ABSTRACT
Glioma is the most common type of primary malignant tumor of the central nervous system (CNS), and is characterized by high malignancy, high recurrence rate and poor survival. Conventional imaging techniques only provide information regarding the anatomical location, morphological characteristics, and enhancement patterns. In contrast, advanced imaging techniques such as dynamic contrast-enhanced (DCE) MRI or DCE CT can reflect tissue microcirculation, including tumor vascular hyperplasia and vessel permeability. Although several studies have used DCE imaging to evaluate gliomas, the results of data analysis using conventional tracer kinetic models (TKMs) such as Tofts or extended-Tofts model (ETM) have been ambiguous. More advanced models such as Brix's conventional two-compartment model (Brix), tissue homogeneity model (TH) and distributed parameter (DP) model have been developed, but their application in clinical trials has been limited. This review attempts to appraise issues on glioma studies using conventional TKMs, such as Tofts or ETM model, highlight advancement of DCE imaging techniques and provides insights on the clinical value of glioma management using more advanced TKMs.
ABSTRACT
BACKGROUND: Parameters from diffusion-weighted imaging (DWI) have been increasingly used as imaging biomarkers for the diagnosis and monitoring of treatment responses in cancer. The consistency of DWI measurements across different centers remains uncertain, which limits the widespread use of quantitative DWI in clinical settings. PURPOSE: To investigate the consistency of quantitative metrics derived from DWI between different scanners in a multicenter clinical setting. MATERIAL AND METHODS: A total of 193 patients with cervical cancer from four scanners (MRI1, MRI2, MRI3, and MRI4) at three centers were included in this retrospective study. DWI data were processed using the mono-exponential and intravoxel incoherent motion (IVIM) model, yielding the following parameters: apparent diffusion coefficient (ADC); true diffusion coefficient (D); pseudo-diffusion coefficient (D*); perfusion fraction (f); and the product of f and D* (fD*). Various parameters of cervical cancer obtained from different scanners were compared. RESULTS: The parameters D and ADC derived from MRI1 and MRI2 were significantly different from those derived from MRI3 or MRI4 (P <0.01 for all comparisons). However, there was no significant difference in cervical cancer perfusion parameters (D* and fD*) between the different scanners (P >0.05). The P values of comparisons of all DWI parameters (D, D*, fD*, and ADC) between MRI3 and MRI4 (same vendor in different centers) for cervical cancer were all >0.05, except for f (P = 0.05). CONCLUSION: Scanners of the same model by the same vendor can yield close measurements of the ADC and IVIM parameters. The perfusion parameters showed higher consistency among the different scanners.
Subject(s)
Diffusion Magnetic Resonance Imaging , Uterine Cervical Neoplasms , Humans , Uterine Cervical Neoplasms/diagnostic imaging , Female , Diffusion Magnetic Resonance Imaging/methods , Middle Aged , Retrospective Studies , Adult , Aged , Aged, 80 and over , Reproducibility of Results , Image Interpretation, Computer-Assisted/methodsABSTRACT
BACKGROUND: Patients with early endometrial carcinoma (EC) have a good prognosis, but it is difficult to distinguish from endometrial polyps (EPs). PURPOSE: To develop and assess magnetic resonance imaging (MRI)-based radiomics models for discriminating Stage I EC from EP in a multicenter setting. MATERIAL AND METHODS: Patients with Stage I EC (n = 202) and EP (n = 99) who underwent preoperative MRI scans were collected in three centers (seven devices). The images from devices 1-3 were utilized for training and validation, and the images from devices 4-7 were utilized for testing, leading to three models. They were evaluated by the area under the receiver operating characteristic curve (AUC) and metrics including accuracy, sensitivity, and specificity. Two radiologists evaluated the endometrial lesions and compared them with the three models. RESULTS: The AUCs of device 1, 2_ada, device 1, 3_ada, and device 2, 3_ada for discriminating Stage I EC from EP were 0.951, 0.912, and 0.896 for the training set, 0.755, 0.928, and 1.000 for the validation set, and 0.883, 0.956, and 0.878 for the external validation set, respectively. The specificity of the three models was higher, but the accuracy and sensitivity were lower than those of radiologists. CONCLUSION: Our MRI-based models showed good potential in differentiating Stage I EC from EP and had been validated in multiple centers. Their specificity was higher than that of radiologists and may be used for computer-aided diagnosis in the future to assist clinical diagnosis.
Subject(s)
Endometrial Neoplasms , Uterine Neoplasms , Female , Humans , Magnetic Resonance Imaging/methods , Endometrial Neoplasms/diagnostic imaging , Endometrial Neoplasms/pathology , ROC Curve , Retrospective StudiesABSTRACT
Introduction: The successful use of machine learning (ML) for medical diagnostic purposes has prompted myriad applications in cancer image analysis. Particularly for hepatocellular carcinoma (HCC) grading, there has been a surge of interest in ML-based selection of the discriminative features from high-dimensional magnetic resonance imaging (MRI) radiomics data. As one of the most commonly used ML-based selection methods, the least absolute shrinkage and selection operator (LASSO) has high discriminative power of the essential feature based on linear representation between input features and output labels. However, most LASSO methods directly explore the original training data rather than effectively exploiting the most informative features of radiomics data for HCC grading. To overcome this limitation, this study marks the first attempt to propose a feature selection method based on LASSO with dictionary learning, where a dictionary is learned from the training features, using the Fisher ratio to maximize the discriminative information in the feature. Methods: This study proposes a LASSO method with dictionary learning to ensure the accuracy and discrimination of feature selection. Specifically, based on the Fisher ratio score, each radiomic feature is classified into two groups: the high-information and the low-information group. Then, a dictionary is learned through an optimal mapping matrix to enhance the high-information part and suppress the low discriminative information for the task of HCC grading. Finally, we select the most discrimination features according to the LASSO coefficients based on the learned dictionary. Results and discussion: The experimental results based on two classifiers (KNN and SVM) showed that the proposed method yielded accuracy gains, compared favorably with another 5 state-of-the-practice feature selection methods.
ABSTRACT
PURPOSE: To compare a deep learning model with a radiomics model in differentiating high-grade (LR-3, LR-4, LR-5) liver imaging reporting and data system (LI-RADS) liver tumors from low-grade (LR-1, LR-2) LI-RADS tumors based on the contrast-enhanced magnetic resonance images. METHODS: Magnetic resonance imaging scans of 361 suspected hepatocellular carcinoma patients were retrospectively reviewed. Lesion volume segmentation was manually performed by two radiologists, resulting in 426 lesions from the training set and 83 lesions from the test set. The radiomics model was constructed using a support vector machine (SVM) with pre-defined features, which was first selected using Chi-square test, followed by refining using binary least absolute shrinkage and selection operator (LASSO) regression. The deep learning model was established based on the DenseNet. Performance of the models was quantified by area under the receiver-operating characteristic curve (AUC), accuracy, sensitivity, specificity and F1-score. RESULTS: A set of 8 most informative features was selected from 1049 features to train the SVM classifier. The AUCs of the radiomics model were 0.857 (95% confidence interval [CI] 0.816-0.888) for the training set and 0.879 (95% CI 0.779-0.935) for the test set. The deep learning method achieved AUCs of 0.838 (95% CI 0.799-0.871) for the training set and 0.717 (95% CI 0.601-0.814) for the test set. The performance difference between these two models was assessed by t-test, which showed the results in both training and test sets were statistically significant. CONCLUSION: The deep learning based model can be trained end-to-end with little extra domain knowledge, while the radiomics model requires complex feature selection. However, this process makes the radiomics model achieve better performance in this study with smaller computational cost and more potential on model interpretability.
Subject(s)
Carcinoma, Hepatocellular , Deep Learning , Liver Neoplasms , Humans , Liver Neoplasms/diagnostic imaging , Carcinoma, Hepatocellular/diagnostic imaging , Retrospective Studies , Magnetic Resonance ImagingABSTRACT
Objectives: This study assessed the clinical value of parameters derived from dynamic contrast-enhanced (DCE) MRI with respect to correlation with angiogenesis and proliferation of cervical cancer, performance of diagnosis and reproducibility of DCE-MRI parameters across MRI scanners. Materials and Methods: A total of 113 patients with cervical carcinoma from two centers were included in this retrospective study. The DCE data were centralized and processed using five tracer kinetic models (TKMs) (Tofts, Ex-Tofts, ATH, SC, and DP), yielding the following parameters: volume transfer constant (Ktrans), extravascular extracellular volume (Ve), fractional volume of vascular space (Vp), blood flow (Fp), and permeability surface area product (PS). CD34 counts and Ki-67 PI (proliferation index) of cervical cancer and normal cervix tissue were obtained using immunohistochemical staining in Center 1. Results: CD34 count and Ki-67 PI in cervical cancer were significantly higher than in normal cervix tissue (p<0.05). Parameter Ve from each TKM was significantly smaller in cervical cancer tissue than in normal cervix tissue (p<0.05), indicating the higher proliferation of cervical cancer cells. Ve of each TKM attained the largest AUC to diagnose cervical cancer. The distributions of DCE parameters for both cervical cancer and normal cervix tissue were not significantly different between two centers (P>0.05). Conclusion: Parameter Ve was similar to the expression of Ki-67 in revealing the proliferation of tissue cells, attained good performance in diagnosis of cervical cancer, and demonstrated consistent findings on measured values across centers.
ABSTRACT
Purpose: The purpose of this study was to design an automated algorithm that can detect fluorescence leakage accurately and quickly without the use of a large amount of labeled data. Methods: A weakly supervised learning-based method was proposed to detect fluorescein leakage without the need for manual annotation of leakage areas. To enhance the representation of the network, a residual attention module (RAM) was designed as the core component of the proposed generator. Moreover, class activation maps (CAMs) were used to define a novel anomaly mask loss to facilitate more accurate learning of leakage areas. In addition, sensitivity, specificity, accuracy, area under the curve (AUC), and dice coefficient (DC) were used to evaluate the performance of the methods. Results: The proposed method reached a sensitivity of 0.73 ± 0.04, a specificity of 0.97 ± 0.03, an accuracy of 0.95 ± 0.05, an AUC of 0.86 ± 0.04, and a DC of 0.87 ± 0.01 on the HRA data set; a sensitivity of 0.91 ± 0.02, a specificity of 0.97 ± 0.02, an accuracy of 0.96 ± 0.03, an AUC of 0.94 ± 0.02, and a DC of 0.85 ± 0.03 on Zhao's publicly available data set; and a sensitivity of 0.71 ± 0.04, a specificity of 0.99 ± 0.06, an accuracy of 0.87 ± 0.06, an AUC of 0.85 ± 0.02, and a DC of 0.78 ± 0.04 on Rabbani's publicly available data set. Conclusions: The experimental results showed that the proposed method achieves better performance on fluorescence leakage detection and can detect one image within 1 second and thus has great potential value for clinical diagnosis and treatment of retina-related diseases, such as diabetic retinopathy and malarial retinopathy. Translational Relevance: The proposed weakly supervised learning-based method that automates the detection of fluorescence leakage can facilitate the assessment of retinal-related diseases.
Subject(s)
Deep Learning , Diabetic Retinopathy , Retinal Diseases , Algorithms , Area Under Curve , Diabetic Retinopathy/diagnostic imaging , Fluorescein Angiography , HumansABSTRACT
BACKGROUND: Most commonly used diffusion-weighted imaging (DWI) models include intravoxel incoherent motion (IVIM), diffusion kurtosis imaging (DKI), stretched exponential model (SEM), and mono-exponential model (MEM). Previous studies of the four models were inconsistent on which model was more effective in distinguishing cervical cancer from normal cervical tissue. PURPOSE: To assess the performance of four DWI models in characterizing cervical cancer and normal cervical tissue. MATERIAL AND METHODS: Forty-seven women with suspected cervical carcinoma underwent DWI using eight b-values before treatment. Imaging parameters, calculated using IVIM, SEM, DKI, and MEM, were compared between cervical cancer and normal cervical tissue. The diagnostic performance of the models was evaluated using independent t-test, Mann-Whitney U test, receiver operating characteristic (ROC) curve analysis, and multivariate logistic regression analysis. RESULTS: All parameters except pseudo-diffusion coefficient (D*) differed significantly between cervical cancer and normal cervical tissue (P < 0.001). Through logistic regression analysis, all combined models showed a significant improvement in area under the ROC curve (AUC) compared to individual DWI parameters. The model with combined IVIM parameters had a larger AUC value compared to those of other combined models (P < 0.05). CONCLUSION: All four DWI models are useful for differentiating cervical cancer from normal cervical tissue and IVIM may be the optimal model.
Subject(s)
Diffusion Magnetic Resonance Imaging/methods , Uterine Cervical Neoplasms/diagnostic imaging , Cervix Uteri/diagnostic imaging , Diagnosis, Differential , Female , Humans , Middle Aged , Reproducibility of Results , Sensitivity and SpecificityABSTRACT
PURPOSE: To develop and validate an MRI-based radiomics model for preoperatively distinguishing endometrial carcinoma (EC) with benign mimics in a multicenter setting. METHODS: Preoperative MRI scans of EC patients were retrospectively reviewed and divided into training set (158 patients from device 1 in center A), test set #1 (78 patients from device 2 in center A) and test set #2 (109 patients from device 3 in center B). Two radiologists performed manual delineation of tumor segmentation on the map of apparent diffusion coefficient (ADC), diffusion-weighted imaging (DWI) and T2-weighted imaging (T2WI). The features were extracted and firstly selected using Chi-square test, followed by refining using binary least absolute shrinkage and selection operator (LASSO) regression. The support vector machine (SVM) was employed to build the radiomics model, which is tuned in the training set using 10-fold cross-validation, and then assessed on the test set. Performance of the model was determined by area under the receiver-operating characteristic curve (AUC), accuracy, sensitivity, specificity and F1-score. RESULTS: Five most informative features are selected from the extracted 3142 features. The SVM with linear kernel was employed to build the radiomics model. The AUCs of the model were 0.989 (95% confidence interval [CI]: 0.968-0.997) for the training set, 0.999 (95% CI: 0.991-1.000) for test set #1, 0.961 (95% CI: 0.902-0.983) for test set #2. Accuracies of the model were 0.937 for the training set (precision: 0.919, recall: 0.963, F1-score: 0.940), 0.974 for test set #1 (precision: 0.949, recall: 1.000, F1-score: 0.974) and 0.908 for test set #2 (precision: 0.899, recall: 0.954, F1-score: 0.925). These results confirmed the efficacy of this model in predicting EC in different centers. CONCLUSION: The MRI-based radiomics model showed promising potential for distinguishing EC with benign mimics and might be useful for surgical management of EC.
Subject(s)
Endometrial Neoplasms , Magnetic Resonance Imaging , Diffusion Magnetic Resonance Imaging , Endometrial Neoplasms/diagnostic imaging , Female , Humans , ROC Curve , Retrospective StudiesABSTRACT
BACKGROUND: To compare different fitting methods for determining IVIM (Intravoxel Incoherent Motion) parameters and to determine whether the use of different IVIM fitting methods would affect differentiation of cervix cancer from normal cervix tissue. METHODS: Diffusion-weighted echo-planar imaging of 30 subjects was performed on a 3.0 T scanner with b-values of 0, 30, 100, 200, 400, 1000 s/mm2. IVIM parameters were estimated using the segmented (two-step) fitting method and by simultaneous fitting of a bi-exponential function. Segmented fitting was performed using two different cut-off b-values (100 and 200 s/mm2) to study possible variations due to the choice of cut-off. Friedman's test and Student's t-test were respectively used to compare IVIM parameters derived from different methods, and between cancer and normal tissues. RESULTS: No significant difference was found between IVIM parameters derived from the segmented method with b-value cutoff of 200 s/mm2 and the simultaneous fitting method (P>0.05). Tissue diffusivity (D) and perfusion fraction (f) were significantly lower in cervix cancer than normal tissue (P< 0.05). CONCLUSIONS: IVIM parameters derived using fitting methods with small cutoff b-values could be different, however, the segmented method with b-value cutoff of 200 s/mm2 are consistent with the simultaneous fitting method and both can be used to differentiate between cervix cancer and normal tissue.
Subject(s)
Diffusion Magnetic Resonance Imaging/methods , Echo-Planar Imaging/methods , Uterine Cervical Neoplasms/diagnostic imaging , Female , Humans , Middle Aged , Motion , Uterine Cervical Neoplasms/pathologyABSTRACT
Previous studies using contrast-enhanced imaging for glioma isocitrate dehydrogenase (IDH) mutation assessment showed promising yet inconsistent results, and this study attempts to explore this problem by using an advanced tracer kinetic model, the distributed parameter model (DP). Fifty-five patients with glioma examined using dynamic contrast-enhanced imaging sequence at a 3.0 T scanner were retrospectively reviewed. The imaging data were processed using DP, yielding the following parameters: blood flow F, permeability-surface area product PS, fractional volume of interstitial space Ve, fractional volume of intravascular space Vp, and extraction ratio E. The results were compared with the Tofts model. The Wilcoxon test and boxplot were utilized for assessment of differences of model parameters between IDH-mutant and IDH-wildtype gliomas. Spearman correlation r was employed to investigate the relationship between DP and Tofts parameters. Diagnostic performance was evaluated using receiver operating characteristic (ROC) curve analysis and quantified using the area under the ROC curve (AUC). Results showed that IDH-mutant gliomas were significantly lower in F (P = 0.018), PS (P < 0.001), Vp (P < 0.001), E (P < 0.001), and Ve (P = 0.002) than IDH-wildtype gliomas. In differentiating IDH-mutant and IDH-wildtype gliomas, Vp had the best performance (AUC = 0.92), and the AUCs of PS and E were 0.82 and 0.80, respectively. In comparison, Tofts parameters were lower in K trans (P = 0.013) and Ve (P < 0.001) for IDH-mutant gliomas. No significant difference was observed in Kep (P = 0.525). The AUCs of K trans, Ve, and Kep were 0.69, 0.79, and 0.55, respectively. Tofts-derived Ve showed a strong correlation with DP-derived Ve (r > 0.9, P < 0.001). K trans showed a weak correlation with F (r < 0.3, P > 0.16) and a very weak correlation with PS (r < 0.06, P > 0.8), both of which were not statistically significant. The findings by DP revealed a tissue environment with lower vascularity, lower vessel permeability, and lower blood flow in IDH-mutant than in IDH-wildtype gliomas, being hostile to cellular differentiation of oncogenic effects in IDH-mutated gliomas, which might help to explain the better outcomes in IDH-mutated glioma patients than in glioma patients of IDH-wildtype. The advantage of DP over Tofts in glioma DCE data analysis was demonstrated in terms of clearer elucidation of tissue microenvironment and better performance in IDH mutation assessment.
Subject(s)
Brain Neoplasms/pathology , Contrast Media/analysis , Glioma/pathology , Isocitrate Dehydrogenase/genetics , Mutation , Adult , Aged , Brain Neoplasms/genetics , Female , Follow-Up Studies , Glioma/genetics , Humans , Magnetic Resonance Imaging , Male , Middle Aged , Prognosis , ROC Curve , Retrospective StudiesABSTRACT
Physiological parameters can be estimated from dynamic contrast enhanced magnetic resonance imaging (DCEMRI) data using pharmacokinetic models. This work evaluates the performance of various pharmacokinetic models through a retrospective study on cervix cancer, including two generalized kinetic models and three 2-compartment exchange models (2CXMs). In the current clinical practice, region of interest (ROI) is treated as a whole and the models are assessed by their top pharmacokinetic parameters. We explore the intervoxel relationship in the pharmacokinetic parameter maps and demonstrate that, for those insignificant parameters, texture descriptors can largely improve their discriminative power. Multi-parametric classifiers are developed to fuse the information carried by physiological parameters and the descriptors. Assessed merely by the top parameter, the DP (distributed parameter) model is the best one with an area under the ROC (receiver operating characteristic) curve (AUC) of 0.80; by combining multiple pharmacokinetic parameters, the ExTofts model is the winner with an AUC of 0.837. Finally, the classifier of the AATH (adiabatic approximation to the tissue homogeneity) model build on combined features achieves an AUC of 0.92.Clinical Relevance - Using data from 36 cervical cancer patients and 17 normal subjects, this work quantitatively compared the various pharmacokinetic models and provided recommendations for model selection in cervical cancer diagnosis.
Subject(s)
Uterine Cervical Neoplasms , Contrast Media , Female , Humans , Magnetic Resonance Imaging , ROC Curve , Retrospective StudiesABSTRACT
Pharmacokinetic parameters estimated from dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) time course data enable the physio-biological interpretation of tissue angiogenesis. This study aims to develop machine learning approaches for cervical carcinoma prediction based on pharmacokinetic parameters. The performance of individual parameters was assessed in terms of their efficacy in differentiating cancerous tissue from normal cervix tissue. The effect of combining parameters was evaluated using the following two approaches: the first approach was based on support vector machines (SVMs) to combine the parameters from one pharmacokinetic model or across several models; the second approach was based on a novel method called APITL (artificial pharmacokinetic images for transfer learning), which was designed to fully utilize the comprehensive pharmacokinetic information acquired from DCE-MRI data. A "winner-takes-all" strategy was employed to consolidate the slice-wise prediction into subject-wise prediction. Experiments were carried out with a dataset comprising 36 patients with cervical cancer and 17 healthy subjects. The results demonstrated that parameter Ve, representing volume fraction of the extracellular extravascular space (EES), attained high discriminative power regardless of the pharmacokinetic model used for estimation. An approximately 10% improvement in the accuracy was achieved with the SVM approach. The APITL method further outperformed SVM and attained a subject-wise prediction accuracy of 94.3%. Our experiment demonstrated that APITL could predict cervical carcinoma with high accuracy and had potential in clinical applications.
Subject(s)
Uterine Cervical Neoplasms , Contrast Media , Female , Humans , Machine Learning , Magnetic Resonance Imaging , Support Vector Machine , Uterine Cervical Neoplasms/diagnostic imagingABSTRACT
PURPOSE: To evaluate whether the analysis of high-temporal resolution DCE-MRI by various tracer kinetic models could yield useful radiomic features in discriminating cervix carcinoma and normal cervix tissue. METHODS: Forty-three patients (median age 51 yr; range 26-78 yr) diagnosed with cervical cancer based on postoperative pathology were enrolled in this study with informed consent. DCE-MRI data with temporal resolution of 2 s were acquired and analyzed using the Tofts (TOFTS), extended Tofts (EXTOFTS), conventional two-compartment (CC), adiabatic tissue homogeneity (ATH), and distributed parameter (DP) models. Ability of all kinetic parameters in distinguishing tumor from normal tissue was assessed using Mann-Whitney U test and receiver operating characteristic (ROC) curves. Repeatability of parameter estimates due to sampling of arterial input functions (AIFs) was also studied using intraclass correlation (ICC) analysis. RESULTS: Fractional extravascular, extracellular volume (Ve) of all models were significantly smaller in cervix carcinoma than normal cervix tissue, and were associated with large values of area under ROC curve (AUC 0.884-0.961). Capillary permeability PS derived from the ATH, CC, and DP models also yielded large AUC values (0.730, 0.860, and 0.797). Transfer constant Ktrans derived from TOFTS and EXTOFTS models yielded smaller AUC (0.587 and 0.701). Repeatability of parameters derived from all models was robust to AIF sampling, with ICC coefficients typically larger than 0.80. CONCLUSIONS: With the use of high-temporal resolution DCE-MRI, all tracer kinetic models could reflect pathophysiological differences between cervix carcinoma and normal tissue (with significant differences in Ve and PS) and potentially yield radiomic features with diagnostic value.
Subject(s)
Contrast Media , Image Processing, Computer-Assisted/methods , Magnetic Resonance Imaging , Uterine Cervical Neoplasms/diagnostic imaging , Adult , Aged , Female , Humans , Middle AgedABSTRACT
A variety of tracer kinetic methods have been employed to assess tumor angiogenesis. The Standard two-Compartment model (SC) used in cervix carcinoma was less frequent, and Adiabatic Approximation to the Tissue Homogeneity (AATH) and Distributed Parameter (DP) model are lacking. This study compares two-compartment exchange models (2CXM) (AATH, SC, and DP) for determining dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) parameters in cervical cancer, with the aim of investigating the potential of various parameters derived from 2CXM for tumor diagnosis and exploring the possible relationship between these parameters in patients with cervix cancer. Parameters (tissue blood flow, F p; tissue blood volume, V p; interstitial volume, V e; and vascular permeability, PS) for regions of interest (ROI) of cervix lesions and normal cervix tissue were estimated by AATH, SC, and DP models in 36 patients with cervix cancer and 17 healthy subjects. All parameters showed significant differences between lesions and normal tissue with a P value less than 0.05, except for PS from the AATH model, F p from the SC model, and V p from the DP model. Parameter V e from the AATH model had the largest AUC (r = 0.85). Parameters F p and V p from SC and DP models and V e and PS from AATH and DP models were highly correlated, respectively, (r > 0.8) in cervix lesions. Cervix cancer was found to have a very unusual microcirculation pattern, with over-growth of cancer cells but without evident development of angiogenesis. V e has the best performance in identifying cervix cancer. Most physiological parameters derived from AATH, SC, and DP models are linearly correlated in cervix cancer.
Subject(s)
Contrast Media/administration & dosage , Magnetic Resonance Imaging , Neovascularization, Pathologic/diagnostic imaging , Uterine Cervical Neoplasms/diagnostic imaging , Adult , Aged , Female , Humans , Middle Aged , Neoplasm Staging , Neovascularization, Pathologic/pathology , Uterine Cervical Neoplasms/pathologyABSTRACT
Dynamic contrast-enhanced (DCE) images can be acquired at multiple time points and multiple slice locations of a tumor. Image segmentation and registration are important preprocessing steps that can improve subsequent analysis of DCE images by kinetic modeling. An automatic system for region-of-interest segmentation and registration of DCE images is presented. Tissue segmentation is performed using a combination of thresholding and morphological operations, and further refined using shape information from consecutive images. The segmented regions are subsequently registered based on a mutual information method that accounts for possible tissue movement between slices. The proposed segmentation and registration methods are applied on actual DCE CT datasets to illustrate feasibility of practical implementation in the clinic.
Subject(s)
Colorectal Neoplasms/diagnosis , Image Enhancement/methods , Magnetic Resonance Imaging/methods , Tomography, Emission-Computed/methods , HumansABSTRACT
A novel thresholding approach to confine the intensity frequency range of the object based on supervision is introduced. It consists of three steps. First, the region of interest (ROI) is determined in the image. Then, from the histogram of the ROI, the frequency range in which the proportion of the background to the ROI varies is estimated through supervision. Finally, the threshold is determined by minimizing the classification error within the constrained variable background range. The performance of the approach has been validated against 54 brain MR images, including images with severe intensity inhomogeneity and/or noise, CT chest images, and the Cameraman image. Compared with nonsupervised thresholding methods, the proposed approach is substantially more robust and more reliable. It is also computationally efficient and can be applied to a wide class of computer vision problems, such as character recognition, fingerprint identification, and segmentation of a wide variety of medical images.
Subject(s)
Algorithms , Artificial Intelligence , Image Enhancement/methods , Image Interpretation, Computer-Assisted/methods , Pattern Recognition, Automated/methods , Imaging, Three-Dimensional/methods , Information Storage and Retrieval/methods , Reproducibility of Results , Sensitivity and SpecificityABSTRACT
This paper presents a method to improve the semi-automatic method for intensity inhomogeneity correction by Dawant et al. through introducing a fully automatic approach to reference points generation, which is based on order statistics and integrates information from the fine to coarse scale representations of the input image. The method has been validated and compared with two popular methods, N3 and BFC. Advantages of the proposed method are demonstrated.
Subject(s)
Artificial Intelligence , Brain/anatomy & histology , Image Enhancement/methods , Image Interpretation, Computer-Assisted/methods , Imaging, Three-Dimensional/methods , Magnetic Resonance Imaging/methods , Pattern Recognition, Automated/methods , Algorithms , Artifacts , Humans , Reproducibility of Results , Sensitivity and Specificity , Subtraction TechniqueABSTRACT
Intensity inhomogeneity (IIH) is often encountered in MR imaging, and a number of techniques have been devised to correct this artifact. This paper attempts to review some of the recent developments in the mathematical modeling of IIH field. Low-frequency models are widely used, but they tend to corrupt the low-frequency components of the tissue. Hypersurface models and statistical models can be adaptive to the image and generally more stable, but they are also generally more complex and consume more computer memory and CPU time. They are often formulated together with image segmentation within one framework and the overall performance is highly dependent on the segmentation process. Beside these three popular models, this paper also summarizes other techniques based on different principles. In addition, the issue of quantitative evaluation and comparative study are discussed.
ABSTRACT
This paper presents a method to segment brain tissue from T1-weighted Magnetic Resonance (MR) images. A modified BayesShrink method is utilized to filter the image in wavelet transform domain before segmentation, where the shrinkage strength is automatically adjusted with respect to noise level. Then the fuzzy c-means clustering is applied to segment brain tissue into cerebrospinal fluid, gray matter and white matter. Comparison with other methods for brain tissue segmentation that remove noise using wavelet or non-wavelet based methods is made on phantom or real data and the advantage of the proposed method is demonstrated.