Improving the accuracy of dose estimates from automatically scored dicentric chromosomes by accounting for chromosome number

Radiation dose estimation with multiple artificial neural networks in dicentric chromosome assay

PURPOSE

The dicentric chromosome assay (DCA), often referred to as the 'gold standard' in radiation dose estimation, exhibits significant challenges as a consequence of its labor-intensive nature and dependency on expert knowledge. Existing automated technologies face limitations in accurately identifying dicentric chromosomes (DCs), resulting in decreased precision for radiation dose estimation. Furthermore, in the process of identifying DCs through automatic or semi-automatic methods, the resulting distribution could demonstrate under-dispersion or over-dispersion, which results in significant deviations from the Poisson distribution. In response to these issues, we developed an algorithm that employs deep learning to automatically identify chromosomes and perform fully automatic and accurate estimation of diverse radiation doses, adhering to a Poisson distribution.

MATERIALS AND METHODS

The dataset utilized for the dose estimation algorithm was generated from 30 healthy donors, with samples created across seven doses, ranging from 0 to 4 Gy. The procedure encompasses several steps: extracting images for dose estimation, counting chromosomes, and detecting DC and fragments. To accomplish these tasks, we utilize a diverse array of artificial neural networks (ANNs). The identification of DCs was accomplished using a detection mechanism that integrates both deep learning-based object detection and classification methods. Based on these detection results, dose-response curves were constructed. A dose estimation was carried out by combining a regression-based ANN with the Monte-Carlo method.

RESULTS

In the process of extracting images for dose analysis and identifying DCs, an under-dispersion tendency was observed. To rectify the discrepancy, classification ANN was employed to identify the results of DC detection. This approach led to satisfaction of Poisson distribution criteria by 32 out of the initial pool of 35 data points. In the subsequent stage, dose-response curves were constructed using data from 25 donors. Data provided by the remaining five donors served in performing dose estimations, which were subsequently calibrated by incorporating a regression-based ANN. Of the 23 points, 22 fell within their respective confidence intervals at p < .05 (95%), except for those associated with doses at levels below 0.5 Gy, where accurate calculation was obstructed by numerical issues. The accuracy of dose estimation has been improved for all radiation levels, with the exception of 1 Gy.

CONCLUSIONS

This study successfully demonstrates a high-precision dose estimation method across a general range up to 4 Gy through fully automated detection of DCs, adhering strictly to Poisson distribution. Incorporating multiple ANNs confirms the ability to perform fully automated radiation dose estimation. This approach is particularly advantageous in scenarios such as large-scale radiological incidents, improving operational efficiency and speeding up procedures while maintaining consistency in assessments. Moreover, it reduces potential human error and enhances the reliability of results.

Persons chronically exposed to low doses of ionizing radiation: A cytogenetic dosimetry study

The dosimetry and control of exposure for individuals chronically exposed to ionizing radiation are important and complex issues. Assessment may be optimized by evaluating individual adaptation and radiosensitivity, but it is not possible for a single model to account for all relevant parameters. Our goal was to develop approaches for the calculation of doses for persons chronically exposed to ionizing radiation, taking their radiosensitivities into consideration. On the basis of ex vivo radiation of blood samples, dose-effect models were constructed for dose ranges 0.01-2.0 and 0.01-0.4 Gy, using different cytogenetic criteria. The frequencies of "dicentric chromosomes and rings" at low doses are too low to have predictive value. The different responses of subjects to radiation made it possible to categorize them according to their radiosensitivities and to generate separate dose-effect curves for radiosensitive, average, and radioresistant individuals, reducing the amount of error in retrospective dosimetry.

International Comparison Exercise for Biological Dosimetry after Exposures with Neutrons Performed at Two Irradiation Facilities as Part of the BALANCE Project

In the case of a radiological or nuclear event, biological dosimetry can be an important tool to support clinical decision-making. During a nuclear event, individuals might be exposed to a mixed field of neutrons and photons. The composition of the field and the neutron energy spectrum influence the degree of damage to the chromosomes. During the transatlantic BALANCE project, an exposure similar to a Hiroshima-like device at a distance of 1.5 km from the epicenter was simulated, and biological dosimetry based on dicentric chromosomes was performed to evaluate the participants ability to discover unknown doses and to test the influence of differences in neutron spectra. In a first step, calibration curves were established by irradiating blood samples with 5 doses in the range of 0-4 Gy at two different facilities in Germany (Physikalisch-Technische Bundesanstalt [PTB]) and the USA (the Columbia IND Neutron Facility [CINF]). The samples were sent to eight participating laboratories from the RENEB network and dicentric chromosomes were scored by each participant. Next, blood samples were irradiated with 4 blind doses in each of the two facilities and sent to the participants to provide dose estimates based on the established calibration curves. Manual and semiautomatic scoring of dicentric chromosomes were evaluated for their applicability to neutron exposures. Moreover, the biological effectiveness of the neutrons from the two irradiation facilities was compared. The calibration curves from samples irradiated at CINF showed a 1.4 times higher biological effectiveness compared to samples irradiated at PTB. For manual scoring of dicentric chromosomes, the doses of the test samples were mostly successfully resolved based on the calibration curves established during the project. For semiautomatic scoring, the dose estimation for the test samples was less successful. Doses >2 Gy in the calibration curves revealed nonlinear associations between dose and dispersion index of the dicentric counts, especially for manual scoring. The differences in the biological effectiveness between the irradiation facilities suggested that the neutron energy spectrum can have a strong impact on the dicentric counts.

Biodose Tools: an R shiny application for biological dosimetry

INTRODUCTION

In the event of a radiological accident or incident, the aim of biological dosimetry is to convert the yield of a specific biomarker of exposure to ionizing radiation into an absorbed dose. Since the 1980s, various tools have been used to deal with the statistical procedures needed for biological dosimetry, and in general those who made several calculations for different biomarkers were based on closed source software. Here we present a new open source program, Biodose Tools, that has been developed under the umbrella of RENEB (Running the European Network of Biological and retrospective Physical dosimetry).

MATERIALS AND METHODS

The application has been developed using the R programming language and the shiny package as a framework to create a user-friendly online solution. Since no unique method exists for the different mathematical processes, several meetings and periodic correspondence were held in order to reach a consensus on the solutions to be implemented.

RESULTS

The current version 3.6.1 supports dose-effect fitting for dicentric and translocation assay. For dose estimation Biodose Tools implements those methods indicated in international guidelines and a specific method to assess heterogeneous exposures. The app can include information on the irradiation conditions to generate the calibration curve. Also, in the dose estimate, information about the accident can be included as well as the explanation of the results obtained. Because the app allows generating a report in various formats, it allows traceability of each biological dosimetry study carried out. The app has been used globally in different exercises and training, which has made it possible to find errors and improve the app itself. There are some features that still need consensus, such as curve fitting and dose estimation using micronucleus analysis. It is also planned to include a package dedicated to interlaboratory comparisons and the incorporation of Bayesian methods for dose estimation.

CONCLUSION

Biodose Tools provides an open-source solution for biological dosimetry laboratories. The consensus reached helps to harmonize the way in which uncertainties are calculated. In addition, because each laboratory can download and customize the app's source code, it offers a platform to integrate new features.

High-precision automatic identification method for dicentric chromosome images using two-stage convolutional neural network

Dicentric chromosome analysis is the gold standard for biological dose assessment. To enhance the efficiency of biological dose assessment in large-scale radiation catastrophes, automatic identification of dicentric chromosome images is a promising and objective method. In this paper, an automatic identification method for dicentric chromosome images using two-stage convolutional neural network is proposed based on Giemsa-stained automatic microscopic imaging. To automatically segment the adhesive chromosome masses, a k-means based adaptive image segmentation and watershed segmentation algorithm is applied. The first-stage CNN is used to identify the dicentric chromosome images from all the images and the second-stage CNN works to specifically identify the dicentric chromosome images. This two-stage CNN identification method can effectively detects chromosome images with concealed centromeres, poorly expanded and long-armed entangled chromosomes, and tricentric chromosomes. The novel two-stage CNN method has a chromosome identification accuracy of 99.4%, a sensitivity of 85.8% sensitivity, and a specificity of 99.6%, effectively reducing the false positive rate of dicentric chromosome. The analysis speed of this automatic identification method can be 20 times quicker than manual detection, providing a valuable reference for other image identification situations with small target rates.

Dicentric chromosome assay using a deep learning-based automated system

The dicentric chromosome assay is the "gold standard" in biodosimetry for estimating radiation exposure. However, its large-scale deployment is limited owing to its time-consuming nature and requirement for expert reviewers. Therefore, a recently developed automated system was evaluated for the dicentric chromosome assay. A previously constructed deep learning-based automatic dose-estimation system (DLADES) was used to construct dose curves and calculate estimated doses. Blood samples from two donors were exposed to cobalt-60 gamma rays (0-4 Gy, 0.8 Gy/min). The DLADES efficiently identified monocentric and dicentric chromosomes but showed impaired recognition of complete cells with 46 chromosomes. We estimated the chromosome number of each "Accepted" sample in the DLADES and sorted similar-quality images by removing outliers using the 1.5IQR method. Eleven of the 12 data points followed Poisson distribution. Blind samples were prepared for each dose to verify the accuracy of the estimated dose generated by the curve. The estimated dose was calculated using Merkle's method. The actual dose for each sample was within the 95% confidence limits of the estimated dose. Sorting similar-quality images using chromosome numbers is crucial for the automated dicentric chromosome assay. We successfully constructed a dose-response curve and determined the estimated dose using the DLADES.

Chronic human exposure to ionizing radiation: Individual variability of chromosomal aberration frequencies and G0 radiosensitivities

Bio-monitoring of human radiation exposure is based, as a rule, on a single analysis of chromosomal aberrations. Factors such as radiosensitivity, adaptation, and the stability of cytogenetic indices are not taken into account. We studied frequency of chromosome aberrations (FCA) and G₀ chromosome radiosensitivity following in vitro γ-exposure, over a 2.5-year period, for 129 residents of the Dolon settlement, part of the extreme radiation risk zone, Semipalatinsk nuclear test site region, Kazakhstan. Radiosensitivity was evaluated on the basis of FCA and dose assessment by physical dosimetry. FCA was 3-fold higher in Dolon inhabitants as in the control group (p ≤ 0.01). The average coefficient of variability of spontaneous FCA was 31 %. In 20 % of the subjects, it was very high (50-70 %). Individual dose estimation in a single study in such individuals may lead to significant errors. Individual G₀-chromosomal radiosensitivity showed less variation (18.7 %). Chronic low-dose irradiation was an adaptive factor to the damaging dose (1 Gy). Three methods of individual radiosensitivity assessment were considered, based on: G₀-chromosomal radiosensitivity under additional in vitro γ-radiation; FCA and average dose per year; FCA and total dose received during years of residence in a radiocontaminated settlement, according to physical dosimetry. There is a significant difference in response (FCA) between radiosensitive and radioresistant individuals. This should be taken into account in individual dosimetry and risk assessment of radiation exposure.

Clinical Applications of Biological Dosimetry in Patients Exposed to Low Dose Radiation Due to Radiological, Imaging or Nuclear Medicine Procedures

Radiation dosimetric biomarkers have found applications beyond radiation protection area and now are actively introduced into clinical practice. Cytogenetic assays appeared to be a valuable tool for individualized quantifying radiation effects in patients, with high capability for assessing genotoxicity of various medical exposure modalities and providing meaningful radiation dose estimates for prognoses of radiation-related cancer risk. This review summarized current data on the use of biological dosimetry methods in patients undergoing various medical irradiations to low doses. The highlighted topics include basic aspects of biological dosimetry and its limitations in the range of low radiation doses, and main patterns of in vivo induction of radiation biomarkers in clinical exposure scenarios, occurring in X-ray diagnostics, computed tomography, interventional radiology, low dose radiotherapy, and nuclear medicine (internally administered ¹³¹I and other radiopharmaceuticals). Additionally, several specific issues, examined by biodosimetry techniques, are analysed, such as contrast media effect, radiation response in pediatric patients, impact of magnetic resonance imaging, evaluation of radioprotectors, detection of patients' abnormal intrinsic radiosensitivity and dose estimation in persons involved in medical radiation incidents. A prognosis of possible directions for further improvements in this area includes the automation of cytogenetic analysis, introduction of molecular biodosimeters and development of multiparametric biodosimetry platforms. A potential approach to the advanced biodosimetry of internal exposure and/or low dose external irradiation is suggested; this can be a multiparametric platform based on the combination of the γ-H2AX foci, dicentric, and translocation assays, each applied in the optimum postexposure time range, with the amalgamation of the dose estimates. The study revealed the necessity of further research, which might clarify medical radiation safety concerns for patients via using stringent biodosimetry methodology.