ABSTRACT
OBJECTIVE@#To propose a semi-supervised material quantitative intelligent imaging algorithm based on prior information perception learning (SLMD-Net) to improve the quality and precision of spectral CT imaging.@*METHODS@#The algorithm includes a supervised and a self- supervised submodule. In the supervised submodule, the mapping relationship between low and high signal-to-noise ratio (SNR) data was constructed through mean square error loss function learning based on a small labeled dataset. In the self- supervised sub-module, an image recovery model was utilized to construct the loss function incorporating the prior information from a large unlabeled low SNR basic material image dataset, and the total variation (TV) model was used to to characterize the prior information of the images. The two submodules were combined to form the SLMD-Net method, and pre-clinical simulation data were used to validate the feasibility and effectiveness of the algorithm.@*RESULTS@#Compared with the traditional model-driven quantitative imaging methods (FBP-DI, PWLS-PCG, and E3DTV), data-driven supervised-learning-based quantitative imaging methods (SUMD-Net and BFCNN), a material quantitative imaging method based on unsupervised learning (UNTV-Net) and semi-supervised learning-based cycle consistent generative adversarial network (Semi-CycleGAN), the proposed SLMD-Net method had better performance in both visual and quantitative assessments. For quantitative imaging of water and bone materials, the SLMD-Net method had the highest PSNR index (31.82 and 29.06), the highest FSIM index (0.95 and 0.90), and the lowest RMSE index (0.03 and 0.02), respectively) and achieved significantly higher image quality scores than the other 7 material decomposition methods (P < 0.05). The material quantitative imaging performance of SLMD-Net was close to that of the supervised network SUMD-Net trained with labeled data with a doubled size.@*CONCLUSIONS@#A small labeled dataset and a large unlabeled low SNR material image dataset can be fully used to suppress noise amplification and artifacts in basic material decomposition in spectral CT and reduce the dependence on labeled data-driven network, which considers more realistic scenario in clinics.
Subject(s)
Tomography, X-Ray Computed/methods , Image Processing, Computer-Assisted/methods , Algorithms , Signal-To-Noise Ratio , PerceptionABSTRACT
The main purpose of this study was to explore the latent relations of the selected morphometric, physiological and biochemical parameters. Thirty-six variables (12 morphometric, 9 physiological and 15 biochemical variables) were measured on 317 male-entities aged 17 - 35 y/o. The obtained data were analysed through the factor analysis of the first and second order. The statistical analyses were performed with the IBM SPSS Statistics software package, version 20. The factorization of the first order enabled extraction of 12 latent factors that explain 74.8 % of the total variance, while the factorization of the second order enabled extraction of five latent components that explain 51.39 % of the total variance. The final results of this study confirm the main hypothesis that there exist the numbers of latent variables that explain the latent structure of selected biometric measures. The nature of the extracted latent factors/ components in both orders of factorization is relatively clear, understandable, and easy to interpret. The higher projections of the manifest biometric variables on the extracted latent factors of the first and second order were accordingly with the nature of the measured variables. The results of this research might be considered as one step more in the holistic approach to the biometric measures.
El objetivo principal de este estudio fue explorar las relaciones latentes de parámetros morfométricos, fisiológicos y bioquímicos seleccionados. Treinta y seis variables (12 morfométricas, 9 fisiológicas y 15 bioquímicas) se midieron en 317 hombres de 17 a 35 años. Los datos obtenidos fueron analizados a través del análisis factorial de primer y segundo orden. Los análisis estadísticos se realizaron con el software IBM SPSS Statistics, versión 20. La factorización del primer orden permitió la extracción de 12 factores latentes que explican el 74,8 % de la varianza total, mientras que la factorización del segundo orden permitió la extracción de cinco componentes latentes que determinaron el 51,39 % de la varianza total. Los resultados finales de este estudio confirmaron la hipótesis principal de que existen números de variables latentes que explican la estructura latente de las medidas biométricas seleccionadas. La naturaleza de los factores/componentes latentes extraídos en ambos órdenes de factorización es relativamente clara, comprensible y fácil de interpretar. Las proyecciones superiores de las variables biométricas manifiestas en los factores latentes extraídos del primer y segundo orden correspondieron a la naturaleza de las variables medidas. Los resultados de esta investigación podrían considerarse como un paso más en el enfoque holístico de las medidas biométricas.