Real-case applications of luminescence retrospective dosimetry in small-scale radiological accidents

Abstract

Retrospective dosimetry based on thermoluminescence (TL) and optically stimulated luminescence (OSL) using personal belongings was applied to a series of small‑scale radiological accidents that occurred in South Korea between 2015 and 2024. Display glasses and surface‑mounted resistors (SMR) extracted from mobile phones and electro-personal dosimeters (EPDs) of exposed individuals were analyzed using measurement protocols employed in multiple interlaboratory comparison (ILC) exercises. The reconstructed doses ranged from below the detection limit to several grays. In low‑dose cases (≤ 320 mGy), TL/OSL estimates aligned with those obtained from the dicentric chromosome assay (DCA). In contrast, discrepancies emerged in some high‑dose cases: in one accident, the mobile‑phone‑derived dose substantially exceeded the DCA estimate, whereas in another, the TL/OSL analysis revealed an exposure that DCA failed to detect. These findings confirm that luminescence techniques cannot yet serve as a stand‑alone tool for representing individual doses, but they can provide critical, scenario‑dependent complementary information to existing retrospective methods. The study also discusses the potential of the techniques and key considerations regarding regulatory and technical perspectives for their practical implementation in real-world scenarios.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-52162-x.

Introduction

Retrospective dosimetry refers to a suite of techniques used to estimate the radiation dose received by individuals from past exposures, such as radiological accidents, where conventional dosimetry methods were absent or insufficient. These techniques primarily focus on quantifying changes in the physical and chemical properties of materials induced by ionizing radiation. Key methods include biological dosimetry, which evaluates DNA damage in blood samples; electron paramagnetic resonance (EPR), which measures radiation-induced free radicals in materials like teeth, bones, and fingernails; and TL/OSL, which detects luminescence emitted from trapped charges in specific materials^1,2.

Among these, TL/OSL techniques stand out for their faster measurement times and lower detection limits compared to the other methods, making them particularly well-suited for triage during the initial stages of large-scale radiation accidents³. Promising materials for TL/OSL dosimetry include smart chip cards from credit or ID cards⁴, SMR⁵, display glasses⁶, and IC chips from mobile phones⁷. Additionally, other materials such as tobacco, salt, dental ceramics, drugs, and pharmaceuticals have been explored for their potential use in this field^8–11.

The effectiveness of retrospective dosimetry has been demonstrated in numerous real-world accidents. DCA is regarded as the reference cytogenetic biodosimetry method because of its high specificity for ionizing radiation and its standardized quantitative dose-estimation framework. It is extensively used to estimate individual radiation doses, triage exposed persons, and support medical management. Notable examples include the Chernobyl disaster^12,13, the Fukushima accident¹⁴, the Istanbul Co-60 accident¹⁵, the Goiânia Cs-137 accident^16–18, the Tokaimura criticality accident^19,20, the exposures in the Marshall Islands due to fallout²¹, and the Fleurus irradiator accident²². EPR has also played a key role in post-accident assessments: following Chernobyl^23,24 and Fukushima^25,26, tooth enamel EPR was applied for long-term dose reconstruction, while EPR using fingernail or bone was used in localized exposure scenarios, such as the Chilca²⁷, Lia²⁸, and Nueva Aldea²⁹ accidents to estimate localized doses for medical planning. TL/OSL techniques have primarily been applied to evaluate environmental doses in large-scale accidents like Chernobyl^30,31 and Fukushima^32,33, often by analyzing building materials such as bricks and ceramics near contaminated zones.

Despite the extensive literature on the application of retrospective dosimetry, real-world cases involving TL/OSL analysis of personal belongings remain scarce or unreported. This study aims to fill that gap by presenting dose assessments conducted using TL/OSL techniques in actual radiation exposure incidents that occurred in South Korea between 2015 and 2024. These efforts were made possible through collaboration between the Korea Institute of Nuclear Safety (KINS), the Korea Atomic Energy Research Institute (KAERI), Korea Institute of Radiological and Medical Sciences (KIRAMS), and the technical support of the European Radiation Dosimetry Group (EURADOS) WG-10 network. Furthermore, we explore the potential of the techniques in small-scale radiation accidents, highlight key considerations in their application, and discuss their implications from a regulatory perspective.

Materials and methods

Regulation and dose evaluation procedure

KINS has been investigating radiation overexposure incidents in South Korea, typically identified through reports of suspected events or confirmed exceedances of occupational dose limits in the workplace. Investigations include document reviews, on-site inspections, interviews, and retrospective dose assessments. When personal dosimeter data are unavailable, unreliable, or insufficient for reconstructing the exposure scenario, individual dose estimates are confirmed by a multidisciplinary expert committee using multiple complementary data sources, as presented in Table 1. This indicates that the retrospective dosimetry techniques have been officially integrated into the regulatory framework, with Table 2 summarizing their practical application over the past decade. The finalized doses are maintained as official national records and used mainly for regulatory purposes and, in high-exposure cases, to support medical management.

Table 1.

Hierarchy of data sources considered for dose assessment in radiological accidents.

Priority	Data	Details
1	Primary individual monitoring	Reading data from TLD (TL dosimeter), OSLD (OSL dosimeter), GD (glass dosimeter), etc
2	Secondary individual monitoring	Record from secondary dosimeters (e.g. EPD, automated dosimetry record (ADR), etc.) Result from retrospective dosimetry techniques (e.g., DCA, FISH, TL/OSL, EPR, etc.) Clinical symptoms, including health examinations
3	Co-worker dose records	Dosimetry data from workers performing the same tasks, based on TLD, OSLD, and GD readings
4	Workplace-based dose reconstruction	Numerical and experimental dose reconstruction incorporating exposure geometry, duration, and source characteristics
5	Personal exposure history	Historical dose data, including the maximum recorded dose over the past five years

Table 2.

Application frequency of retrospective dosimetry techniques in radiation exposure investigations conducted over the past 10 years.

Year	DCA	FISH	TL/OSL	EPR	Computational
2014	34	8
2015	22	6	2	1	6
2016	16	3	1	1	7
2017	14		2
2018	12		1
2019	9	4	2		2
2020	8		1
2021	12	1			1
2022	4	1	2
2023	4				1
2024	9		1	3	3
Sum	144	23	12	5	20

All methods presented in Table 1 were conducted in accordance with the guidelines and regulations outlined in the ‘Regulations on the Evaluation and Management of Individual Radiation Exposure Dose’ issued by the Nuclear Safety and Security Commission (NSSC), after written informed consent had been obtained from all subjects or their legal guardians. The health examinations included complete blood count (CBC) testing for radiation workers and were performed according to standard procedures at authorized medical institutions. Cytogenetic assays were performed at KIRAMS, a nationally accredited medical testing institution certified by the Korea Laboratory Accreditation Scheme in accordance with ISO 15,189 requirements. For both DCA and fluorescence in situ hybridization (FISH), phytohemagglutinin-stimulated T lymphocytes were cultured for 48 h. For DCA, metaphase chromosomes were Giemsa-stained, and dicentric chromosomes were scored under the assumption of whole-body exposure, in accordance with ISO 19,238. For FISH, whole-chromosome painting probes for chromosomes 1, 2, and 4 were used to support long-term or retrospective exposure assessment, in accordance with ISO 20,046. Owing to the relatively long half-life of the aberrations compared to the measurement timeline, the influence of time elapsed since irradiation was not considered in the dose estimation³⁴. All measurement data derived from human biological materials were anonymized prior to analysis, and the study was approved by the KIRAMS Institutional Review Board (IRB No. 2025-12-006-001). Moreover, due to KINS’s guidelines on confidentiality in incident investigations, this study presents only limited information related to dose assessments, such as a general overview of selected incidents, TL/OSL dose evaluation results obtained from personal belongings, and partial comparisons with other dosimetric techniques. Detailed methodologies of other dose assessment techniques are beyond the scope of this study.

TL/OSL techniques

TL/OSL techniques were applied to personal belongings that were either carried at the time of exposure or presumed to have been carried, such as mobile phones and EPDs. These items were voluntarily provided by the exposed individuals and their employers with informed consent. During the collection process, the data inside was typically backed up and protected before acquisition.

In the laboratory, the items were disassembled under red light conditions, without any heat treatment, and target samples were extracted and pretreated according to the protocols described later. The primary materials analyzed were display glass (the bottom substrate of the display unit) and SMRs. Measurements using mobile phone components were conducted at KAERI, while those using EPD components were carried out at KINS. Figure 1 shows one of the disassembled mobile phones along with the tools used for disassembly.

Fig. 1.

A disassembled mobile phone and tools used for dose reconstruction in radiation accidents.

Measurements were conducted using automated TL/OSL readers available at each institute. At KAERI, a Risø TL/OSL-DA-20 system equipped with a DASH (Detection and Stimulation Head) upgrade was used. Optical stimulation was performed using an integrated blue LED module (470 nm, 20 nm bandwidth) capable of delivering up to 80 mW/cm². The emitted luminescence was filtered through a 7.5 mm Hoya U-340 filter (effective in the 280–380 nm range) and detected using a blue-sensitive photomultiplier tube (PMT; Electron Tube PDM9107Q-AP-TTL-03). Samples were irradiated with a built-in beta source (Sr-90/Y-90). At KINS, a similar Risø TL/OSL-DA-20 reader was employed, also configured with 470 nm blue LEDs (80 mW/cm²), a Hoya U-340 filter, a Bialkali EMI 9235QB PMT, and an identical type of beta source.

The luminescence signals from the analyzed materials were converted into absorbed dose using a single-aliquot regenerative-dose protocol with the built-in beta source of each reader. The beta source dose rate was calibrated in terms of air kerma using display glass and resistor samples previously irradiated with a Cs-137 gamma source at the secondary standard dosimetry laboratory of KAERI. A 2 mm PMMA buildup layer was applied during irradiation to ensure charged-particle equilibrium. The calibration glass samples were extracted from a Samsung Galaxy S3 mobile phone display, while fresh 0402-size reel resistors were used for resistor calibration; both types of samples were annealed up to 450 °C prior to irradiation. Separate material-specific dose rates were established for glass and resistor samples, and source decay was corrected for according to the elapsed time after calibration.

Display glasses

For display glass samples, a pre-bleaching TL protocol was applied based on previously published methodologies^6,35. The glass pieces were initially immersed in acetone for 15 min using an ultrasonic bath to remove any adhesive residues. Following this, they were etched with 48% hydrofluoric acid (HF) for 2 to 4 min, depending on the thickness, to reduce the native TL signal³⁶. A second ultrasonic cleaning in ethanol for 15 min was performed to eliminate any remaining impurities. The samples were then cut and selected to a size of approximately 5 × 5 mm².

Prior to TL measurement, optical pre-bleaching was conducted using the reader’s built-in blue LEDs for 500 s at room temperature. The TL signals were subsequently recorded while heating the samples at a constant rate of 2 °C/s up to 450 °C. The thermal background was subtracted by acquiring a subsequent TL glow curve without prior irradiation. All measurements were carried out under nitrogen gas flow. The TL signal was integrated over the temperature range of 100–250 °C. Regenerative doses ranging from several hundred mGy to a few Gy were applied between measurements.

The final dose (D_glassTL) was calculated by correcting the uncorrected dose (D_unc.)—obtained from the ratio of luminescence signals to the corresponding regenerative doses—using fading (f) and zero-dose (D_BG) correction factors, as expressed in Eq. (1). Accordingly, the total uncertainty in dose estimation incorporates contributions from the luminescence intensity, regenerative dose calibration, zero-dose correction, and fading compensation. The minimum detectable dose (MDD) assessed shortly after exposure has been reported to be approximately 76–90 mGy^35,36.

1

Display glass samples were primarily classified into Categories A to D based on the shape of their TL glow curves, which are known to exhibit distinct dosimetric characteristics⁶. After determining the category of each sample, corresponding fading factors and zero-dose values were applied. In this study, only Categories A and B glass were identified. Because Category B glass predominantly exhibits red emission around 600 nm, the U‑340 filter was not used during measurement.

Surface-mounted resistors

Due to the limited number of suitable resistors available on mobile phone printed circuit boards (PCBs), only one to three sample cups could be prepared per device. Typically, around ten resistors of the 0402 package (1.00 × 0.5 mm²) were placed in a single cup. In cases where insufficient 0402 resistors were available, smaller 0201 resistors (0.60 × 0.3 mm²) were used. Resistor samples were cleaned by immersing them in acetone and sonicating for 15 min to remove any remaining glue.For the OSL protocol, a pre-heat step at 120 °C for 10 s was applied to eliminate unstable signal components, following the methodology described by Bassinet et al.³⁷. OSL readout was performed at 100 °C for 30 s. Signal integration was conducted over the 0–6 s range, with background subtraction using the 6–12 s interval. Signal fading corrections were applied using a universal fading curve as suggested by Woda et al³⁸. Accordingly, the total uncertainty in the OSL dose estimates for resistors (D_resistorOSL), as described in Eq. (2), included contributions from luminescence intensity, regenerative dose calibration, and fading correction.

2

The TL protocol for resistors also included a 10 s preheating step at 120 °C. TL measurements were performed up to 400 °C at 2 °C/s. A second, non-irradiated TL measurement was used for thermal background subtraction, and nitrogen gas was purged during the readout. The U-340 filter was removed during luminescence measurements to enhance signal intensity. TL signals were integrated over the 100–200 °C range. Corrections were applied for fading, intrinsic zero doses, and dose overestimation (C) based on literature-reported data^38,39. Thus, the overall uncertainty in resistor TL dose estimation (D_resistorTL), as defined in Eq. (3), accounts for luminescence intensity, calibration dose, zero-dose correction, fading rate, and overestimation correction.

3

The MDD can be estimated to range from 20–30 mGy for the OSL protocol and 30–70 mGy for the TL protocol, based on a three-sigma threshold derived from zero dose signal variation^38,40,41.

Fading corrections

In most cases, exposure was confirmed through the quarterly reported personal dose. As a result, the exact time of exposure remained unclear, typically falling within one to three months prior to the measurement. For incidents with known exposure times, a single fading correction was applied; for those with unknown times, multiple fading corrections were used. For the multiple fading correction, the average and standard deviation were calculated based on values at the beginning, middle, and end points of the assumed exposure period. Specifically, the approximate fading ratios (i.e., the ratio of the remaining signal to the initial signal) for one and three months from the irradiation are as follows: 51% and 45% for the glass TL, 48% and 41% for the resistor TL, and 37% and 29% for the resistor OSL^36,38,42.

Reliability of the techniques

Until now, various luminescence dosimetry protocols have been developed using personal belongings. Studies have proposed the use of TL from surface-mounted inductors^43,44, OSL from smart chip cards^4,45, TL/OSL from integrated circuit (IC) chips extracted from PCBs^7,46, and OSL from pharmaceuticals^10,47. In addition, different measurement techniques have been proposed for the same type of material—for example, residual TL⁴⁸ and PTTL (photo-transferred TL) for resistors⁴⁹, or PTTL and TaOSL (thermally assisted OSL) for glass samples^50,51.

While these approaches demonstrate considerable potential, most have not yet undergone comprehensive reproducibility assessments through inter-laboratory comparison studies. As will be discussed later, the components of personal belongings are often available only in limited quantities, restricting the number of measurable samples. Therefore, rather than applying a broad range of newly developed protocols, this study employed protocols that have already been validated through comparative studies.

The measurement protocols described in the “TL/OSL techniques” section have been used in ILC exercises organized by international/domestic networks such as the International Atomic Energy Agency (IAEA), EURADOS WG-10, Running the European Network of Biological and Retrospective Physical Dosimetry (RENEB), and Korea Retrospective Dosimetry Network (KREDOS). These exercises have ranged from laboratory-based blind tests using standard irradiation setups^37,52 to field exercises simulating hypothetical exposure accident scenarios^40,42,53. To better reflect real-world conditions, some exercises incorporated variables such as different mobile phone models, post-exposure use, and light exposure of phones, the application of identical protocols across multiple laboratories, and varying time intervals between exposure and measurement depending on the laboratory. Consequently, the results of these exercises were considered to provide a minimum level of reliability to luminescence techniques, which currently lack internationally standardized protocols.

Results and discussions

Statistics on the use of luminescence techniques

Table 3 summarizes the cases in which luminescence dosimetry using personal belongings was applied to radiation-exposed workers in South Korea between 2015 and 2024. Most of the incidents occurred during industrial non-destructive testing (NDT) operations involving gamma radiography using sources such as Se-75, Ir-192, and Co-60. Additionally, one incident involving X-ray irradiation devices was identified. On average, the TL/OSL techniques have been employed in approximately one case per year, with more than 20 items analyzed in total. These items were collected from different individuals involved in each incident. Among the disassembled mobile phones, Samsung models were the most frequently encountered, followed by those from LG and Apple. For EPDs, devices manufactured by Iljinrad Co., Ltd. were used.

Table 3.

Summary of dose assessments using luminescence dosimetry on personal belongings from radiation exposure incidents in South Korea (2015–2024).

Case	Year	ID	Model	Protocol	Time since irradiation	TL/OSL dose (mGy)	DCA dose (mGy)	Personal dosimeter (mSv)
1a,b)	2015	1-1	Unidentifiable	Resistor OSL	14 d	210 (131–288)	< MDD	155
2b)	2015	2-1	Unidentifiable	Resistor OSL	30 d	< MDD	168 (27–340)	146
3b)	2016	3-1	LG G3	Resistor TL Resistor OSL Glass TL	120 d	< MDD	< MDD	82
4b)	2017	4-1	Samsung Galaxy Note 5	Resistor TL	30–90 d	10,811 (9,446–12,176)	1,191 (865–1,541)	1201)
		4-2	Samsung Galaxy S7	Resistor TL	30–90 d	3,356 (2,488–4,224)	672 (60–840)	2.61)
		4-3	LG V20	Resistor TL	30–90 d	341 (276–405)	115 (30–272)	< MDD
		4-4	Samsung Galaxy S7	Resistor TL	30–90 d	312 (280–345)	171 (18–531)	1.4
		4-5	Samsung Galaxy A7	Resistor TL	30–90 d	< MDD	153 (44–306)	3.0
		4-6	LG Stylus 2	Resistor TL	30–90 d	< MDD	153 (44–306)	5.0
		4-7	Samsung Galaxy Note 4	Resistor TL	30–90 d	< MDD	188 (82–345)	< MDD
		4-8	Samsung Galaxy A8	Resistor TL	30–90 d	< MDD	188 (82–345)	0.3
		4-9	Apple iPhone 6 s	Resistor TL	30–90 d	< MDD	115 (30–273)	0.3
		4-10	Samsung Galaxy S6	Resistor TL	30–90 d	< MDD	171 (66–333)	0.7
5b)	2017	5-1	Samsung Galaxy S6	Resistor TL	30–90 d	211 (126–296)	115 (30–273)	4.6
		5-2	Samsung Galaxy Note 3 Neo	Resistor TL	30–90 d	241 (101–382)	188 (82–345)	2.3
6b)	2018	6-1	ILJINRAD Co., Ltd. CLOVER ADR-20	Resistor OSL	30–90 d	< MDD	< MDD	< MDD1)
7b)	2019	7-1	Apple iPhone 7	Glass TL	30–60 d	< MDD	115 (0–504)	-2)
8b)	2019	8-1	ILJINRAD Co., Ltd. CLOVER ADR-20	Resistor OSL	90–120 d	176 (159–193)	115 (30–273)	118
9c)	2020	9-1	ILJINRAD Co., Ltd. CLOVER ADR-20	Resistor TL	60–90 d	136 (117–150)	115 (30–273)	3381)
10a)	2022	10-1	Samsung Galaxy S9	Resistor OSL	70–99 d	145 (103–187)	115 (30–273)	-2)
11b)	2022	11-1	ILJINRAD Co., Ltd. CLOVER ADR-20	Resistor OSL	90–120 d	203 (0–412)	293 (64–674)	-2)
12d)	2024	12-1	Samsung Galaxy S10 plus	Resistor OSL	8 d	380 (240–521)	< MDD	-3)

For comparison, dose estimates from DCA and personal dosimeter readings are also provided. Annotations indicate cases where the personal dosimeter was: 1) suspected to be unworn, 2) lost, or 3) not required to be worn. The radiation sources involved in each case were: a) Se-75, b) Ir-192, c) Co-60, or d) an X-ray. The values given in parentheses indicate the uncertainty range.

The reported doses (i.e., TL/OSL dose) represent the average value calculated from more than two sample cups per item. The uncertainty range was propagated to account for the sample variation and, in cases where the time of exposure was given as a range, for the multiple fading corrections applied. For comparison, the dose estimates obtained using DCA and a personal dosimeter are presented in the rightmost columns, while the results from the FISH technique are provided in Supplementary Table S1. It should be noted that doses reconstructed from resistors, display glass, and blood samples are expressed as absorbed dose (Gy), whereas the personal dosimeter recordings are reported as personal dose equivalent (i.e., Hp(10)); therefore, a direct comparison between these quantities is not appropriate. The detector materials of personal dosimeters varied depending on the service provider (e.g., Parasonic (Li₂B₄O₇:Cu and CaSO₄:Tm) and ILJINRAD (LiF:Mg,Cu,Si) systems). The MDDs for DCA and personal dosimeter are approximately 100 mGy and 0.1 mSv, respectively. Meanwhile, the MDDs of TL/OSL dose represent values assumed to be assessed immediately after exposure, as suggested in the “TL/OSL techniques” section. However, it should be noted that effective MDDs of TL/OSL may increase over time due to signal fading. For example, in Case 6, when considering the beginning and end points of the time since irradiation, the MDD increases from approximately 80 mGy to 100 mGy.

Preference of protocols

Case 3 was the only instance in which all three protocols were applied simultaneously, and the results were found to be consistent, with all dose estimates falling below the MDD. In contrast, in most other cases, only one protocol was applied, with protocols based on resistors being used more frequently than those based on glasses. This was partly due to the presence of flexible active matrix organic light-emitting diode (AMOLED) displays in some recent mobile phones, which lacked glass substrates altogether. In addition, the frequent presence of Category B glass, which had been less studied at that time, contributed to the preference for resistor-based protocols. The glass TL protocol was applied only in two cases throughout the study: the sample from ID 3-1 was identified as Category A, and the sample from ID 7-1 as Category B. For resistors, either the TL or OSL protocols were applied, depending on the signal fading. When the time elapsed was less than approximately two months, OSL was preferred due to its low zero-dose characteristics. For longer delays, TL was used, owing to its relatively lower signal fading properties.

Use of EPDs

An EPD is recommended or mandated to be worn during tasks, and its fixed wearing position allows for a more reasonable representation of whole-body dose compared to mobile phones. Although it provides real-time dose monitoring, in cases where the recorded doses are significantly lower than expected based on the exposure scenario, it is assumed that the device may have been misused, the dose record may have been incorrectly written or omitted, or the internal log of the device may have been overwritten or erased. In such instances, luminescence dosimetry techniques using EPD can be employed as a supplementary assessment tool³⁹, as shown in Cases 6,8,9, and 11. Although EPD data have been used for individual dose assessment (see Table 1), they were not included in Table 3 due to the limited comparability and interpretation challenges relative to other dose metrics. Compared to mobile phones, the use of EPD offered the technical advantage of having a greater number of resistors and larger-sized components than the 0402 type. In addition, EPDs tended to have more consistent device types over time, whereas mobile phone models frequently change. On the other hand, the monochrome display glass used in EPDs exhibited poor radiation sensitivity⁶. Besides, there are several considerations: the devices may be used interchangeably by different individuals depending on the workplace, and although the accumulated dose during routine calibration processes (e.g., every six months in South Korea) is typically below a few millisieverts, it may still contribute to the total dose and should therefore be taken into account. In this study, however, only EPDs without prior high-dose exposure histories were used, and any previous dose signals were assumed to have sufficiently faded and therefore not to significantly influence the assessment of recent exposures.

Dose comparison

Annual dose limit exceedances in TLD (i.e., personal dosimeter) readings were observed in Cases 1, 2, 3, 4 (ID 4-1), 8, and 9. However, in most instances listed in the table, the TLD was suspected to have been unworn, lost, not required due to the absence of a mandatory dosimeter policy, or even stored near the radiation source. As a result, dose estimation based solely on TLD readings was challenging. From this perspective, the TL/OSL and DCA methods have served as complementary tools, providing additional clues regarding potential radiation exposure. For instance, among the cases with TLD doses exceeding 100 mGy, exposure was also suspected based on TL/OSL results in Case 1 and on DCA findings in Case 2. In Cases 4, 8, and 9, both methods indicated evidence of exposure.

When comparing the TL/OSL and DCA methods, cases with TL/OSL doses below 320 mGy generally showed considerable consistency, with dose estimates falling within each other’s uncertainty ranges or exhibiting overlapping uncertainties, including expanded uncertainties with a coverage factor of k = 2 in some cases. In these cases, the ratio of TL/OSL dose to DCA dose, based on their mean values or MDDs, ranged approximately from 0.5 to 2. These results suggest that luminescence dosimetry can be used for cross-validation of existing methods. In contrast, significant discrepancies were observed in high-dose cases. Specifically, Case 4 exhibited TL/OSL doses approximately 3 to 9 times higher than the corresponding DCA doses, while Case 12 showed a dose roughly four times greater than the DCA detection limit. Detailed case study analyses of these two cases are presented in the following sections.

Case study [Case #4]

Overview

In January 2017, KINS received a report of radiation overexposure involving Worker 1 at a chemical plant under construction (Fig. 2). The personal dosimeter reading indicated an accumulated dose of 120.45 mSv. Medical examination revealed hematological abnormalities, including pancytopenia and aplastic anemia, requiring a bone marrow biopsy and emergency blood transfusions.

Fig. 2.

The chemical plant construction site where the overexposure accident occurred (Case #4).

A subsequent KINS investigation found poor working conditions and serious violations of radiation safety regulations. Worker 1 had performed NDT radiography using an Ir-192 source (< 20 Ci) during inspections of piping, tanks, and pipelines. Although both a legal dosimeter and a secondary dosimeter were required, they were found not to have been worn in most cases. The work was often performed in close proximity to the exposed source, sometimes at distances as short as 30 cm, with limited shielding and escape routes, especially at elevated locations, and shielding tools such as collimators and lead plates were rarely used. These problems were attributed to insufficient radiation protection education, inadequate operational training, and the lack of pre-job site inspections.

The investigation also identified possible exposure among other workers. Health screening (e.g., CBC) and cytogenetic analyses were conducted for 35 related personnel, of whom 9 individuals (Workers 2 to 10) showed notable findings and underwent additional dose assessments.

Dose assessment

Table 4 presents the various dose assessments for the ten workers in Case #4. Considering the exposure geometry and the accumulation of dose due to multiple tasks, a homogeneous whole-body exposure was assumed. Given Worker 1’s clinical symptoms—such as pancytopenia and aplastic anemia, which are typically associated with exposures exceeding 500 mGy⁵⁴—and the inconsistent statements regarding personal dosimeter usage (initially claiming it was worn to avoid administrative penalties, but later acknowledging that it was not worn during most tasks), the recorded dosimeter reading was considered unreliable. Biological dose assessments estimated 1191 mGy (range: 865–1541 mGy) using the DCA and 2695 mGy (range: 2278–3146 mGy) using FISH. EPR analysis of tooth enamel also indicated a dose exceeding 1000 mGy. Additionally, based on the worker’s interview regarding exposure geometry and workload, the numerically reconstructed effective dose was roughly estimated at ~ 994 mSv. In contrast, the dose reconstructed from the worker’s mobile phone (ID 4-1) ranged from 9.5 to 12 Gy, showing a substantial discrepancy from the other methods. However, testimony confirmed that Worker 1 had occasionally stored the mobile phone near the exposed radiation source, suggesting that the dose measured from the device may not be representative of the worker’s body dose.

Table 4.

Summary of dose assessments for Case #4.

Worker	TLD	Blood test (CBC)	DCA	TL/OSL (mobile phone)	EPR (tooth)	Workplace-based dose reconstruction
1	120	Pancytopenia Aplastic anemia	1,191	10,811	> 1000	994
2	MDD to 5	Thrombocytopenia Leukocytosis	672	3,356		5 to 1,307
3			115	341		0.2 to 4.5
4			171	312
5–10		No significant findings	115–188	< MDD

To illustrate individual trends, some of the data presented in Table 3 are reused. (Unit: mGy, except for TLD and workplace-based dose reconstruction, which are expressed in mSv).

The chromosomal analyses of the remaining nine workers (Workers 2 to 10) revealed signs of radiation-induced damage. Among them, three individuals (Workers 2 to 4) also exhibited abnormal clinical findings in CBC, including hematological abnormalities such as thrombocytopenia and leukocytosis. In particular, Worker 2—who performed radiography without wearing a personal dosimeter—showed the highest estimated dose among the three, with 672 mGy based on DCA and up to 1307 mGy according to workplace-based dose reconstruction. Interestingly, the assessments using the workers’ mobile phones also yielded dose estimates ranging from 300 mGy to 3.3 Gy for the same three individuals (ID 4-2 to 4-4), while the remaining six (ID 4-5 to 4-10) showed doses below the MDD. Luminescence dosimetry using Worker 2’s mobile phone indicated ~ 3.3 Gy, again representing the highest values among the three. Once again, according to the worker’s statement that the mobile phone has been occasionally left near the source, it is presumed that this condition led to the discrepancy between the DCA and the mobile phone dose estimates. These findings suggest that while the dose estimated from a mobile phone may not directly correspond to the whole-body dose, it can still exhibit dose-proportional characteristics in clinical and biological observations under specific exposure conditions.

TL/OSL method

The resistor TL protocol was applied to all mobile phone samples. Depending on the model, both 0402- and 0201-sized package types were extracted due to an insufficient number of resistors, and they were separated and measured according to size. In the results, no significant dependence on resistor size was observed, and the standard deviations among the sample cups ranged from 7 to 25%, depending on the mobile phone. These values exceeded the uncertainty of ~ 7% associated with the multiple fading correction for 1–3 months. No notable outliers deviating substantially from the averages were found in the dataset. Figure 3 shows the TL glow curves from some of the measured samples. The doses were evaluated using one-point calibration (i.e., a single regenerative dose), with the regenerative dose varying by mobile phone. The level of the regenerative doses ranged from 700 mGy to 3.5 Gy, depending on the estimated level of exposure as inferred from KINS’s investigation, including clinical symptoms. All measured samples exhibited a typical TL peak near 180 °C, commonly observed in resistors⁴⁴. Some samples showed a secondary peak around 320 °C in the first TL reading. However, this peak mostly disappeared in the following TL readings, suggesting that it likely corresponds to intrinsic background signals varying by sample.

Fig. 3.

TL glow curves of resistor samples extracted from mobile phones (ID 4-1 and 4-2) after the accident (solid line) and following a test dose (dashed line). Each glow curve was normalized with respect to the maximum TL intensity obtained from the accident.

Case study [Case #12]

Overview

In 2024, KINS received a report of a suspected radiation exposure accident involving two workers (Worker A and Worker B) during the maintenance of a radiation-generating device shown in Fig. 4. The workers encountered a stage-jamming issue inside a cabinet-type X-ray instrument (40 kV, 70 mA) used for silicon wafer inspection and decided to examine the interior of the device prior to beginning repairs. Since the worker believed that the malfunction could not be observed with the power turned off, the examination was carried out with the device powered on, and the door opened, while the X-ray interlock remained engaged. Unfortunately, due to a failure in the shutter base interlock, the workers were unaware that X-rays were being emitted during the inspection. Over an approximately 15-min period, the workers manually examined the internal components and took photographs using a mobile phone. After the task, Worker A noticed edema in the fingers, which led to the suspicion of radiation exposure and subsequent reporting of the incident.

Fig. 4.

Cabinet-type X-ray instrument involved in the overexposure accident (Case #12).

Dose assessment

Table 5 presents the dose assessments for the workers in Case #12. Based on these, localized exposures on the order of several tens of Gy were suspected for both workers’ hands. During interviews, Worker A reported that their fingers had approached ~ 4.9 cm from the source center (Fig. 5a) and remained there for 1–2 s. Worker B stated that their fingers were positioned within 8–10 cm from the source (Fig. 5b) for approximately 4–6 s, and the mobile phone photography was conducted at a distance of 30–40 cm from the source (Fig. 5c) for approx. 8–13 s.

Table 5.

Summary of dose assessments for Case #12. To illustrate individual trends, some of the data presented in Table 3 are reused.

Worker	Clinical symptoms	DCA	TL/OSL (mobile phone)	EPR (tooth)	Workplace-based dose reconstruction
A	Pain, blistering, edema, maceration, vascular damage	< MDD	–	< MDD	~ 15 mSv (Effective dose) 68–123 Sv (local)
B	Erythema, edema, moist desquamation		380 mGy		~ 130 mSv (Effective dose) 26–30 Sv (local)

Fig. 5.

Schematic illustration of hand positions during internal examination of the instrument, based on worker interviews.

As the X-ray instrument was not subject to personal monitoring due to the low dose rate outside of the instrument, no personal dosimeter readings were available. Additionally, the DCA using blood and EPR using tooth enamel results were below the detection limit, and EPR analysis using fingernails could not be performed due to the workers’ extremely short nails. Consequently, the only available measurable dose estimate was derived from luminescence measurements of the mobile phone used during the accident.

Due to the limited dose information for the localized exposure of the extremities, both experimental dose reconstruction and numerical assessments were conducted, referencing the threshold doses associated with clinical symptoms as well as dose estimates derived from the luminescence technique. The exposure scenario was reenacted multiple times by the involved workers, and dose rate distributions near the X-ray source were measured using Al₂O₃:C dosimeters. Energy response corrections were applied to account for the enhanced sensitivity of the dosimeters at low photon energies. For Worker A, the localized skin equivalent dose was estimated at 68–123 Sv, with an effective dose of 15 mSv. For Worker B, the corresponding values were 26–30 Sv and 130 mSv. These estimates exhibit wide ranges, as they are primarily based on statements with considerable uncertainty. In addition, the discrepancy between the reconstructed skin dose and the mobile phone dose was likely attributable to the phone being used only during part of the task, at a greater distance from the source, with possible shielding by surrounding instruments.

TL/OSL method

As the mobile phone used during the accident did not contain a display substrate glass due to the presence of a flexible AMOLED display, only ~ 20 resistors of the 0402 type could be extracted and measured across two sample cups. Figure 6 shows the OSL decay curves of the two sample cups, and their standard deviation was ~ 37%, which is notably higher than the uncertainty of about 10–13% associated with individual measurements. Although the randomly extracted resistor samples from the PCB can be assumed to represent an average dose over the entire surface area of the mobile phone, the limited sample quantity and the potential partial shielding by the hand or phone housing are expected to result in notable deviations between sample cups, particularly under low-energy X-ray exposure conditions. Additionally, in this case, the time since irradiation was short and clearly defined (i.e., 8 days), and therefore a single fading correction factor of 0.51 was applied.

Fig. 6.

OSL decay curves of two resistor samples (S1 and S2) extracted from the mobile phone (ID 12-1) after the accident and following a 1 Gy test dose. Each decay curve was normalized with respect to the maximum OSL intensity obtained following the test dose irradiation.

Given that the average X-ray energy applied during the accident was estimated to be ~ 20 keV, two key factors were considered. First, since the resistor material is primarily composed of Al₂O₃, which exhibits more than three times higher response below 100 keV⁵⁵, there is a potential for dose overestimation. Second, the mobile phone housing, composed of materials such as glass and aluminum, could have attenuated the low-energy X-rays, leading to possible dose underestimation. To account for both effects, an energy response correction factor was experimentally derived using KAERI’s standard X-ray reference field (ANSI M30, mean energy: ~ 20 keV). The mobile phone was reassembled, and the same resistors were repositioned on the same PCB, after which a standard irradiation of 1 Gy (air kerma) was performed. The measured signal was then used to recalibrate the dose rate of the beta source in the Risø reader. The dose rate of the beta source was determined to be 6.2 mGy/s in Cs-137 equivalent air kerma and 4.6 mGy/s in M30 equivalent air kerma, corresponding to a relative response of approximately 1.35. After applying this energy correction, the dose was re-evaluated as ~ 380 mGy from ~ 505 mGy.

Discussions

Regulatory perspectives

From a regulatory perspective, personal belongings-based dose information cannot fully substitute for human body dose estimates. However, it can serve as the minimum accessible information under the following scenarios:

• When personal dosimeters are damaged, lost, not worn, or their use is uncertain.

• When dosimeter readings are unreasonably low compared to the exposure scenario

• When cross-validation is needed with results from other dosimetry techniques

• When assessing the possibility of localized exposure

• When the exposure scenario cannot be reconstructed or reproduced

Given the time and cost required for biological dosimetry, luminescence dosimetry may also serve as preliminary data to support early-stage triage decisions. For instance, in Case #4, health screenings and cytogenetic analyses were conducted for 35 workers, but if luminescence dosimetry had been performed in advance, it could have facilitated the prioritization or selective application of further dose assessments. This interpretation is supported by the results in Table 4. Although the TL/OSL doses tended to be higher than the DCA doses, such conservative estimates are often preferred in triage to avoid underestimating potential radiation exposure.

Moreover, considering the MDDs of luminescence dosimetry, it remains challenging to assess whether the annual effective dose limit (20–50 mSv) has been exceeded. However, it can be feasible to evaluate whether the threshold for the annual skin dose limit (500 mSv) has been surpassed. In industrial settings where the use of personal electronic devices during work is increasingly common—as exemplified by Case #12—this approach is expected to see wider application in the future.

Meanwhile, whether individuals involved in large-scale radiation emergencies would voluntarily submit their personal items for dose assessment remains uncertain. In contrast, in small-scale accidents, mobile phone collection and analysis were found to be more feasible, partly due to concerns over potential regulatory penalties, administrative actions, or increased public sensitivity to radiation-related health risks. This observation underscores the pragmatic advantage of applying such techniques for regulatory investigations.

Technical perspectives

From a technical standpoint, several improvements are needed to enhance the reliability and applicability of luminescence dosimetry in practical settings:

• As recent mobile phones often lack display glass due to the use of flexible AMOLED displays and contain fewer or smaller resistors, continued investigation and identification of next-generation radiation-sensitive marker materials are required.

• It is necessary to verify whether the dosimetric properties of marker materials established in previous studies remain valid for current electronic components.

• When small-sized samples are extracted, validation procedures—such as analyzing glow curve raw data—should be performed to confirm whether the extracted components are indeed the target material.

• Additional studies are needed to characterize the energy response of materials to low-energy X-rays, especially in light of the increasing prevalence of low-energy X-ray equipment. In addition, for low-energy photon exposures, the shielding effect of device housings should be taken into account.

• It is necessary to simplify the measurement protocol or make it easier to operate. This would enhance reproducibility across more laboratories and enable its implementation in situations requiring mass measurements, such as large-scale exposure scenarios.

• In this study, dose reconstruction based on the carrying position of personal items was not performed in most cases due to insufficient information on the accident scenarios. However, when detailed information, such as working posture, exposure geometry, source terms, exposure time, and the carrying positions of the items, is available, numerical techniques can be applied to convert measured doses into organ or whole-body doses. In this context, recent research has proposed technical approaches such as deformable mesh-type phantoms⁵⁶, artificial intelligence-based dose conversion methods⁵⁷, and ILC on dose conversion techniques⁵⁸.

• Discussions on standardized methods for handling outliers are essential, particularly in cases where different materials from the same mobile phone yield divergent dose values. This possibility has been noted in previous ILC studies^58,59. Accordingly, it may be appropriate to assign weighting factors—based on the reliability of the applied protocol—prior to averaging the dose values.

• A consistent framework for uncertainty estimation is required to address scenarios such as: (1) when the approach to uncertainty estimation varies depending on the applied protocols, or (2) when uncertainty should be estimated due to an unclear time of exposure.

• For laboratory-developed protocols to be reliably applied under diverse real-world scenarios, international cooperation, such as standardization of protocols and periodic intercomparison exercises, is essential.

Conclusions

The present study evaluated the applicability, field utility, and key considerations of luminescence dosimetry using personal belongings in the context of real radiological accidents that occurred in South Korea between 2015 and 2024. Among the 12 investigated cases, the reconstructed doses varied, ranging from below MDD to several Gy. In most cases, the estimated doses showed good agreement with results obtained from DCA. However, notable discrepancies were observed in a few high-dose cases, mostly under conditions of localized or inhomogeneous exposure. For instance, in one case, the dose derived from luminescence measurements significantly exceeded the DCA-based estimate, while in another, evidence of exposure that was not captured by either DCA or personal dosimeters was revealed through mobile phone analysis. These findings suggest that while luminescence dosimetry may not serve as a standalone definitive tool for dose determination, it holds substantial value as a complementary method in retrospective dose assessments. Finally, although research on luminescence-based dose reconstruction using personal electronic belongings has progressed steadily since the 2010s, no prior studies have reported its application to actual radiological accidents. The results of this study represent the first attempt to bridge that gap and demonstrate the practical potential of this technique in real-world settings. It is hoped that these findings will stimulate further development and broader implementation of luminescence dosimetry in radiation emergency response and retrospective exposure evaluation.

Author contributions

H.K. conducted the literature review, measurements, data analysis, visualization, drafted the initial manuscript, and contributed to review and editing. H.Y., A.R.K., J.T.L., H.L., and J.K. carried out site inspections, conducted interviews, summarized data, and performed both experimental and numerical dose reconstruction. H.K., Y.H.K., M.C., J.L., I.C., and S.K.L participated in measurements and data analysis. M.D. and C.W. provided technical support, including supervision, software assistance, and methodological guidance. All authors reviewed the manuscript.

Funding

The study was carried out under the National Research Foundation of Korea (NRF) grant funded by the Korean government (Ministry of Science and ICT) (RS-2022-00144350, RS-2022-00144210); the Korea Atomic Energy Research Institute (project number: 522410-25); and the Korea Institute of Nuclear Safety through the research project “Construction of System for Response against Radiological Accidents and Terrorism” (A4GB24410); and the Korea Institute of Radiological and Medical Sciences (No. 50445-2026).

Data availability

The datasets generated and/or analyzed during the current study are not publicly available due to restrictions related to consent for third-party disclosure of personal information of the individuals, but are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.