Tony Pridmore



	Tony Pridmore
Home Projects People Publications Teaching
	BioImage Analysis Projects

	I am heavily involved in the Nottingham Centre for Plant Integrative Biology (CPIB). The Centre’s immediate goal is to understand how a very simple plant, the weed Arabidopsis thaliana, grows. Within CPIB plant biologists breed and grow plants in various situations, and mathematical modelers attempt to produce usually differential equations describing the processes behind the observed behaviours. My work fits between the two; biologists typically produce images of plant components, but modelers need quantitative data on size, shape, growth, deformation, etc. The bioimage analysis group produces techniques and software to extract quantitative measurements from a wide variety of images of plants. The following project suggestions all arise from CPIB work and related problems. Most involve image segmentation/feature extraction or tracking features through video sequences. All are real problems, of the right level for final year undergraduate or MSc student projects. A good solution to any is likely to be included in a CPIB publication, with the student named as co-author.

		Plant descriptions from Kinect data - A key goal in the development of new crops is an understanding of the factors that affect the formation of the plant canopy - the set of leaves it produces. Ideally, 3D descriptions of the shape of individual leaves are required. The Microsoft Kinect produces a registered colour image and depth map which may be of high enough quality to allow these descriptions to be formed. The goal of this project is to interpret Kinect data acquired from plants and identify and describe individual leaves.

	Smartphone app to identify common plant leaf diseases This project will consist of developing an image-analysis based smartphone app to aid the user in diagnosing plant diseases based on leaf appearance. The user will take a photo of a leaf, and tell the app which species the leaf is from (rice, potato, etc.). The photo can be taken under controlled conditions, for example removed from the plant and placed flat against a black background. The software will then analyse the image to provide a diagnosis of either healthy leaf, or diseased/stressed, and if so which it is - we can assume they have very different visual appearances on the leaf. For this initial work, we would want to demonstrate the software working on 3 species of plant with at least 2 or 3 common plant stresses each (including diseases, and environmental stresses such as lack of water). The image analysis will consist of colour- and texture-based segmentation methods, and clustering results into categories, and these processes will need to be built into a an app. A successful project will hopefully lead to development of the app into a more complete tool, and the student will be co-author of any relevant publications produced.

	Initialise RootTrace - RootTrace is a software tool which extracts the length, curvature and other properties of seedling roots from images taken with a normal CCD camera. The images show several tens of seedlings lying on laboratory plates. In the current version the user must click once on each root; the software then extracts the root from each of a time-ordered sequence of images. This video shows RootTrace in use. RootTrace is available via SourceForge and to date has been downloaded some 2000 times.The aim of this project is to remove the need for any user interaction by automatically initializing RootTrace. This is essentially an image segmentation project; it requires at least some sections of the roots to be identified in the first image, and start points for Root trace to be generated from those segments.

		Learning to segment scanning electron microscope images - High quality microscope images can show detailed properties of individual cells. The image to the left shows the junction between three plant cells, with the Y-shaped object at the centre being the point of intersection of several cell walls. The practical goal of this project is to develop methods and software capable of identifying the cell walls so that their thickness can be estimated. The complexity of the images and variation makes this a challenging task. Staff at CPIB have, however, recently segmented a large set of images by hand. This provides a training set which can be used, along with machine learning techniques, to produce an effective segmentation method. This project will look at how to learn to identify cell walls in these images, and so would suit a CS with AI student.

	Recognising Seed Germination Events - A key goal of research in crop science is to identify the conditions which affect seed germination; the ability to increase germination ratios would potentially allow crop, and so global food, yields to be increased. Crop scientists regularly work with large sets of seeds, monitoring their development and germination by capturing images. This project builds on a project from 2010/11 which developed image segmentation software capable of dividing the input image into regions which seem to correspond to different seedling components (see figure). The aim of the new project is to analyse the region descriptions obtained over time to identify events like germination (root emerging from seed) and establishment (appearence of leaves). Image data will be provided by colleagues in Crop Science, and if successful, the resulting software will be distributed to the crop science research community.

		Smartphone app to automatically digitise lab book pages and upload them to a database It is important in science to document all work thoroughly. Typically though, this is still done with paper and pen in a lab book. At the Centre for Plant Integrative Biology we are developing a database to store experimental results, images, procedures software etc. in. Being able to link written work in lab books to this database will be crucial in the future. We propose an app where a user can photograph with their smartphone a page of their lab book, and this photo will be analysed to extract as much information as possible (for example, character recognition to pick out page numbers, bar code reading to be able to tie the page to a database entry etc.). This information, together with the digital image will then be uploaded to CPIB’s DRIVE-db and stored. A successful project will hopefully lead to development of the app into a more complete tool, and the student will be co-author of any relevant publications produced.

	Automatically determining the orientation of fluorescing microtubules in images of plant cells - Microtubules are scaffolding-like structural elements in plant cells, which help cells to maintain their shape. We have seen that the orientation of these elements can change, and we need a way to quantify their orientation over patches of the image. To achieve this, there are a number of image analysis techniques which could be applied, and there is the potential to develop new algorithms as well. Ideally this would be built into a software tool for use by biologists in the Centre For Plant Integrative Biology (CPIB) to help analyse these effects.	$E:\Images\MAP4-GFP\210708 MAP4-GFP averages\Images for Spinning Disc Proposal\MAP4GFP-mature cells1.tif$