Paper Proffered Program Radiopharmaceuticals, Theranostics, and Nuclear Medicine

A Radionuclide-Independent Deep Learning–Based Patient-Specific Dosimetry

Abstract
Purpose

Conventional image-based dosimetry models for radiopharmaceutical therapy are generally radionuclide-specific and require extensive nuclear medicine (NM) datasets for training. We developed a deep learning (DL) model trained solely on a single whole-body CT image that predicts radiation doses for previously unseen radionuclides.

Methods

The DL model was trained to predict voxel dose kernels (VDKs) for beta-emitters using CT–derived density map patches and energy spectra. The DL model-derived VDKs were convolved with time-integrated activity to generate patient-specific dose maps. Zero-shot prediction was tested on unseen radionuclides (Ho-166 and Lu-177). The performance of DL model was validated at kernel, phantom, and patient levels. The proposed DL model, a conventional DL model trained on Y-90 NM/CT images of 22 patients, voxel S-value method (VSV), and organ-level dosimetry (OLINDA/EXM) were benchmarked against Monte Carlo (MC) simulation.

Results

The proposed DL model-based VDKs showed strong correlation with MC-based VDKs (R²: 0.99 and 0.91 for unseen Ho-166 and Lu-177) and, at the phantom level, revealed overall lower errors than VSV in heterogeneous region. At the patient level, the proposed DL model maintained good agreement with the MC simulation-derived horizontal dose profile for the unseen Lu-177 (mean difference: 2.1 Gy, 11.6%), whereas the conventional DL model trained using Y-90 data exhibited large discrepancies (mean difference: 129.4 Gy, 657.4%). The proposed DL model also predicted organ doses with the smallest differences from MC while reducing computation time from 4.6 hours to 2.6 minutes. For the mean tumor dose averaged across the patient cohort for unseen Ho-166, the DL model achieved the most accurate results with a difference of only 1.1 Gy (2.9%), outperforming multiple VSV (3.8 Gy, 7.2%), single VSV (3.8 Gy, 7.3%), and OLINDA/EXM (1.4 Gy, 4.5%).

Conclusion

The proposed DL model enables dose prediction for unseen radionuclides without requiring patient NM images for training.

People
Sangseok HaPresenting Author · Department of Medical Sciences, Graduate School of The Catholic University of Korea, Seoul, South Korea; Department of Biomedical Engineering, College of Medicine, The Catholic University of Korea, Seoul, South Korea Wonmo Sung, PhDCorrespondings · Department of Medical Sciences, Graduate School of The Catholic University of Korea, Seoul, South Korea; Department of Biomedical Engineering, College of Medicine, The Catholic University of Korea, Seoul, South Korea Jung Suk Oh, MD, PhDAuthors · Department of Radiology, Seoul St. Mary’s Hospital, The Catholic University of Korea Hyera KangAuthors · Department of Medical Sciences, Graduate School of The Catholic University of Korea, Seoul, South Korea; Department of Biomedical Engineering, College of Medicine, The Catholic University of Korea, Seoul, South Korea Seohan KimAuthors · Department of Medical Sciences, Graduate School of The Catholic University of Korea, Seoul, South Korea; Department of Biomedical Engineering, College of Medicine, The Catholic University of Korea, Seoul, South Korea Jinhee Jang, MD, PhDAuthors · Department of Radiology, Seoul St. Mary’s Hospital, The Catholic University of Korea Dong-san KangAuthors · Department of Medical Sciences, Graduate School of The Catholic University of Korea, Seoul, South Korea; Department of Biomedical Engineering, College of Medicine, The Catholic University of Korea, Seoul, South Korea

Related

Similar sessions

Poster Poster Program
Jul 19 · 07:00
Adverse Events in Targeted Radionuclide Therapy

Radiopharmaceutical therapy (RPT) plays an important role in the management of oncology patients, particularly those with thyroid cancer, prostate cancer, and neuroendocrine tumor. The use of radionuclide therapy has expanded rap...

Harrison L. Agordzo
Radiopharmaceuticals, Theranostics, and Nuclear Medicine 0 people interested
Poster Poster Program
Jul 19 · 07:00
Development of a Web-Based Theranostic Workflow Management Tool

To develop a Web-Based Theranostic Workflow Management Tool (TWMT) to efficiently manage Theranostic program in the department of radiation oncology (RadOnc).

Ling Zhuang, PhD
Radiopharmaceuticals, Theranostics, and Nuclear Medicine 0 people interested
Poster Poster Program
Jul 19 · 07:00
Epidseg-Net：the Multi-Modal Fusion Framework Based on Drr Guidance In Radiotherapy Is Used for Precise Segmentation of Epid Lung Targets

The proposed multimodal segmentation framework, named EPIDSeg-Net, comprises an encoder, a multi-scale feature layer, and a decoder. The encoder utilizes a dual-branch architecture: a CNN branch for extracting local texture featu...

Huang Qian Qianjia, M.Eng
Radiopharmaceuticals, Theranostics, and Nuclear Medicine 0 people interested