Pixaire1: Evaluation of automated chronic wound surface measurement systems.

Purpose. To evaluate two smartphone-based methods for measuring the surface area of chronic wounds using : Woundtrack (semi-automated measurement: WT) and Woundsize (automated measurement: WS), and comparing them with the reference technique: digitized planimetry (PL). Population and methods. Pixaire 1 is an open-label, single-center study involving 42 patients, from May to June 2023. Wound surfaces were measured using the three methods by two independent experts. We realized a four steps statistical analysis: multivariate analysis of variance; correlation between the two experts (precision); agreement between the two evaluated methods and the reference (accuracy); analysis of non-conformities (differences of more than 20% in absolute values compared with the PL measurement) in a subset of wound less than 8 cm2. Results. Of the 42 patients, 6 were excluded from the statistical analysis (multiplanar wound: 4; difficult edge delineation: 2). We found no difference in multivariate analysis We showed excellent agreement (ICC > 0, 9) of repeated measures (precision) for all three protocols. We also demonstrated excellent agreement (ICC > 0, 9) between WT and WS measurements versus PL (accuracy). However, accuracy and precision were better for WT than for WS. Analysis of non-conformities in small areas wounds showed no difference in variance and distribution between WT and PL, and showed a significant difference between WS and PL. Conclusion. Woundtrack is close to Digitized Planimetry, in terms of precision (reproductibility of the measure) and accuracy (correlation of measures with digitized planimetry). Despite the existence of non-conformities in small wounds, WT does not significantly differ of PL in this subset. WT should be considered as an effective method to measure the area of the wound, similar to PL, with a real benefit in implementation in current care setting (easy to realize, less time consuming). Woundsize showed less consistent results, despite a reliability and an accuracy that remains good. Its integration in a "Algorithm: propose then Clinician: correct and validate" procedure seems most efficient way to implement such methods.