Image-Based Approach for Fungal Network Analysis: Reconstructing Connectivity with Occluded Information

Understanding the growth dynamics of filamentous fungi can potentially improve sustainability and climate resilience in agriculture and forestry. Controlled experiments provide the possibility to gain insight into how such dynamics are affected by diverse conditions. In this paper, we propose a method for automatically detecting fungal mycelium in in-vitro samples, focusing on reconstructing hype connectivity in occluded regions caused in our samples by spores. This is achieved by 1) detecting spores and mycelium, 2) identifying points connecting mycelium to spores, 3) computing vectors approximating the local geometry, and 4) identifying the pairs of connection points most likely to be connected. We demonstrate that the proposed algorithm is highly effective in reconstructing the mycelium structure occluded by the spores in the cases where the mycelium representation is accurate. Our results suggest that the proposed method can improve network topology estimations in fungal images by removing artifacts and reconstructing connectivity occluded by spores.