Sangsoo Han

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. 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. 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. 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.

회전익 드론(Rotary-wing drone)은 제자리비행(Hovering)이 가능하기 때문에 다양한 산업분야에 활용되고 있다. 하지만 회전익드론은 에너지 효율이 좋지 않으며, 이러한 문제점을 해결하기 위해 무인 드론스테이션(Drone-station)이 개발되고 있다. 본 논문은 드론 순환임무 시스템을 구현하고 기능을 검증하였다. 드론 순환임무 시스템은 다수의 드론과 무인 드론스테이션, 그리고 관제시스템으로 구성된다. 본 논문은 자율비행 드론, 자동 배터리 교체식 드론스테이션, 그리고 관제시스템을 실제로 제작하였으며, 드론 순환임무 시스템의 검증을 위해 드론과 드론스테이션을 대신할 수 있는 더미(Dummy) 프로그램을 함께 개발하였다. 본 논문은 제작된 실제 드론 및 드론스테이션, 그리고 더미 드론 및 드론스테이션을 이용하여 드론들의 임무중단 및 임무재개 상황을 모의실험 하였다. 실험결과, 드론 순환임무 시스템의 드론들은 끊임없이 비행이 연속되는 것을 확인했다.

Background: Recently, biological adsorbents have been developed for removing radionuclides from radioactive liquid waste due to their high selectivity, eco-friendliness, and renewability. However, since they can be damaged by radiation in radioactive waste, a method for estimating the bio-adsorbent performance as a time should consider the radiation damages in terms of their renewability. This paper aims to develop a simulation method that applies a deep learning technique to rapidly and accurately estimate the adsorption performance of bio-adsorbents when inserted into liquid radioactive waste.Materials and Methods: A model that describes various interactions between a bio-adsorbent and liquid has been constructed using numerical methods to estimate the adsorption capacity of the bio-adsorbent. To generate datasets for machine learning, Monte Carlo N-Particle (MCNP) simulations were conducted while considering radioactive concentrations in the adsorbent column.Results and Discussion: Compared with the result of the conventional method, the proposed method indicates that the accuracy is in good agreement, within 0.99% and 0.06% for the R2 score and mean absolute percentage error, respectively. Furthermore, the estimation speed is improved by over 30 times.Conclusion: Note that an artificial neural network can rapidly and accurately estimate the survival rate of a bio-adsorbent from radiation ionization compared with the MCNP simulation and can determine if the bio-adsorbents are reusable.

산림지역은 다양한 지형, 장애물 등의 요인에 따라 산불 발생 시에 비가시권에서 드론의 수동 조작을 통한 화재진압은 상당히 어렵다. 특히, 야간의 경우 안전상의 이유로 헬기 등의 항공지원이 불가능하며, 사람 또한 접근이 어려워 드론을 통한 화재진압의 필요성이 점차 커지고 있다. 본 연구에서는 비전문가도 쉽게 드론을 산불진압에 활용할 수 있는 드론 제어 기술을 제안하였다. 제안된 기술은 영상-열화상 이미지 정보 기반의 딥러닝을 활용하여 사용자가 지정한 화재원점을 지속적으로 추적하며, 화재원점으로 유도 비행을 수행한다. 유도비행 시, 충돌 회피 기능이 적용되었다. 또한, 유도기술은 영상기반 정밀착륙에 동시 적용되어 딥러닝으로 인한 임무 컴퓨터에서의 메모리 소모를 최소화하였다. 본 연구는 산불진압용 드론에 직접 활용될 수 있을 것으로 기대되며, 향후 사용자 친화적 드론 제어 시스템의 기술로 활용될 수 있을 것으로 기대된다.

일반적으로 산불진화용 무인기는 사람의 원격 조종에 의해 운영되고 있으나, 산불의 특성상 다수 지점의 화재진압 및 통신이 원활하지 않는 산림환경에서는 활용이 어렵다. 이에 무인기를 이용하여 산불진압 임무를 수행하기 위해서는 진화지점의 자동 예측을 통한 무인기 완전 자동화가 필요하다. 본 연구에서는 딥러닝 기술 중 하나인 Semantic Segmentation 기술을 이용하여 소화탄 투화지점 예측 알고리즘을 제안하였다. 제안된 방법을 통하여 기계학습에 사용되지 않은 독립 데이터를 통한 화재 지점 국지화 및 소화탄 투하지점의 추천 평가를 수행하였으며, 90% 이상의 정확도로 소화탄 투하 필요 지점을 예측하였다. 본 연구에서 개발된 방법은 향후 무인기를 이용한 화재진압 완전 자동화를 위해 활용될 수 있을 것으로 기대된다.

Bismuth-functionalized graphene oxide (Bi-GO) was successfully synthesized and showed both high iodide and iodate removal efficiencies from radioactive wastewater. Batch experiments for kinetic and selectivity tests were performed, respectively. Additional SEM, XRD, FT-IR, and XPS analyses were performed for characterization of a sorbent and bismuth on the GO surface and this confirmed that bismuth on the GO surface reacted with iodine species by surface complexation (or precipitation). Dominant surface species are BiOI and Bi(IO₃)₃ for iodide and iodate removal, respectively. After the selectivity test using a KCl background solution with varying concentrations, Bi-GO still showed higher removal efficiencies (≥95%) for both iodide and iodate than the commercial silver-exchanged zeolite (≥95% for iodide and ≤25% for iodate). Our study suggests the potential application of Bi on graphene-based materials for selective removal of both iodide and iodate from radioactive wastewater.

Batch adsorption, batch diffusion, and flow-through column experiments were conducted using groundwater and fractured rock collected in unsaturated zone to increase our understanding of sorption and transport behavior of radionuclides. Increasing K_d values were observed in the sequence ⁹⁰Sr, ⁹⁹Tc, and ³H regardless of the geological media tested. For all sorbing radionuclides, K_d values for the fracture-filling/coating material were observed to be higher than those for without fracture-filling/coating material regardless of the groundwater. These higher K_d values are the result of zeolite mineral in filling/coating material of fractured rock. The batch diffusion and flow-through column experiments were also conducted using the same fractured rock sample, and the results of diffusion and column experiments showed similar trend of radionuclide sorption and transport to sorption experiment. In this study, sorption K_d of radionuclide was determined and used to increase our understanding of radionuclide retardation through fracture-filling/coating materials.