A semi-supervised material quantitative intelligent imaging algorithm for spectral CT based on prior information perception learning / 南方医科大学学报

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.

A nonlocal spectral similarity-induced material decomposition method for noise reduction of dual-energy CT images / 南方医科大学学报

OBJECTIVE@#To propose a nonlocal spectral similarity-induced material decomposition network (NSSD-Net) to reduce the correlation noise in the low-dose spectral CT decomposed images.@*METHODS@#We first built a model-driven iterative decomposition model for dual-energy CT, optimized the objective function solving process using the iterative shrinking threshold algorithm (ISTA), and cast the ISTA decomposition model into the deep learning network. We then developed a novel cost function based on the nonlocal spectral similarity to constrain the training process. To validate the decomposition performance, we established a material decomposition dataset by real patient dual-energy CT data. The NSSD-Net was compared with two traditional model-driven material decomposition methods, one data-based material decomposition method and one data-model coupling-driven material decomposition supervised learning method.@*RESULTS@#The quantitative results showed that compared with the two traditional methods, the NSSD-Net method obtained the highest PNSR values (31.383 and 31.444) and SSIM values (0.970 and 0.963) and the lowest RMSE values (2.901 and 1.633). Compared with the datamodel coupling-driven supervised decomposition method, the NSSD-Net method obtained the highest SSIM values on water and bone decomposed results. The results of subjective image quality assessment by clinical experts showed that the NSSD-Net achieved the highest image quality assessment scores on water and bone basis material (8.625 and 8.250), showing significant differences from the other 4 decomposition methods (P < 0.001).@*CONCLUSION@#The proposed method can achieve high-precision material decomposition and avoid training data quality issues and model unexplainable issues.

Texture analysis of iodine-based material decomposition images with spectral CT imaging for predicting microsatellite instability status in colorectal cancer / 中国医学影像技术

Objective: To investigate the value of texture analysis of iodine-based material decomposition images with spectral CT imaging for predicting microsatellite instability (MSI) status in colorectal cancer (CRC). Methods: Data of 23 patients with MSI status CRC and 46 patients with microsatellite stability (MSS) status CRC confirmed by postoperative pathology were retrospectively analyzed. All CRC patients underwent preoperative abdominal gemstone spectral imaging. Iodine-based material decomposition images in arterial and venous phases were produced with Viewer software, and the images were imported into Omni-Kinetics software for ROI sketching and feature extraction. The texture parameters included minimum intensity, maximum intensity, mean intensity, median intensity, standard deviation, kewness, kurtosis, uniformity, energy and entropy. The differences of parameters between the two groups were compared. Logistic regression was used to combine texture parameters. Diagnostic performances of various texture parameters and the combination of multiple parameters were studied with ROC analysis. Results: Both in arterial and venous phases, the minimum, maximum, mean, median, and uniformity in MSI group were significantly lower than those in MSS group (all P0.05). In venous phase, entropy in MSI group was significantly higher than that in MSS group (t=1.81, P=0.04). In arterial phase, there was no significant difference in entropy between the two groups (t=0.22, P=0.80). ROC analysis showed that the range of AUC for predicting MSI status in CRC patients using single texture parameter as minimum, maximum, mean, median, uniformity in arterial and venous phase or entropy in venous phase was 0.64~0.82. Multi-parameter combined diagnosis Logistic regression model was -2.598-0.124×arterial phase minimum-0.039×arterial phase maximum-0.774×arterial phase median+1×arterial phase mean-1.892×arterial phase uniformity+0.14×venous phase minimum+0.2×venous phase maximum+0.343×venous phase median-0.61×venous phase mean+13.711×venous phase uniformity-2.598×venous phase entropy. When combined multiple texture parameters, the AUC was 0.83. Conclusion: Texture analysis of iodine-based material decomposition image with spectral CT can serve as a preoperative non-invasive method for predicting MSI status in CRC patients. And the optimal predictive value was observed when combined all significant texture parameters.

Clinical assessment of spectrum CT with material decomposition technique in overcoming renal cyst pseudoenhancement / 中国医学影像技术

Objective To evaluate the value of material decomposition imaging of spectrum CT in overcoming renal cyst pseudoenhaneement.Methods Totally 80 patients with renal cysts (total 75 cysts) who underwent CT imaging with GSI mode were collected.The renal cysts were divided into 3 groups according to diameters,group A (diameters 0.5-＜1.5 cm,n=25),B (1.5-＜2.5 cm,n=25) and C (2.5-＜3.5 cm,n=25) respectively.The iodine-water density imaging was reconstructed by using the GSI Viewer analysis software.The CT value and iodine-water concentration of the cysts were recorded.The difference of CT value,iodine-water concentration in unenhanced and enhanced dual phases in each group was compared.Results The difference of CT value between plain scan and parenchyma phase among the 3 groups had statistically significant difference (F=204.128,P＜0.001),and the differences comparing any two were statistically significant (all P＜0.05).The postcontrast attenuation increased more than 10 HU in group A and B,indicating renal cyst pseudoenhancement,and less than 10 HU in group C,which had no pseudoenhancement.There were statistical difference in iodine concentration of the cysts of the 3 groups in unenhance,cortical and parenchyma phases (all P＜0.001),but the difference value in unenhance,cortical and parenchyma phases were less than 10 (100 μg/cm3),and the difference value of the 3 group was group A＞group B＞group C (all P＜0.05).The water concentration of the cysts in group A descend in renal cortical and parenchyma phase with statistical difference (P＜0.001),but the difference value was less than 10 mg/cms.Conclusion The measurements of iodine-water concentration appear to drift as well,the smaller the greater,The degree of the iodine concentration shifting is more obvious than water concentration.

Dual-Energy CT: New Horizon in Medical Imaging

Dual-energy CT has remained underutilized over the past decade probably due to a cumbersome workflow issue and current technical limitations. Clinical radiologists should be made aware of the potential clinical benefits of dual-energy CT over single-energy CT. To accomplish this aim, the basic principle, current acquisition methods with advantages and disadvantages, and various material-specific imaging methods as clinical applications of dual-energy CT should be addressed in detail. Current dual-energy CT acquisition methods include dual tubes with or without beam filtration, rapid voltage switching, dual-layer detector, split filter technique, and sequential scanning. Dual-energy material-specific imaging methods include virtual monoenergetic or monochromatic imaging, effective atomic number map, virtual non-contrast or unenhanced imaging, virtual non-calcium imaging, iodine map, inhaled xenon map, uric acid imaging, automatic bone removal, and lung vessels analysis. In this review, we focus on dual-energy CT imaging including related issues of radiation exposure to patients, scanning and post-processing options, and potential clinical benefits mainly to improve the understanding of clinical radiologists and thus, expand the clinical use of dual-energy CT; in addition, we briefly describe the current technical limitations of dual-energy CT and the current developments of photon-counting detector.

Spectral CT: Preliminary Studies in the Liver Cirrhosis

OBJECTIVE: To investigate the value of spectral CT imaging in the diagnosis and classification of liver cirrhosis during the arterial phase (AP) and portal venous phase (PVP). MATERIALS AND METHODS: Thirty-eight patients with liver cirrhosis (Child-Pugh class A/B/C: n = 10/14/14), and 43 patients with healthy livers, participated in this study. The researchers used abdominal spectral CT imaging during AP and PVP. Iodine concentration, derived from the iodine-based material-decomposition image and the iodine concentration ratio (ICratio) between AP and PVP, were obtained. Statistical analyses {two-sample t test, One-factor analysis of variance, and area under the receiver operating characteristic curve (A [z])} were performed. RESULTS: The mean normalized iodine concentration (NIC) (0.5 +/- 0.12) during PVP in the control group was significantly higher than that in the study group (0.4 +/- 0.10 on average, 0.4 +/- 0.08 for Class A, 0.4 +/- 0.15 for Class B, and 0.4 +/- 0.06 for Class C) (All p < 0.05). Within the cirrhotic liver group, the mean NIC for Class C during the AP (0.1 +/- 0.05) was significantly higher than NICs for Classes A (0.1 +/- 0.06) and B (0.1 +/- 0.03) (Both p < 0.05). The ICratio in the study group (0.4 +/- 0.15), especially for Class C (0.5 +/- 0.14), was higher than that in the control group (0.3 +/- 0.15) (p < 0.05).The combination of NIC and ICratio showed high sensitivity and specificity for differentiating healthy liver from cirrhotic liver, especially in Class C cirrhotic liver. CONCLUSION: Spectral CT Provides a quantitative method with which to analyze the cirrhotic liver, and shows the potential value in the classification of liver cirrhosis.

IMAGE-QUALITY OPTIMIZATION FOR DUAL ENERGY COMPUTED TOMOGRAPHY (DECT) THREE-MATERIAL DECOMPOSITION / OPTIMIZACIÓN DE LA CALIDAD DE LA IMAGEN PARA LA SEPARACIÓN DE TRES MATERIALES UTILIZANDO LA TOMOGRAFÍA COMPUTARIZADA DE DOS ENERGÍAS (TCDE)

Dual energy computed tomography, which consists of the acquisition of two images of a given region of interest using two different x-ray energies, has been used to decompose images. For example, it has been proposed to estimate the degree of stenosis of blood vessels with calcified plaques. Pragmatic realities, though, such as beam hardening, scattered radiation and mostly quantum noise, reduce the ideal of a perfect decomposition. In this study, a dual energy separation method for iodinated contrast media, cortical bone and soft tissue was implemented; afterwards, it was tested in simulated noiseless and noisy situations. The noise propagation was modeled mathematically, and an image-quality optimization technique regarding the right distribution of radiation dose between the high and low energy images was proposed. The results obtained suggest that in the absence of noise and using mono-energetic beams, an accurate separation is possible, but when noise is added and poly-chromatic spectra used this decomposition becomes more challenging.

La tomografía computarizada de dos energías, que consiste en la adquisición de dos imágenes de una región de interés dada usando rayos X de dos energías distintas, ha sido utilizada para descomponer imágenes. Por ejemplo, esta técnica ha sido propuesta para estimar el grado de estenosis de vasos sanguíneos obstruidos con placas de calcio. Sin embargo, fenómenos como el endurecimiento del rayo, la radiación dispersada y principalmente el ruido cuántico, impiden que esta separación sea perfecta. En este estudio, se implementó un método de separación de medio de contraste yodado, tejido blando y hueso cortical; este se evaluó en simulaciones, tanto en presencia como en ausencia de ruido. Posteriormente, se modeló matemáticamente la propagación del ruido y, con base en los resultados de estos modelos, se propuso una técnica de optimización de la imagen basada en la distribución adecuada de la dosis de radiación entre las imágenes de energías alta y baja. Los resultados evidencian que en ausencia de ruido y con rayos mono-energéticos, es posible obtener una separación precisa, pero cuando se adiciona ruido a las imágenes y se trabaja con espectros policromáticos, la descomposición resulta más complicada.

Estimation and Application of HU Values for Various Materials as Function of Physical Factor / 의학물리

This study aims to evaluate CT (Computed Tomography) characteristics through the estimation of HU (Hounsfield Unit) and the corresponding variations using coefficient of variation values for various materials as a function of physical factor. HU values for various materials with varying densities as a function of physical factor were measured using MDCT (Siemens SOMATOM Sensation 4, Germany). The results showed that the HU values were decreased and increased as a function of kVp and material density, respectively. Especially, the HU values for bone and iodine at 140 kVp were 32% and 42% smaller than those at 80 kVp, respectively. In case of iodine, the HU values also decreased and increased as a function of kVp and concentration, respectively. While the HU values were fixed as a function of mAs. The decreased ratio of HU values between 80 keV and 140 keV was different at various concentration and maximum difference was shown as 1.73 at 3% concentration. These results indicated that it may be possible to separate composition of materials, e.g. iodine and bone, using single source CT. The results showed that dual energy techniques using single source CT can be applied to material separation and expand CT imaging techniques to other practical applications.

Estimation and Application of HU Values for Various Materials as Function of Physical Factor / 의학물리

This study aims to evaluate CT (Computed Tomography) characteristics through the estimation of HU (Hounsfield Unit) and the corresponding variations using coefficient of variation values for various materials as a function of physical factor. HU values for various materials with varying densities as a function of physical factor were measured using MDCT (Siemens SOMATOM Sensation 4, Germany). The results showed that the HU values were decreased and increased as a function of kVp and material density, respectively. Especially, the HU values for bone and iodine at 140 kVp were 32% and 42% smaller than those at 80 kVp, respectively. In case of iodine, the HU values also decreased and increased as a function of kVp and concentration, respectively. While the HU values were fixed as a function of mAs. The decreased ratio of HU values between 80 keV and 140 keV was different at various concentration and maximum difference was shown as 1.73 at 3% concentration. These results indicated that it may be possible to separate composition of materials, e.g. iodine and bone, using single source CT. The results showed that dual energy techniques using single source CT can be applied to material separation and expand CT imaging techniques to other practical applications.