High-Throughput Digitisation of Natural History Specimens

Abstract

The Natural History Museum, London (NHM) has embarked on an ambitious Digital Collections Programme to digitise its collections. One aim of the programme has been to improve the workflows and infrastructure needed to support high-throughput digitisation and create comprehensive digital inventories of large scientific collections. Pilot projects have been carried out for a variety of collection types, from which high-throughput imaging workflows have been developed and refined. These workflows have focused on pinned insect specimens (Blagoderov et al. 2012, Paterson et al. 2016, Blagoderov et al. 2017, Price et al. 2018), microscope slides (whole slide and specimen imaging; Allan et al. 2018, Allan et al. 2019) and herbarium sheets. The rate and time taken to digitise specimens is influenced by a number of factors that include, among others, the level of preparation and post-processing required, imaging approach, the type of specimens as well as the complexity and condition of the collection. As part of this presentation we will include information on the rate, cost and time to digitise various NHM collections, illustrating how our processes have improved digitisation efficiency and allowed us to maintain quality. The programme has run a variety of digitisation projects, gathering data about rates of digitisation (preparation, imaging, transcription etc.) and developing improvements. Collection types such as microscope slides and herbarium sheets lend themselves to higher imaging rates, while other collections such as pinned insects, which require greater amounts of specimen handling to remove labels, tend to have lower imaging rates (Fig. 1). In order to increase efficiency, we have developed approaches that minimise specimen handling. For example, workflows for pinned insects such as the Angled Label Image Capture and Extraction (ALICE) do not require the removal of specimen labels from the pin as the system can capture angled images of the labels, thus increasing the imaging rate three-fold (Fig. 1).  Another approach taken is to semi-automate mass digitisation using a combination of  temporary and permanent Data Matrix barcode labels (Allan et al. 2019). By using multiple barcodes at the imaging stage to encode information associated with each specimen (i.e. unique identifier, location in the collection, taxonomic name, type status etc.; Fig. 2), we can run a series of automated processes, including file renaming, image processing and bulk import into the Museum’s collection management system. Through adaptation of our workflows with this new approach we have increased the efficiency of digitisation processes, illustrating how simple activities, like automated file renaming, reduces image post-processing time, minimises human error and can be applied across multiple collection types.