Swift prediction of personalized head and chest organ doses from CT examinations via neural networks with optimized quantity of hidden layers and radiomics features

Objective: To utilize radiomics features to enhance the prediction of personalized organ doses from CT scans, in order to explore methods for improving neural network-based models. Methods: Patient CT DICOM files were processed using DeepViewer to define regions of interest (ROIs) in their organs. R...

Full description

Saved in:
Bibliographic Details
Main Authors: Wencheng Shao, Xin Lin, Ying Huang, Liangyong Qu, Weihai Zhuo, Haikuan Liu
Format: Article
Language:English
Published: Elsevier 2025-04-01
Series:Radiation Medicine and Protection
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2666555725000218
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849698441562685440
author Wencheng Shao
Xin Lin
Ying Huang
Liangyong Qu
Weihai Zhuo
Haikuan Liu
author_facet Wencheng Shao
Xin Lin
Ying Huang
Liangyong Qu
Weihai Zhuo
Haikuan Liu
author_sort Wencheng Shao
collection DOAJ
description Objective: To utilize radiomics features to enhance the prediction of personalized organ doses from CT scans, in order to explore methods for improving neural network-based models. Methods: Patient CT DICOM files were processed using DeepViewer to define regions of interest (ROIs) in their organs. Radiomics features were extracted from the CT images and ROIs, and benchmark organ doses were calculated using Monte Carlo simulations. Fully-connected neural networks (FCNN) were trained with radiomics features to predict organ doses. The FCNN model was optimized by adjusting the number of input radiomics features and FCNN layers. Performance was evaluated using relative root mean squared error (RRMSE) and R-squared (R2). Results: Higher RRMSE and lower R2 values are observed when fewer than 30 input radiomics features are used for head CTs and fewer than 10 for chesst CTs. Increasing input features didn't significantly improve FCNN's performance. For head CTs, FCNN's layer quantities affected predictive stability, with better robustness observed with 4- and 5-layer FCNN. Specifically, the median RRMSE was reduced to 8.14% for the brain, 10.27% for the left eye, and 10.16% for the right eye when using 30 or more radiomics features. For chest CTs, the model's predictive stability was less sensitive to the number of layers, with median RRMSE values of 9.58% for the left lung and 9.44% for the right lung, and R² values of 0.76 for both lungs. Conclusions: Optimizing feature quantities and neural network layers enhances performance in predicting organ doses from CT scans. Specifically, head CTs show optimal results with 4–5 layers, while chest CTs do not significantly benefit from increased layers.
format Article
id doaj-art-2d4cb3997eb8405a8b9545d3da75e393
institution DOAJ
issn 2666-5557
language English
publishDate 2025-04-01
publisher Elsevier
record_format Article
series Radiation Medicine and Protection
spelling doaj-art-2d4cb3997eb8405a8b9545d3da75e3932025-08-20T03:18:54ZengElsevierRadiation Medicine and Protection2666-55572025-04-0162819010.1016/j.radmp.2025.02.002Swift prediction of personalized head and chest organ doses from CT examinations via neural networks with optimized quantity of hidden layers and radiomics featuresWencheng Shao0Xin Lin1Ying Huang2Liangyong Qu3Weihai Zhuo4Haikuan Liu5Institute of Radiation Medicine, Fudan University, Shanghai 200032, ChinaInstitute of Radiation Medicine, Fudan University, Shanghai 200032, ChinaDepartment of Nuclear Science and Technology, Institute of Modern Physics, Fudan University, Shanghai 200433, ChinaDepartment of Radiology, Shanghai Zhongye Hospital, Shanghai 201900, ChinaInstitute of Radiation Medicine, Fudan University, Shanghai 200032, China; Corresponding author.Institute of Radiation Medicine, Fudan University, Shanghai 200032, China; Corresponding author.Objective: To utilize radiomics features to enhance the prediction of personalized organ doses from CT scans, in order to explore methods for improving neural network-based models. Methods: Patient CT DICOM files were processed using DeepViewer to define regions of interest (ROIs) in their organs. Radiomics features were extracted from the CT images and ROIs, and benchmark organ doses were calculated using Monte Carlo simulations. Fully-connected neural networks (FCNN) were trained with radiomics features to predict organ doses. The FCNN model was optimized by adjusting the number of input radiomics features and FCNN layers. Performance was evaluated using relative root mean squared error (RRMSE) and R-squared (R2). Results: Higher RRMSE and lower R2 values are observed when fewer than 30 input radiomics features are used for head CTs and fewer than 10 for chesst CTs. Increasing input features didn't significantly improve FCNN's performance. For head CTs, FCNN's layer quantities affected predictive stability, with better robustness observed with 4- and 5-layer FCNN. Specifically, the median RRMSE was reduced to 8.14% for the brain, 10.27% for the left eye, and 10.16% for the right eye when using 30 or more radiomics features. For chest CTs, the model's predictive stability was less sensitive to the number of layers, with median RRMSE values of 9.58% for the left lung and 9.44% for the right lung, and R² values of 0.76 for both lungs. Conclusions: Optimizing feature quantities and neural network layers enhances performance in predicting organ doses from CT scans. Specifically, head CTs show optimal results with 4–5 layers, while chest CTs do not significantly benefit from increased layers.http://www.sciencedirect.com/science/article/pii/S2666555725000218CTPersonalized organ doseRadiomics featuresNeural networkOptimization
spellingShingle Wencheng Shao
Xin Lin
Ying Huang
Liangyong Qu
Weihai Zhuo
Haikuan Liu
Swift prediction of personalized head and chest organ doses from CT examinations via neural networks with optimized quantity of hidden layers and radiomics features
Radiation Medicine and Protection
CT
Personalized organ dose
Radiomics features
Neural network
Optimization
title Swift prediction of personalized head and chest organ doses from CT examinations via neural networks with optimized quantity of hidden layers and radiomics features
title_full Swift prediction of personalized head and chest organ doses from CT examinations via neural networks with optimized quantity of hidden layers and radiomics features
title_fullStr Swift prediction of personalized head and chest organ doses from CT examinations via neural networks with optimized quantity of hidden layers and radiomics features
title_full_unstemmed Swift prediction of personalized head and chest organ doses from CT examinations via neural networks with optimized quantity of hidden layers and radiomics features
title_short Swift prediction of personalized head and chest organ doses from CT examinations via neural networks with optimized quantity of hidden layers and radiomics features
title_sort swift prediction of personalized head and chest organ doses from ct examinations via neural networks with optimized quantity of hidden layers and radiomics features
topic CT
Personalized organ dose
Radiomics features
Neural network
Optimization
url http://www.sciencedirect.com/science/article/pii/S2666555725000218
work_keys_str_mv AT wenchengshao swiftpredictionofpersonalizedheadandchestorgandosesfromctexaminationsvianeuralnetworkswithoptimizedquantityofhiddenlayersandradiomicsfeatures
AT xinlin swiftpredictionofpersonalizedheadandchestorgandosesfromctexaminationsvianeuralnetworkswithoptimizedquantityofhiddenlayersandradiomicsfeatures
AT yinghuang swiftpredictionofpersonalizedheadandchestorgandosesfromctexaminationsvianeuralnetworkswithoptimizedquantityofhiddenlayersandradiomicsfeatures
AT liangyongqu swiftpredictionofpersonalizedheadandchestorgandosesfromctexaminationsvianeuralnetworkswithoptimizedquantityofhiddenlayersandradiomicsfeatures
AT weihaizhuo swiftpredictionofpersonalizedheadandchestorgandosesfromctexaminationsvianeuralnetworkswithoptimizedquantityofhiddenlayersandradiomicsfeatures
AT haikuanliu swiftpredictionofpersonalizedheadandchestorgandosesfromctexaminationsvianeuralnetworkswithoptimizedquantityofhiddenlayersandradiomicsfeatures