Figure 3

Cellular and mitochondrial morphometric characteristics in fibroblasts derived from patients with IPD and healthy controls as well as classification into the two groups based on these characteristics. (A) Segmentation examples. Please note the reduced branching of mitochondria in the IPD example compared to the Ctrl example. Scale bars: 20 or 200 µm, as labelled. White boxes in stitched mosaic overlays indicate coordinates of raw image examples. (B) Individual subject analysis of mitochondrial morphometric features. The mean TMRM intensity within mitochondria at the base level is shown as TMRM_MitoMask. (C) Individual subject analysis of cellular morphometric features. The mean TMRM intensity within cells at the base level is shown as TMRM_Cell. Statistics and significance: without Bonferroni correction: [*P < 0.05; **P < 0.01; ***P < 0.001]; with Bonferroni correction: [^#P < 0.05; ^##P < 0.01; ^###P < 0.001]. The minimum number of cells analysed per individual: 175. (D) Classification into patients with IPD and healthy controls by morphofunctional characteristics. A random classification would result in a diagonal line (shown in black) in the receiver operating characteristic (ROC) and an area under receiver operating characteristic curve (AUC) of 0.5, whereas a perfect classification would result in AUC = 1. This ROC shows that predictions learned from pooled data that take into account mitochondrial and cellular morphometrics including both baseline and FCCP conditions, provide a better predictive power than baseline data alone (baseline alone, cyan, AUC = 0.83; baseline and FCCP, dark blue, AUC = 0.87).