Automated Prediction of Bacterial Exclusion Areas on SEM Images of Graphene–Polymer Composites

To counter the rising threat of bacterial infections in the post-antibiotic age, intensive efforts are invested in engineering new materials with antibacterial properties. The key bottleneck in this initiative is the speed of evaluation of the antibacterial potential of new materials. To overcome this, we developed an automated pipeline for the prediction of antibacterial potential based on scanning electron microscopy images of engineered surfaces. We developed polymer composites containing graphite-oriented nanoplatelets (GNPs). The key property that the algorithm needs to consider is the density of sharp exposed edges of GNPs that kill bacteria on contact. The surface area of these sharp exposed edges of GNPs, accessible to bacteria, needs to be inferior to the diameter of a typical bacterial cell. To test this assumption, we prepared several composites with variable distribution of exposed edges of GNP. For each of them, the percentage of bacterial exclusion area was predicted by our algorithm and validated experimentally by measuring the loss of viability of the opportunistic pathogen Staphylococcus epidermidis. We observed a remarkable linear correlation between predicted bacterial exclusion area and measured loss of viability (R(2) = 0.95). The algorithm parameters we used are not generally applicable to any antibacterial surface. For each surface, key mechanistic parameters must be defined for successful prediction.