Automated Dicentric Chromosome Identifier

and Radiation Dose Estimator


Automated Dicentric Chromosome Identifier and Radiation Dose Estimator (ADCI) software was developed by Cytognomix Inc. in response to a need expressed by colleagues in cytogenetic biodosimetry laboratories. They desired a method to improve their limited capacity to rapidly estimate radiation dose of multiple exposed individuals in an emergency situation. The cytogenetic dicentric chromosome assay is the gold standard of both the World Health Organization and International Atomic Energy Agency for assessing biological radiation exposure in individuals. However, even when taking into account the advent of microscope systems which automate some aspects of image selection, performing dicentric chromosome counts is a labor-intense task requiring manual examination of hundreds to thousands of metaphase images by trained personnel. ADCI automates the processes of selecting appropriate images for examination, dicentric chromosome detection, dicentric frequency calculation, calibration curve creation, and dose estimation. A typical sample can be processed in 10-20 minutes with accuracy comparable to an experienced cytogeneticist.

More information about all ADCI functionality can be found within our online wiki. For a brief introduction to the ADCI's main workflow and user interface consult our one page shortcut document. To learn more about the underlying algorithm employed by ADCI to perform dicentric chromosome detection or to read about the performance of ADCI on datasets supplied to us, view the publication history of ADCI. View a 10 minute video available in the Journal of Visualized Experiments (JoVE) for a brief demonstration of the software. Please contact us with any additional queries. Visit our frequently asked questions page to read answers to questions we’ve previously received.

A demonstration version of ADCI is available at no cost. Click here to request a download link. A quote for the licensed version of ADCI can be requested here.

We have also developed gene expression signatures based on machine learning of radiation responsive genes. Please see:

Zhao JZL, Mucaki EJ and Rogan PK. Predicting ionizing radiation exposure using biochemically-inspired genomic machine learning [version 2; referees: 3 approved]. F1000Research 2018, 7:233 (doi: 10.12688/f1000research.14048.2)


A straightforward single-screen interface provides access to all samples, curves, plots, console, and one-click sample processing.

To maximize accuracy, optimal images are automatically selected.

Calibration curves are generated using a streamlined wizard interface.

Extensive documentation covering these and all other features of ADCI software can be found in our online documentation.

Dose estimate calculations can take into account uncertainty due to the calibration curve and uncertainty due to the Poisson nature of dicentric yield.

Customized sample, curve, and dose estimation reports are generated on demand in several clicks.

Log files contain actions executed in a software session, enabling auditing capabilities and error recovery.

software demonstration

This 10 minute video published in the Journal of Visualized Experiments demonstrates how ADCI can be used to estimate the dose (in Gy) of test samples exposed to an unknown dose of radiation. Steps within the software necessary to do so, such as generation of calibration curves and selection of images, are described and demonstrated. The full text of the article provides additional details. Note there is an option to expand the video to a full screen view after the play button is clicked.


I have assessed ADCI v1.9 and find its speed and accuracy is suitable for emergency response biodosimetry. Health Canada will be adding it to our toolkit for managing performing dose estimates after exposures from large scale radiological or nuclear events.

— Ruth Wilkins, Director, Radiobiology Laboratory, Consumer and Clinical Radiation Protection Bureau, Health Canada

ADCI is an important development for radiation dose estimation. It is essential that we advocate for its adoption to improve radiation safety internationally.

— Edward Waller, NSERC Industrial Research Chair, Faculty of Energy Systems and Nuclear Science, Ontario Institute of Technology

Thank you again for coming here and demonstrating and delivering the software, it was beneficial to all of us.

— Farrah Flegal, Director, Biodosimetry Laboratory, Canadian Nuclear Laboratories

This product is a game changer in the biodosimetry field.

— Ruth Wilkins, Director, Radiobiology Laboratory, Consumer and Clinical Radiation Protection Bureau, Health Canada

I think ADCI will be better than DCscore software (Metasystem​s​) because you have integrated many image selection models. By this way, the dicentric detection ability of your system will ​be ​more effective.

— Pham Ngoc Duy, Biodosimetry Section, Center for Biotechnology, Nuclear Research Institute, Vietnam

This paper [F1000Research article] presents a considerable amount of work devoted to improving the accuracy of existing automatic dicentric systems for biological dosimetry. The authors have developed some very interesting ideas in constructing filters that detect false positive dicentrics and thereby improve the accuracy of 'hands-off' microscopy. The work has demonstrated that inter-laboratory differences, commonly reported in biological dosimetry, and evident between the two labs here, can be considerably improved by application of the filters. It is particularly interesting to see how the linear dose response reported by one lab, over a dose range that should show a lin/quadratic curve, can indeed be converted to linear quadratic by filtering out the false positives. Moreover they have demonstrated that much of the improvement can be achieved by a sub-set of their filters which should simplify future developments by not needing to employ all the methods.

The single most pressing remaining problem is the selection of metaphases of sufficient quality for passing onto the filtration procedures. Manual selection still seems better for removing the wide range of unsuitable material although the authors have demonstrated that the automated approach is getting there. ... I agree that fully automated image selection would be particularly advantageous when rapid triage dosimetry sorting of many cases is needed.

It is gratifying to see that the accuracy of dose estimations using the procedures described here falls well within the requirements for triage sorting following a major radiological incident.

Overall they are to be congratulated on a well-presented account of an improved approach to automated ‘dicentric-hunting’.

— David Lloyd, Centre for Radiation, Chemical and Environmental Hazards (CRCE), Public Health England, Didcot, UK

This paper [F1000Research article] presents an approach for the improvement of automatic detection of dicentrics and particularly the removal of some kinds of False Positives (FP) ... the manual scoring of dicentric chromosome is time consuming and the improvement of the result time is necessary in biological dosimetry. For dose assessment in case of triage of radiation exposed population, automatization of dicentric detection is very useful.

In order to increase the dicentric detection, the authors concentrate the study on the removal of FPs as well as on the selection of quality images. Apparently the main FPs in the metaphases of the 2 study laboratories are Sister Chromatid Separation (SCS) (84% of the FPs). This study shows that the authors reach to eliminate them with filters (about 55% of SCS removed). This is great and encouraging to remove other kind of FPs.

The study showed also the importance of selecting metaphases before the analysis and the detection. The quality of images and chromosomes is an important factor. The dose assessment becomes better when the images were selected than when they are not selected. The data show a difference in dose assessment between the 2 labs. The filters gave better results for CNL images than for HC images particularly on low doses.

However the selection of good quality metaphases is also time-consuming. There is a good comparison between manual and automatic triage of metaphases for the 2 labs highlighted by the statistical tests.

... Overall the authors are to be congratulated on a well-presented study of their work.

Eric Gregoire, Biological Dosimetry Laboratory (LDB), Institute for Radiological Protection and Nuclear Safety (IRSN), Fontenay-aux-Roses, France