WEKO3
アイテム
The development of a DNN-assisted track analysis system for CR-39-based space radiation dosimetry
https://repo.qst.go.jp/records/2001936
https://repo.qst.go.jp/records/2001936958d4163-19d7-42fd-9155-9f1a6556b04b
| アイテムタイプ | 学術雑誌論文 / Journal Article(1) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| 公開日 | 2025-12-02 | |||||||||
| タイトル | ||||||||||
| タイトル | The development of a DNN-assisted track analysis system for CR-39-based space radiation dosimetry | |||||||||
| 言語 | en | |||||||||
| 言語 | ||||||||||
| 言語 | eng | |||||||||
| 資源タイプ | ||||||||||
| 資源タイプ識別子 | http://purl.org/coar/resource_type/c_6501 | |||||||||
| 資源タイプ | journal article | |||||||||
| 著者 |
Hu Jun
× Hu Jun
× Kodaira Satoshi
|
|||||||||
| 抄録 | ||||||||||
| 内容記述タイプ | Abstract | |||||||||
| 内容記述 | CR-39 plastic nuclear track detectors are widely used in space radiation dosimetry due to their high sensitivity to charged particles and ability to record linear energy transfer (LET) information. Conventional CR-39 analysis tools can rapidly detect particle tracks in general dosimetry, however, they are not tailored for space radiation applications that require classification of valid tracks and extraction of morphology parameters to estimate LET, motivating the use of an automated approach. In this study, we propose a Deep Neural Network (DNN)-assisted track analysis system based on the Mask Region-based Convolutional Neural Network (R-CNN) framework to automate the detection and segmentation of particle tracks in CR-39 detectors.The model was trained using a transfer learning strategy with weights pretrained on the MS COCO dataset and fine-tuned on annotated CR-39 images. It achieved mean average precision (mAP) values of 81.5 % for detection and 81.4 % for segmentation. The system demonstrated robust performance in handling overlapping tracks and significantly improved analysis efficiency, reducing the processing time for a typical sample from several hours to approximately 2 min.A comparative analysis with conventional methods showed strong agreement in track counts and absorbed dose estimates, with an average difference of 4.7 % in valid track numbers and 11.2 % in absorbed dose. The discrepancy in dose estimation was primarily attributed to differences in ellipse fitting methods used for LET calculations. Despite challenges in classifying small or low-contrast tracks, the proposed DNN-assisted method offers a promising solution for high-throughput, reliable, and standardized CR-39 track analysis in space radiation dosimetry. | |||||||||
| 書誌情報 |
Radiation Measurements 巻 190, p. 107565, 発行日 2026-01 |
|||||||||
| 出版者 | ||||||||||
| 出版者 | Elsevier | |||||||||
| ISSN | ||||||||||
| 収録物識別子タイプ | ISSN | |||||||||
| 収録物識別子 | 1879-0925 | |||||||||
| DOI | ||||||||||
| 識別子タイプ | DOI | |||||||||
| 関連識別子 | 10.1016/j.radmeas.2025.107565 | |||||||||
