Project promotion materials:

Project Homepage:

Data collections can be seen on:

Software is available at:

Project Members:

Anitha Kannan (Project Manager,

Michael Eager (Developer,

Simran Kaur (Lead Business Analyst,

Dr David Barnes (Senior Research Fellow,

ANDS Contact:

Mingfang Wu (

Project Status:


Multimodal Kidney Image Analysis Project

Monash University

Project Description:

There is growing evidence that the number of glomeruli in the kidney – which can range from 200,000 to two million – is strongly linked, not only to kidney disease, but to cardiovascular health. Until recently, stereological examination of sectioned sub-samples of ex-vivo, post-mortem kidney tissue was the only proven technique for estimating the total glomerular population, and therefore of assessing disease state. Prof. John Bertram’s group at Monash is one of the world-leaders in this technique ‘Counting in the kidney', which is entirely manual and “wet-lab-based”. In 2011, the possibility of counting every individual glomerulus in a whole kidney was demonstrated for the first time, by two groups collaborating with Prof Bertram, using high-field magnetic resonance (MR) imaging and a threshold-based segmentation and counting algorithm ‘Measuring glomerular number and size in perfused kidneys using MRI’ and ‘Quantification of glomerular number and size distribution in normal rat kidneys using magnetic resonance imaging’. Total counts from MR images were shown to be in excellent agreement with totals estimated (extrapolated) from subsequent stereological examination and sampled counting, for the same physical specimens. Note that in stereology, the technique is performed on a selection of the total number of 2D tissue sections. The complete manual sectioning of a murine kidney and subsequent counting of all glomeruli by stereology is not feasible as there would be typically 500-1000 sections to be viewed, and of order 30,000 individual glomeruli to count. However, small ‘representative’ subsections of kidney tissue have been sectioned and imaged with virtual microscope technology, providing small but representative volumetric image stacks containing typically ~30 glomeruli. Such volumetric, 3D digital image stacks, presently of small kidney sub-samples, but ultimately of entire organs, open the way to automated counting of every individual glomerulus in physically sectioned kidneys.

Progress towards successful in-vivo imaging of human kidneys for diagnostic purposes and an absolute quantification of renal disease state requires a staged program of:
(a) 3D MR imaging of ex-vivo animal models and associated automated glomerular segmentation and counting;
(b) acceleration, if not automation using e.g. virtual microscopy and image alignment to create volumetric (3D) digital image stacks, of the 2D tissue section microscope analysis for quantifying whole-kidney glomerular count and distribution;
(c) synchrotron-based phase-contrast tomography for fast, ultra-high-resolution imaging of ex-vivo animal model kidneys providing an independent, exquisitely-detailed characterisation of the vascular and glomerular 3D structure;
(d) integration of the data from (a), (b) and (c), analysis of the combined data types and calibration of these new techniques against current best-practice manual kidney histology and stereological analysis; and
(e) translation of ex-vivo work through to in-vitro and in-vivo animal models, and ultimately in-vivo human kidneys.

This project has given Australian biomedical researchers a significant advantage in the nascent field of the quantification of glomerular distribution and its relationship to nephron number and chronic kidney disease using ex-vivo kidney imaging and image analysis. This in turn, will make them perfectly placed to translate the work to in-vitro animal organs, and ultimately in-vivo animal models and then humans.

Project Impetus
This project aimed to deploy a multi-modal kidney image integration and analysis pipeline to place Australian researchers at the absolute forefront of the nascent field of clinical diagnostic kidney imaging. The convergence in the Monash Clayton precinct of world-class kidney research teams, multi-scale imaging modalities (including the immediately-adjacent Australian Synchrotron and Monash Biomedical Imaging facilities), and computational expertise and capability (including the Multimodal Australian ScienceS Imaging and Visualisation Environment - MASSIVE), made this the best site worldwide to fully-establish and lead the field.

Project Outcomes
This project has produced components that enable:
(i) the deposition and management of kidney images from MR, VM and CTX modalities in one software system (DaRIS),
(ii) the automatic segmentation, counting and population analysis of glomeruli in rat and mouse kidneys using MR and CTX data, and semi-automatic processing of VM data, and
(iii) co-registration of the multi-modal image data and unique analysis and visualisation capabilities for kidney data.
The completed system, Xglom, also enables and simplifies future strategies towards identifying appropriate techniques for in-vivo clinical kidney imaging, such as exploring a range of contrast agents for MR, staining agents for VM, and imaging techniques for CTX.

Research Champion:

Professor John Bertram
Department of Anatomy and Developmental Biology, Faculty of Medicine,
Nursing and Health Sciences, Monash University

Professor Gary Egan
Monash Biomedical Imaging, Monash University & Faculty of Medicine,
Nursing and Health Sciences, Monash University

Dr David Barnes
Monash Biomedical Imaging and Monash e-Research Centre, Monash University

Dr James Pearson
Australian Synchrotron &
Monash Biomedical Imaging, Monash University

Associate Professor Roger Evans
Department of Physiology, Faculty of Medicine,
Nursing and Health Sciences, Monash University

Data Type:

Input Data:
We will integrate imaging data from the Agilent MR scanner, a Slide scanner and the Australian Synchrotron Imaging and Medical Beamline

Output Data:
Quantitative analysis reports of kidney images from each modality including: number of glomeruli, size, position and density.

High Level Software Functionality:

Technical outcomes/solutions developed by this project
• Simple workflow of 3D image processing for ultra-high field MR and micro CT imaging data. The Xglom pipeline can analyse MR and CT images for the number of glomeruli, their distribution and an estimate their volume. This analysis contains valuable information for understanding kidney function and development. The Xglom software enables researchers with limited technical abilities in HPC or image processing to gain a distinct advantage in their research.
• 40 times speed up in the analysis stage of the processing pipeline with more efficient methods (the original methods took over 20 minutes to analyse the processed image, whereas the new methods took as short as 0.25 seconds to analyse).
• Pre-processing and bias-correction of raw images dramatically improved the output of the image cleaning stage, which also improved the outcomes of the glomeruli quantification and analysis stage. The pre-processing and bias-correction methods have further implications for MBI and the Agilent 9.4T MR scanner.
• Simple interactive searching of dataset database and local storage. Searching terms included image types, modification status, metadata information, processing status and output statistics.
• The virtual microscopy plugin, GlomViewer, was a great advancement in the automation and simplification of the “Gold-standard” techniques currently being used in the Bertram Lab.

Progress made / challenges faced in developing the software and integrating data
• The Xglom application integrates the cutting-edge medical image database, DaRIS, into a simple graphical user interface in MATLAB. This system allows secure, permission-controlled access to DaRIS projects directly controlled by kidney researchers. The unique feature of the kidney project was the different methods for each image modality. The features within Xglom application allowed for filtering data based on metadata information, downloading of images and manipulation of metadata properties (i.e. whether the dataset has been modified or processed).
• Multiple modalities presented issues regarding different image file formats, different metatdata and different contrast and SNR of image values. The Xglom pipeline was designed to accommodate different types of MR and micro-CT images, each with different image properties, to be processed and show similar counting and quantification outputs.
• The challenges with viewing virtual microscopy image stacks (over 30 GB) were overcome with simple, efficient methods to load and view the images. The VM module of Xglom, GlomViewer, was distint from the Xglom pipeline in that the user manually selected the stereology grid points. The UI development presented additional challenges to satisfy the stakeholders’ requirements.


270503 Animal Anatomy and Histology
300505 Anatomy and Physiology
111401 Foetal development and medicine
110320 Radiology and Organ Imaging
321012 Nephrology and Urology