Skip to main content
eScholarship
Open Access Publications from the University of California

Scholarly Works (13 results)

  • Thesis
  • Peer Reviewed
UCLA Electronic Theses and Dissertations (2012)

Glioblastoma multiforme (GBM) is the most common and aggressive type of primary brain tumor. Magnetic resonance (MR) imaging plays an important role in the detection of brain tumors and treatment response assessment of drugs in clinical trials. Diffusion weighted magnetic resonance imaging (DW-MRI) has the potential to work as surrogate biomarker to reveal early changes in the tumor microenvironment that precede morphologic tumor changes. In this dissertation, we developed a computer-aided therapeutic response system (CADrx) for GBM brain tumors using T1w post-contrast MR and diffusion-weighted (DW) MR images in clinical trials. There are two components: 1) semi-automated segmentation of GBM brain tumors on T1w post-contrast MR images; 2) prediction of volumetric treatment response using early ADC values derived from the DW-MRI. The first component is the main focus of the dissertation. The overall goal is to first facilitate radiologists in the time-consuming task of tumor contouring and generate as reproducible segmentation as possible, and then collect potential features automatically and use machine learning techniques to explore patterns in the large-scale dataset. By doing this, we aim to provide radiologists a second opinion in tumor contouring and therapy response classification.

    Cover page: Computer Aided Segmentation and Early Therapeutic Response Classification (CADrx) for Glioblastoma Multiforme (GBM) Brain Tumors with Magnetic Resonance Imaging
    • Thesis
    • Peer Reviewed
    UCLA Electronic Theses and Dissertations (2015)

    Quantitative computed tomography (CT) plays a major role in large-scale, longitudinal, multi-center clinical trials. Minimizing measurement variation by identifying robust CT imaging biomarkers and developing robust techniques for quantitative CT has implications for clinical trial management and for patient care. We investigated robustness with respect to two sources of measurement variation in quantitative CT: repeat-scan variation (reproducibility) and variation due to changing CT technical parameters.

    In this dissertation, we conducted two separate but related studies in the area of quantitative CT robustness. In the first, we characterized and compared the reproducibilities of several widely-accepted measures of emphysema by examining repeat CT images from a multi-center clinical trial taken one week apart. We investigated the influence of breathhold on reproducibility of emphysema measures. We also investigated variations in reproducibility characteristics across sites. Our results have implications for multi-center clinical trials that rely on accurate and reproducible measures of emphysema.

    In the second study, we investigated feature and classifier robustness with respect to slice thickness, reconstruction kernel, and tube current in the setting of classification of fibrotic interstitial lung disease (FILD). We developed a quantitative Robustness Index measure by examining the stability of imaging features across multiple systematic reconstructions of CT raw sinogram data. We proposed a novel Robustness-Driven Feature Selection (RDFS) method for identifying a subset of robust features, then used these features to develop a robust support vector classifier for lung structure and parenchymal abnormalities in FILD. We demonstrated the superior robustness of this classifier compared to a similar classifier that did not utilize RDFS. Our results have implications for improving the robustness of classifier-based CT CAD systems, which is of importance in multi-center clinical trials that rely on imaging biomarkers that can be generalized across many sites and timepoints.

      Cover page: Measurement Variation and Robustness in Quantitative Thoracic Computed Tomography
      • Thesis
      • Peer Reviewed
      UCLA Electronic Theses and Dissertations (2021)

      Radiomic features are quantitative metrics calculated over regions of interest on medical images. Tumor-specific radiomic features can describe tumor characteristics such as shape, attenuation, and tissue heterogeneity. The promise of radiomics to link with tumor biology, treatment outcome, and pathology has been explored extensively. However, radiomics is not yet fully validated as a clinical biomarker. Two crucial steps in validation of radiomics are the assessment of its clinical utility and technical validity. Large multicenter trials are still required to ensure clinical utility of radiomics and the technical validity of radiomics has not been adequately addressed. Radiomic features are quantitative metrics calculated over regions of interest on medical images. Tumor-specific radiomic features can describe tumor characteristics such as shape, attenuation, and tissue heterogeneity. The promise of radiomics to link with tumor biology, treatment outcome, and pathology has been explored extensively. However, radiomics is not yet fully validated as a clinical biomarker. Two crucial steps in validating radiomics are the assessment of its clinical utility and technical validity. Large multicenter trials are still required to ensure the clinical utility of radiomics, and the technical validity of radiomics has not been adequately addressed. Radiomics is data-driven and can get influenced by inconsistencies in image acquisition, image analysis, etc. While recent studies have demonstrated the susceptibility of radiomics to image acquisition, the reproducibility of CT radiomic features is not well established yet. Due to the unavailability of highly controlled datasets, previous efforts have been restricted to phantom data, limited patient cohorts representing narrow CT parameter ranges, or univariable analysis of a few CT parameters. Furthermore, enforcement of harmonization strategies is needed to handle related inconsistencies. Thus far, only a few limited efforts have explored such strategies; however, harmonization of radiomics is not resolved yet, and continued research and evaluations are necessary. This dissertation addressed the existing knowledge gap in understanding the variability of radiomic features and investigated potential strategies for harmonizing the radiomics approach. We investigated the effects of a wide range of CT acquisition and reconstruction parameters (dose, kernel, and slice thickness) on radiomic features in a realistic setting using clinical low-dose lung cancer screening cases. A computational pipeline was used that generated a unique and highly controlled dataset suitable for assessing the technical validity of radiomic features. We performed univariable and multivariable exploration of reproducibility of well-known radiomic features. Only a few features were reproducible in response to variation of dose and kernel, and the majority of radiomic features were impacted by slice thickness. Multivariable analyses revealed interactions among CT parameters, suggesting that selecting specific combinations of CT parameters can adjust for (or worsen) the impact of CT condition variations. We tested and compared two harmonization methods of Generative Adversarial Networks (GAN) and ComBat. A previously developed GAN model, Pix2Pix, was applied to sub-volumes surrounding lung nodules to transform lung nodule images at different CT conditions into harmonized images with radiomic features similar to a designated baseline CT condition. The ComBat method was applied separately to the radiomic feature data to estimate and adjust the deviations of radiomic features of non-baseline CT conditions to the baseline. The two mitigation techniques reduced radiomic feature variabilities at specific dose, kernel, and slice thickness ranges. Our findings advise on the inclusion of a harmonization procedure in the radiomics approach to avoid facing technical challenges in multicenter studies. Harmonization can be achieved via careful radiomic feature selection based on reproducibility or by applying an effective mitigation technique. While further evaluation remains a future, we illustrated the possibility of alleviating some variabilities due to CT image acquisition variations. Hence, there is a potential for the inclusion of these techniques in harmonization procedures. If validated, radiomics can be a valuable tool for clinical decision-making. Our explorations into the reproducibility and harmonization of radiomics contribute to enabling meaningful validation of radiomics.

        Cover page: Computed Tomography Radiomic Features of Lung Nodules: Characterizing Feature Reproducibility Due to Variations in Image Acquisition and Reconstruction Parameters and Investigations into Mitigation Methods
        • Article
        • Peer Reviewed
        UCLA Previously Published Works (2014)

        Purpose

        To assess the intra- and interscan reproducibility of LV twist using FAST. Assessing the reproducibility of the measurement of new MRI biomarkers is an important part of validation. Fourier Analysis of STimulated Echoes (FAST) is a new MRI tissue tagging method that has recently been shown to compare favorably with conventional estimates of left ventricular (LV) twist from cardiac tagged images, but with significantly reduced user interaction time.

        Materials and methods

        Healthy volunteers (N = 10) were scanned twice using FAST over 1 week. On day 1, two measurements of LV twist were collected for intrascan comparisons. Measurements for LV twist were again collected on day 8 for interscan assessment. LV short-axis tagged images were acquired on a 3 Tesla (T) scanner to ensure detectability of tags during early and mid-diastole. Peak LV twist is reported as mean ± SD. Reproducibility was assessed using the concordance correlation coefficient (CCC) and the repeatability coefficient (RC) (95% confidence interval [CI] range).

        Results

        Mean peak twist measurements were 13.4 ± 4.3° (day 1, scan 1), 13.6 ± 3.7° (day 1, scan 2), and 13.0 ± 2.7° (day 8). Bland-Altman analysis resulted in intra- and interscan bias and 95% CI of -0.6° [-1.0°, 1.6°] and 1.4° (-1.0°, 3.0°), respectively. The Bland-Altman RC for peak LV twist was 2.6° and 4.0° for intra- and interscan, respectively. The CCC was 0.9 and 0.6 for peak LV twist for intra- and interscan, respectively.

        Conclusion

        FAST is a semi-automated method that provides a quick and quantitative assessment of LV systolic and diastolic twist that demonstrates high intrascan and moderate interscan reproducibility in preliminary studies.
          Cover page: Intra‐ and interscan reproducibility using Fourier Analysis of STimulated Echoes (FAST) for the rapid and robust quantification of left ventricular twist
          • Article
          • Peer Reviewed
          UCLA Previously Published Works (2019)
          Idiopathic pulmonary fibrosis (IPF) is a fatal lung disease characterized by an unpredictable progressive declinein lung function. Natural history of IPF is unknown and the prediction of disease progression at the time ofdiagnosis is notoriously difficult. High resolution computed tomography (HRCT) has been used for the diagnosisof IPF, but not generally for monitoring purpose. The objective of this work is to develop a novel predictivemodel for the radiological progression pattern at voxel-wise level using only baseline HRCT scans. Mainly, thereare two challenges: (a) obtaining a data set of features for region of interest (ROI) on baseline HRCT scans andtheir follow-up status; and (b) simultaneously selecting important features from high-dimensional space, andoptimizing the prediction performance. We resolved the first challenge by implementing a study design andhaving an expert radiologist contour ROIs at baseline scans, depending on its progression status in follow-upvisits. For the second challenge, we integrated the feature selection with prediction by developing an algorithmusing a wrapper method that combines quantum particle swarm optimization to select a small number of featureswith random forest to classify early patterns of progression. We applied our proposed algorithm to analyzeanonymized HRCT images from 50 IPF subjects from a multi-center clinical trial. We showed that it yields aparsimonious model with 81.8% sensitivity, 82.2% specificity and an overall accuracy rate of 82.1% at the ROIlevel. These results are superior to other popular feature selections and classification methods, in that ourmethod produces higher accuracy in prediction of progression and more balanced sensitivity and specificity witha smaller number of selected features. Our work is the first approach to show that it is possible to use onlybaseline HRCT scans to predict progressive ROIs at 6 months to 1year follow-ups using artificial intelligence.
            Cover page: Prediction of progression in idiopathic pulmonary fibrosis using CT scans atbaseline: A quantum particle swarm optimization - Random forest approach
            • Article
            • Peer Reviewed
            UCLA Previously Published Works (2012)

            Objectives

            To propose and evaluate a technique for automatic quantification of fissural completeness from chest computed tomography (CT) in a database of subjects with severe emphysema.

            Methods

            Ninety-six CT studies of patients with severe emphysema were included. The lungs, fissures and lobes were automatically segmented. The completeness of the fissures was calculated as the percentage of the lobar border defined by a fissure. The completeness score of the automatic method was compared with a visual consensus read by three radiologists using boxplots, rank sum tests and ROC analysis.

            Results

            The consensus read found 49% (47/96), 15% (14/96) and 67% (64/96) of the right major, right minor and left major fissures to be complete. For all fissures visually assessed as being complete the automatic method resulted in significantly higher completeness scores (mean 92.78%) than for those assessed as being partial or absent (mean 77.16%; all p values <0.001). The areas under the curves for the automatic fissural completeness were 0.88, 0.91 and 0.83 for the right major, right minor and left major fissures respectively.

            Conclusions

            An automatic method is able to quantify fissural completeness in a cohort of subjects with severe emphysema consistent with a visual consensus read of three radiologists.

            Key points

            • Lobar fissures are important for assessing the extent and distribution of lung disease • Modern CT allows automatic lobar segmentation and assessment of the fissures • This segmentation can also assess the completeness of the fissures. • Such assessment is important for decisions about novel therapies (eg for emphysema).
              Cover page: A method for the automatic quantification of the completeness of pulmonary fissures: evaluation in a database of subjects with severe emphysema
              • Article
              • Peer Reviewed
              UCLA Previously Published Works (2016)
              Lack of classifier robustness is a barrier to widespread adoption of computer-aided diagnosis systems for computed tomography (CT). We propose a novel Robustness-Driven Feature Selection (RDFS) algorithm that preferentially selects features robust to variations in CT technical factors. We evaluated RDFS in CT classification of fibrotic interstitial lung disease using 3D texture features. CTs were collected for 99 adult subjects separated into three datasets: training, multi-reconstruction, testing. Two thoracic radiologists provided cubic volumes of interest corresponding to six classes: pulmonary fibrosis, ground-glass opacity, honeycombing, normal lung parenchyma, airway, vessel. The multi-reconstruction dataset consisted of CT raw sinogram data reconstructed by systematically varying slice thickness, reconstruction kernel, and tube current (using a synthetic reduced-tube-current algorithm). Two support vector machine classifiers were created, one using RDFS ("with-RDFS") and one not ("without-RDFS"). Classifier robustness was compared on the multi-reconstruction dataset, using Cohen's kappa to assess classification agreement against a reference reconstruction. Classifier performance was compared on the testing dataset using the extended g-mean (EGM) measure. With-RDFS exhibited superior robustness (kappa 0.899-0.989) compared to without-RDFS (kappa 0.827-0.968). Both classifiers demonstrated similar performance on the testing dataset (EGM 0.778 for with-RDFS; 0.785 for without-RDFS), indicating that RDFS does not compromise classifier performance when discarding nonrobust features. RDFS is highly effective at improving classifier robustness against slice thickness, reconstruction kernel, and tube current without sacrificing performance, a result that has implications for multicenter clinical trials that rely on accurate and reproducible quantitative analysis of CT images collected under varied conditions across multiple sites, scanners, and timepoints.
                • Article
                • Peer Reviewed
                UCLA Previously Published Works (2014)

                Objectives

                The purpose of this study was to define clinically appropriate, computer-aided lung nodule detection (CAD) requirements and protocols based on recent screening trials. In the following paper, we describe a CAD evaluation methodology based on a publically available, annotated computed tomography (CT) image data set, and demonstrate the evaluation of a new CAD system with the functionality and performance required for adoption in clinical practice.

                Methods

                A new automated lung nodule detection and measurement system was developed that incorporates intensity thresholding, a Euclidean Distance Transformation, and segmentation based on watersheds. System performance was evaluated against the Lung Imaging Database Consortium (LIDC) CT reference data set.

                Results

                The test set comprised thin-section CT scans from 108 LIDC subjects. The median (±IQR) sensitivity per subject was 100 (±37.5) for nodules ≥ 4 mm and 100 (±8.33) for nodules ≥ 8 mm. The corresponding false positive rates were 0 (±2.0) and 0 (±1.0), respectively. The concordance correlation coefficient between the CAD nodule diameter and the LIDC reference was 0.91, and for volume it was 0.90.

                Conclusions

                The new CAD system shows high nodule sensitivity with a low false positive rate. Automated volume measurements have strong agreement with the reference standard. Thus, it provides comprehensive, clinically-usable lung nodule detection and assessment functionality.

                Key points

                • CAD requirements can be based on lung cancer screening trial results. • CAD systems can be evaluated using publically available annotated CT image databases. • A new CAD system was developed with a low false positive rate. • The CAD system has reliable measurement tools needed for clinical use.
                  • Article
                  • Peer Reviewed
                  UCLA Previously Published Works (2023)

                  Background

                  Idiopathic pulmonary fibrosis (IPF) is a progressive, irreversible, and usually fatal lung disease of unknown reasons, generally affecting the elderly population. Early diagnosis of IPF is crucial for triaging patients' treatment planning into anti-fibrotic treatment or treatments for other causes of pulmonary fibrosis. However, current IPF diagnosis workflow is complicated and time-consuming, which involves collaborative efforts from radiologists, pathologists, and clinicians and it is largely subject to inter-observer variability.

                  Purpose

                  The purpose of this work is to develop a deep learning-based automated system that can diagnose subjects with IPF among subjects with interstitial lung disease (ILD) using an axial chest computed tomography (CT) scan. This work can potentially enable timely diagnosis decisions and reduce inter-observer variability.

                  Methods

                  Our dataset contains CT scans from 349 IPF patients and 529 non-IPF ILD patients. We used 80% of the dataset for training and validation purposes and 20% as the holdout test set. We proposed a two-stage model: at stage one, we built a multi-scale, domain knowledge-guided attention model (MSGA) that encouraged the model to focus on specific areas of interest to enhance model explainability, including both high- and medium-resolution attentions; at stage two, we collected the output from MSGA and constructed a random forest (RF) classifier for patient-level diagnosis, to further boost model accuracy. RF classifier is utilized as a final decision stage since it is interpretable, computationally fast, and can handle correlated variables. Model utility was examined by (1) accuracy, represented by the area under the receiver operating characteristic curve (AUC) with standard deviation (SD), and (2) explainability, illustrated by the visual examination of the estimated attention maps which showed the important areas for model diagnostics.

                  Results

                  During the training and validation stage, we observe that when we provide no guidance from domain knowledge, the IPF diagnosis model reaches acceptable performance (AUC±SD = 0.93±0.07), but lacks explainability; when including only guided high- or medium-resolution attention, the learned attention maps are not satisfactory; when including both high- and medium-resolution attention, under certain hyperparameter settings, the model reaches the highest AUC among all experiments (AUC±SD = 0.99±0.01) and the estimated attention maps concentrate on the regions of interests for this task. Three best-performing hyperparameter selections according to MSGA were applied to the holdout test set and reached comparable model performance to that of the validation set.

                  Conclusions

                  Our results suggest that, for a task with only scan-level labels available, MSGA+RF can utilize the population-level domain knowledge to guide the training of the network, which increases both model accuracy and explainability.
                    Cover page: Multi‐scale, domain knowledge‐guided attention + random forest: a two‐stage deep learning‐based multi‐scale guided attention models to diagnose idiopathic pulmonary fibrosis from computed tomography images
                    • Article
                    • Peer Reviewed
                    UCLA Previously Published Works (2018)

                    Purpose

                    We present a method for generating a T2 MR-based probabilistic model of tumor occurrence in the prostate to guide the selection of anatomical sites for targeted biopsies and serve as a diagnostic tool to aid radiological evaluation of prostate cancer.

                    Materials and methods

                    In our study, the prostate and any radiological findings within were segmented retrospectively on 3D T2-weighted MR images of 266 subjects who underwent radical prostatectomy. Subsequent histopathological analysis determined both the ground truth and the Gleason grade of the tumors. A randomly chosen subset of 19 subjects was used to generate a multi-subject-derived prostate template. Subsequently, a cascading registration algorithm involving both affine and non-rigid B-spline transforms was used to register the prostate of every subject to the template. Corresponding transformation of radiological findings yielded a population-based probabilistic model of tumor occurrence. The quality of our probabilistic model building approach was statistically evaluated by measuring the proportion of correct placements of tumors in the prostate template, i.e., the number of tumors that maintained their anatomical location within the prostate after their transformation into the prostate template space.

                    Results

                    Probabilistic model built with tumors deemed clinically significant demonstrated a heterogeneous distribution of tumors, with higher likelihood of tumor occurrence at the mid-gland anterior transition zone and the base-to-mid-gland posterior peripheral zones. Of 250 MR lesions analyzed, 248 maintained their original anatomical location with respect to the prostate zones after transformation to the prostate.

                    Conclusion

                    We present a robust method for generating a probabilistic model of tumor occurrence in the prostate that could aid clinical decision making, such as selection of anatomical sites for MR-guided prostate biopsies.
                      Cover page: Building a high-resolution T2-weighted MR-based probabilistic model of tumor occurrence in the prostate
                      Top