses standard-light models trained for 200 epochs and polarized models trained for 500 epochs for each target; checkpoints were selected solely by validation mean absolute error (MAE). On the 48 test images, the polarized 500-epoch Red model achieved an MAE of 0.004063 ± 0.000287 and a structural similarity index measure (SSIM) of 0.987013 ± 0.001073; the corresponding Brown model achieved an MAE of 0.025414 ± 0.009142 and an SSIM 0.960360 ± 0.008078. Fixed geometric skin-candidate masks yielded an SSIM of 0.983198 ± 0.001172 for Red and 0.953726 ± 0.011496 for Brown. Values are image-level means ± sample standard deviations, not estimates from 48 independent participants. A separate two-participant cross-camera experiment used earlier model checkpoints. These findings support device-image reconstruction under the evaluated conditions; they do not establish chromophore concentration accuracy or clinical diagnostic performance.

Förderung

  • National Science and Technology Council (NSTC), Taiwan

Zitieren

APA

Huang, P. P., Tu, Y., & Wang, C. (2026). Development and Evaluation of a Facial Skin RBX Image Generation System Based on Pix2Pix Image Translation. Life, 16(11), 1887. https://doi.org/10.3390/life16111887

Vancouver

Huang PP, Tu Y, Wang C. Development and Evaluation of a Facial Skin RBX Image Generation System Based on Pix2Pix Image Translation. Life. 2026;16(11):1887. doi:10.3390/life16111887

BibTeX
@article{Huang2026,
  author = {Huang, Patrick Po-Han and Tu, Yu-Ting and Wang, Chih-Yu},
  title = {Development and Evaluation of a Facial Skin RBX Image Generation System Based on Pix2Pix Image Translation},
  journal = {Life},
  year = {2026},
  volume = {16},
  number = {11},
  pages = {1887},
  doi = {10.3390/life16111887}
}
Kostenlos starten →