Mycelium characterization and Image Analysis

Mycelium is a new biomaterial whose microstructure consists of fibers, the hyphae, that grow and eventually connect, creating a mesh that resembles an open cell foam structure. The initial purpose of this project was to extract structural features and develop a QSPR for mycelium transport properties with the aim to design a mycelium-based high performance membrane. Seven key features of transport scaling laws in porous media were extracted from SEM images of a mycelium sample.

This research project started in 2018 as a Master project under supervision of Dr. P. Nalam and Dr. O. Wodo, in collaboration with Ecovative Design.


TRANSCRIPT OF TALK: AUTOMATED MYCELIUM MICROSTRUCTURAL CHARACTERIZATION WITH IMAGE PROCESSING ARTIFACT PREDICTION

T. Stona de Almeida. Cambridge Open Engage (2024). doi.org/10.33774/coe-2024-jj4g6 (Preferred source)
T. Stona de Almeida. Zenodo (2024). doi.org/10.5281/zenodo.10712067

Link to the talk on YouTube: Lightning Talks, February 2024

Abstract of Talk:This talk discusses fungal mycelium as a biomaterial, its applications, and the process of microstructure characterization based on scanning electron microscope images, the challenges of micrograph knowledge extraction and how Machine Learning can help automate the identification of imaging artifacts.


MINING ARTIFACTS IN MYCELIUM SEM MICROGRAPHS
T. Stona de Almeida. Preprint v.1 (v.2 to be submitted soon): arxiv.org/abs/2103.07573

Abstract: Mycelium is a promising biomaterial based on fungal mycelium, a highly porous, nanofibrous structure. Scanning electron micrographs are used to characterize its network, but the currently available tools for nanofibrous microstructures do not contemplate the particularities of biomaterials. The adoption of a software for artificial nanofibrous microstructure for mycelium characterization adds the uncertainty of imaging artifact formation to the analysis. The reported work combines supervised and unsupervised machine learning methods to automate the identification of artifacts in the mapped pores of mycelium microstructure.

Conclusion: The proposed protocol for automated detection of imaging artifacts is consistent with the analytical approach of data cleaning based on expert knowledge and its extension to a larger dataset and double-supervised classification of artifacts by experts is encouraged. Transfer learning could be a potential machine learning method to generalize and automate the classification to future micrograph additions to the database.

Keywords: Machine learning; unsupervised learning; image processing; mycelium; microstructure informatics.


GRADIENT POROUS STRUCTURES OF MYCELIUM: A QUANTITATIVE STRUCTURE-MECHANICAL PROPERTY ANALYSIS
Scientific Reports 13, 19285 (2023): www.nature.com/articles/s41598-023-45842-5
E Oliverio, E Gawronska, P Manimuda, D Jivani, F Zullfikar Chaggan, Z Corey, T Stona De Almeida, J Kaplan-Bie, G McIntyre, O Wodo*, P C Nalam*

Abstract: Gradient porous structures (GPS) are characterized by structural variations along a specific direction, leading to enhanced mechanical and functional properties compared to homogeneous structures. This study explores the potential of mycelium, the root part of fungus, as a biomaterial for generating GPS. During intentional growth of mycelium, the filamentous network undergoes structural changes as the hyphae grow away from the feed substrate. Through microstructural analysis of sections obtained from the mycelium tissue, systematic variations in fiber characteristics (such as fiber radii distribution, crosslink density, network density, segment length) and pore characteristics (including pore size, number, porosity) are observed. Furthermore, the mesoscale mechanical moduli of the mycelium networks exhibit a gradual variation in local elastic modulus, with a significant change of approximately 50% across a 1.2-inch-thick mycelium tissue. The structure-property analysis reveals a direct correlation between the local mechanical moduli and the network density of the mycelium. This study presents the potential of controlling growth conditions to generate mycelium-based GPS with desired functional properties. This approach, which is both sustainable and economically viable, expands the applications of mycelium-based GPS to include filtration membranes, bio-scaffolds, tissue regeneration platforms, and more.

Keywords: Gradient Porous Structures; Fibrous Networks; Mycelium; Micromechanics; Microstructure Informatics.

*Corresponding authors


AUTOMATIC HIGH THROUGHPUT STRUCTURE FEATURE EXTRACTION FROM SEM IMAGES OF MYCELIUM: TOWARDS QSPR OF FIBROUS STRUCTURE FOR HIGH PERFORMANCE MEMBRANES
Master thesis supervised by Dr. P. Nalam and Dr. O. Wodo

Mycelium is a renewable biomaterial, a composite often considered as an alternative to synthetic polymers. It can be prepared by growing the structure in molds: agricultural waste, nutrients and liquid mushroom mycelium are mixed and put in a mold. Once its growth has achieved the desired size, the material is demold and baked. The baking process makes the material inert and dry, killing the mushroom and keeping the designed mold shape. When disposed, it can be used as plant nutrient. Mycelium has been employed in faux leather, structural boards and packaging. Mushroom mycelium is the root structure of mushrooms. It consists of a porous structure of fibers, the hyphae, which have typical diameter in the range of a few to several microns, and length in the range of a few microns to several meters, depending on the species and growth conditions. Mechanics of mycelium have been studied, however transport properties have not been explored to our best knowledge. The objective is to characterize SEM images of several mycelium samples of the species Ganoderma resinaceum. To achieve this aim we leveraged tools of materials informatics e.g. skeletonization, feature extraction, clustering. The goal is to design mycelium high performing membranes.



COMPREHENSIVE QUANTIFICATION OF THE HETEROGENEOUS STRUCTURE OF MYCELIUM
Poster presented at MRS Fall 2019 (MT03.10.03) with E. Oliverio, O. Wodo, P. Nalam, J. Bie-Kaplan and G. McIntyre.

The discovery and design of novel structures for reactive membranes which purify or enrich contaminated air, either without or with limited use of toxic chemicals, still have a significant environmental impact. Airborne byproducts of manufacturing and fuel combustion such as particulate matter (PM2.5; particulate diameter < 2.5 micron) have proven to be a global health risk and while current filtration membrane materials such as polyester and fiberglass benefit from tunable pore areas and fine fiber diameters, they are non-recyclable and must be replaced regularly due to fouling caused by the accumulation of pollutants. In collaboration with Ecovative Design, a bio-fabrication company working with mycelium (the root structure of mushroom), we studied the application of mycelium films as air-based filtration membranes. Surface proteins on mycelium hyphae are bio-adsorbants of several heavy metals and air contaminants and are therefore ideal candidates for membrane development. Like other naturally growing materials, mycelium has a heterogeneous structure, and in its optimization for membrane design with high filtration efficiency, a comprehensive quantification of its structure is necessary. In this study, through a combination of high-resolution and high-throughput imaging of the membrane, across several location and depths of the membrane, we quantitatively estimated several physical parameters such as pore area, fiber diameter, network topology, and fiber orientation of these heterogeneous membranes. Scanning Electron Microscope and atomic force microscope images were acquired to provide micron-level details of the mycelium network. These images were sampled across a range of magnifications, and image processing techniques such as statistical region mapping and axial thinning were employed for feature extraction. By obtaining a distribution of the fibers radii, Gaussian mixed models were used to identify three unique fibers indicating bifurcation as the main network growth mechanism. Additionally, unsupervised learning tools were employed to appropriately identify pores from the processed images, which showed a positively skewed data with an average pore area of 4micron^2 and mode 0.5micron^2 across the growth. These pore areas put mycelium in the magnitude for PM2.5 filtration, verifying mycelium’s potential as an air filtration membrane. The results accelerate the development of mycelium-based biofiltration products by establishing a feedback loop with Ecovative Design to optimize their growth conditions and species selection to generate optimal filtration microstructures.