Rapid estimation of patient‐specific organ doses using a deep learning network

Abstract

BackgroundPatient‐specific organ‐dose estimation in diagnostic CT examinations can provide useful insights on individualized secondary cancer risks, protocol optimization, and patient management. Current dose estimation techniques mainly rely on time‐consuming Monte Carlo methods or/and generalized anthropomorphic phantoms.PurposeWe proposed a proof‐of‐concept rapid workflow based on deep learning networks to estimate organ doses for individuals following thorax Computed Tomography (CT) examinations.MethodsCT scan data from 95 individuals undergoing thorax CT examinations were used. Monte Carlo simulations were performed and three‐dimensional (3D) dose distributions for each patient were obtained. A fully connected sequential deep learning network model was constructed and trained for each organ considered in this study. Water‐equivalent diameter (WED), scan length, and tube current were the independent variables. Organ doses for heart, lungs, esophagus, and bones were calculated from the Monte Carlo 3D distribution and used to train the deep learning networks. Organ dose predictions from each network were evaluated using an independent data set of 19 patients.ResultsThe trained networks provided organ dose predictions within a second. There was very good agreement between the deep learning network predictions and reference organ dose values calculated from Monte Carlo simulations. The average difference was −1.5% for heart, −1.6% for esophagus, −1.0% for lungs, and −0.4% for bones in the 95 patients dataset, and −5.1%, 4.3%, 0.9%, and 1.4% respectively in the 19 patients test dataset.ConclusionsThe proposed workflow demonstrated that patient‐specific organ‐doses can be estimated in nearly real‐time using deep learning networks. The workflow can be readily implemented and requires a small set of representative data for training.

Image detection of aortic dissection complications based on multi-scale feature fusion

“…While Gerard et al [ 5 ] proposed Reg3DNet+ residual regression convolutional neural network to directly regress high-resolution images of local tissue volume changes from CT images.Through image registration between total vital capacity and lung volume under functional residual capacity, using tissue mass and structure-preserving registration algorithms, it was possible to obtain a direct estimation of regional lung volume changes from both paired and single CT images.Furthermore, Marios [ 6 ] developed a fast proof-of-concept workflow based on deep learning networks for assessing individual organ doses after chest computed tomography(CT) examinations.The study leveraged an independent dataset consisting of 19 patients for each network for organ dose prediction.Huo et al [ 7 ] employed several machine learning algorithms to construct a classification and prediction model for early aortic dissection patients,to facilitate the identification of potential dissection cases among misdiagnosed patients.Among the models built,the Bayesian network achieved an 84.55% accuracy, with an Area Under Curve(AUC) of 85.7%.Singh et al [ 8 ],on the other hand, developed a deep learning model to automatically assess aortic dissection in chest CT images.They trained the model with 4235 images from 30 patients (15 dissection cases and 15 non-dissection cases) and tested it on 3423 images obtained from 40 patients (20 dissection cases and 20 non-dissection cases),using a pretrained inception-v3 [ 9 ] network as the classification model.The study demonstrated the efficacy of deep convolutional neural networks in the recognition of aortic dissection, achieving an AUC of 97% and a sensitivity of 100%.In their research, Lee et al [ 10 ] utilized a generative-discriminative learning approach to predict object boundaries in medical image datasets, applying it to true and false lumen segmentation in aortic dissection. Kovács et al [ 11 ] employed De-scoteaux's method to measure the intimal flap in the segmented CT image after segmenting the aortic image.…”

Fully automated explainable abdominal CT contrast media phase classification using organ segmentation and machine learning

“…For instance, the management of patients treated with selective internal radiation therapy (SIRT) for the treatment of liver malignancies requires CE-CT for the delineation of the tumoral tissues, perfused liver lobe and organs at risk towards personalized dosimetry. 3 Artificial intelligence (AI) and particularly deep learning has shown very promising performance in multiple tasks, including image segmentation, [4][5][6] image generation, 7,8 dosimetry, [9][10][11] and classification. 12,13 However, the number of clean and reliable data available is still the bottle neck for generalizability and robustness of deep learning models.…”