Three-dimensional characterizations of two-photon excitation fluorescence images of elastic fibers affected by cutaneous scar duration

Background: The type or duration of a scar determines the choice of therapy available. Traditional detection methods can easily cause secondary trauma, so there is an urgent need for a non-invasive, rapid diagnostic approach. Methods: A strategy for quantitative analysis of three-dimensional (3D) elastic fibers in human cutaneous scars was designed, which included 3D reconstruction, skeleton extraction, quantitative analysis, and random forest regression. Results: Four reconstruction methods were used to reconstruct 3D two-photon excitation fluorescence images of elastic fibers for comparison. In the skeleton extraction stage, the 3D thinning algorithm was improved to prepare for accurate quantitative analysis, in which eight parameters comprising branches number (B-NUM), nodes number (N-NUM), averaged branch broken-line length (AB-BL), averaged linear branch length (AB-LL), averaged branch tortuosity (AB-T), branch direction consistency (B-DC), averaged branch volume (AB-V), and averaged branch sectional area (AB-SA) were presented. Six of them, except averaged branch tortuosity (AB-T) and branch direction consistency (B-DC), showed an explicit tendency to change with scar duration. In the random forests regression analysis, the six extracted parameters could be used to predict scar duration with R-2=0.981 and RMSE=0.513. Conclusions: The parameters we extracted had a distinct relationship with scar duration, and random forests regression showed better performance in forecasting scar duration than unitary models.

Automated summary

A new method for quantitatively analyzing three-dimensional elastic fibers in human cutaneous scars was designed, with extracted parameters being able to predict scar duration. Random forests regression showed better performance in forecasting scar duration than unitary models.