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

Scholarly Works (27 results)

  • Thesis
  • Peer Reviewed
UC San Francisco Electronic Theses and Dissertations (2024)

Prostate cancer progression after radical prostatectomy poses a significant risk to patient health. The ability to predict which patients are at a higher risk of progression is crucial for determining appropriate adjuvant therapies. This study investigates the application of convolutional neural networks (CNNs) to pre-surgical multiparametric MRI (mpMRI) for predicting post-surgical prostate cancer progression. The study utilizes a retrospective patient cohort and explores the performance of different CNN architectures (ResNet and DenseNet), normalization methods, and slice selection techniques. The results demonstrate the potential of CNNs in predicting prostate cancer progression, with the best-performing model achieving an accuracy of 0.712. The study highlights the importance of appropriate image normalization and slice selection methods for optimal performance. The findings suggest that CNNs could serve as a valuable tool for aiding clinical decision-making in prostate cancer management.

    Cover page: Application of Convolutional Neural Networks in Multiparametric MR Imaging to Predict Prostate Cancer Progression
    • Article
    • Peer Reviewed
    UC San Francisco Previously Published Works (2017)
    The purpose of this study was to determine whether 5α-reductase inhibitors (5-ARIs) affect the discrimination between low-grade prostate cancer and benign tissues on multiparametric MRI (mpMRI). Twenty men with biopsy-proven Gleason 3 + 3 prostate cancer and 3 T mpMRI were studied. Ten patients (Tx) had been receiving 5-ARIs for at least a year at scan time. Ten untreated patients (Un) were matched to the treated cohort. For each subject two regions of interest representing cancerous and benign tissues were drawn within the peripheral zone of each prostate, MR measures evaluated, and cancer contrast versus benign (contrast = (MRTumor  - MRHealthy )/MRHealthy ) calculated. Decreased cancer contrast was noted on T2 -weighted images: 0.4 (Un) versus 0.3 (Tx). However, for functional MR measures, a better separation of cancerous and benign tissues was observed in the treated group. Cancer contrast on high-b diffusion-weighted imaging (DWI) was 0.61 (Un) versus 0.99 (Tx). Logistic regression analysis yielded higher AUC (area under the curve) values for distinguishing cancerous from benign regions in treated subjects on high-b DWI (0.71 (Un), 0.94 (Tx)), maximal enhancement slope (0.95 (Un), 1 (Tx)), peak enhancement (0.84 (Un), 0.93 (Tx)), washout slope (0.78 (Un), 0.99 (Tx)), Ktrans (0.9 (Un), 1 (Tx)), and combined measures (0.86 (Un), 0.99 (Tx)). Coefficients of variation for MR measures were lower in benign and cancerous tissues in the treated group compared with the untreated group. This study's results suggest an increase in homogeneity of benign and malignant peripheral zone prostatic tissues with 5-ARI exposure, observed as reduced variability of MR measures after treatment. Cancer discrimination was lower with T2 -weighted imaging, but was higher with functional MR measures in a 5-ARI-treated cohort compared with controls.
      Cover page: Improved multiparametric MRI discrimination between low‐risk prostate cancer and benign tissues in a small cohort of 5α‐reductase inhibitor treated individuals as compared with an untreated cohort
      • Thesis
      • Peer Reviewed
      UC San Francisco Electronic Theses and Dissertations (2014)

      The purpose of this study was to create a semi-automatic segmentation of the prostate for an accurate estimation of prostate volume and modeling of longitudinal changes in patient data. Segmentation algorithms are available based on axial T2-weighted imaging, but decrease in accuracy in abnormal or treated data, and can be expensive. In this work, we segment based on a fat fraction map (FF map), due to the smaller change in image intensities after treatment and large contrast between the prostate and surrounding fatty tissue. The algorithm consists of five parts: a global filtering of the image; region growth from a predetermined seed point; spline interpolation of the region-enclosing polygon; a comparison of distances between boundary points on adjacent slices; and backward mapping to an upsampled plane with a final volume extrusion. Image artifact is removed by morphological opening after the initial region growth, and by removing boundary points with a shortest distance to an adjacent slice boundary greater than a predetermined threshold. FF maps emphasizing low resolution in favor of high contrast (FF_hc maps), and FF maps emphasizing low contrast in favor of high resolution (FF_hr maps) using different acquisition parameters were tested for accuracy against a manual segmentation drawn on a T2-weighted image in fourteen patients receiving a multiparametric MR exam for confirmed or suspected prostate cancer. There was no significant difference in the volumes recorded from the trial data for the semi-automatic segmentation of the FF_hr map and manual segmentation of the axial T2-weighted image (n=14, p<0.84, paired students t-test), but FF_hr map segmentation had a trend to underestimate the size of the prostate gland. The FF_hc map segmentation volumes were significantly different from those of the axial T2-weighted images (n=14, p<0.03, paired student's t-test). The segmentations of FF_hc maps showed psoas muscle invasion into the prostate region as the most common artifact due to the anatomic proximity and signal similarity, and both FF map types showed ambiguity as the prostate base abuts the bladder, leading to both over- and underestimation. This quick segmentation using FF_hr maps creates accurate prostate volume estimations and should be pursued.

        Cover page: Semi-automatic Segmentation of the Prostate by MR Fat Fraction Map
        • Thesis
        • Peer Reviewed
        UC San Francisco Electronic Theses and Dissertations (2024)

        Disease conditions like obstructive sleep apnea (OSA) and human immunodeficiency virus (HIV) infections are characterized by chronic low-grade inflammation, leading to poor metabolic health. This work is focused on comparing MR-based fat measures to indicators of worse metabolic health, namely OSA severity and hydroxyproline, a biomarker of subcutaneous fat (SAT) fibrosis. The study also explored the utility of novel MRI methods of diffusion-weighted imaging of fat and T1 mapping to detect fibrosis in the SAT. There were 33 participants with OSA and 58 participants with or without HIV infection who had hydroxyproline measured via SAT biopsy, 13 of whom had the novel MRI measures of fibrosis. The liver, visceral, and SAT volumes were segmented using artificial intelligence-based methods, and the pancreas was manually drawn on proton density fat fraction (PDFF) images. Apparent diffusion coefficient (ADC) and T1 were measured in regions of interest drawn in the SAT. Liver fat fraction, liver fat content, and pancreatic fat fraction were higher with increased severities of sleep apnea, p0.01. Furthermore, liver fat fraction, pancreatic fat fraction, and visceral fat volume were higher in subjects with high hydroxyproline levels on biopsy, p0.03. The ADC of SAT at b-values of 3×10-3 s/mm2 were negatively correlated with hydroxyproline levels. It can be concluded that more severe OSA and higher hydroxyproline were associated with worse metabolic health. The role of diffusion-weighted imaging of fat with high b-values to detect SAT fibrosis is also encouraging and opens up avenues for future research.

          Cover page: MRI - Based Measures of Metabolic Health in the Assessment of Patients with Chronic Inflammatory States
          • Article
          • Peer Reviewed
          UC San Francisco Previously Published Works (2015)
          The purposes of this study were to determine if image distortion is less in prostate MR apparent diffusion coefficient (ADC) maps generated from a reduced-field-of-view (rFOV) diffusion-weighted-imaging (DWI) technique than from a conventional DWI sequence (CONV), and to determine if the rFOV ADC tumor contrast is as high as or better than that of the CONV sequence. Fifty patients underwent a 3T MRI exam. CONV and rFOV (utilizing a 2D, echo-planar, rectangularly-selective RF pulse) sequences were acquired using b=600, 0s/mm(2). Distortion was visually scored 0-4 by three independent observers and quantitatively measured using the difference in rectal wall curvature between the ADC maps and T2-weighted images. Distortion scores were lower with the rFOV sequence (p<0.012, Wilcoxon Signed-Rank Test, n=50), and difference in distortion scores did not differ significantly among observers (p=0.99, Kruskal-Wallis Rank Sum Test). The difference in rectal curvature was less with rFOV ADC maps (26%±10%) than CONV ADC maps (34%±13%) (p<0.011, Student's t-test). In seventeen patients with untreated, biopsy confirmed prostate cancer, the rFOV sequence afforded significantly higher ADC tumor contrast (44.0%) than the CONV sequence (35.9%), (p<0.0012, Student's t-test). The rFOV sequence yielded significantly decreased susceptibility artifact and significantly higher contrast between tumor and healthy tissue.
            Cover page: Reduced-FOV excitation decreases susceptibility artifact in diffusion-weighted MRI with endorectal coil for prostate cancer detection
            • Article
            • Peer Reviewed
            UC San Francisco Previously Published Works (2012)

            Purpose

            To determine if the concordance of magnetic resonance (MR) imaging and MR spectroscopic data with histologic measures of steatosis is affected by histologic magnification level, tissue heterogeneity, or assessment of tissue area versus that of hepatocytes.

            Materials and methods

            This study was institutional review board approved and HIPAA compliant. Written informed consent was obtained. In- and out-of-phase MR imaging and MR spectroscopic measures of steatosis were compared in 33 patients with nonalcoholic fatty liver disease and in 15 healthy volunteers. Concordance of MR imaging and MR spectroscopic data with histologic findings was assessed for (a) histologic examination at standard (×40 and ×100) versus high magnification (×200 and ×400), (b) heterogeneity and homogeneity of livers, and (c) percentage of tissue and hepatocytes that contained lipids. Evaluations included linear regression and Fisher exact tests.

            Results

            In- and out-of-phase MR imaging and MR spectroscopic data were well correlated (R2=0.93) and generally concordant with histologic measures. Patients in whom MR fat fractions were higher than expected compared with steatosis grades at standard magnification histologic examination were upgraded significantly more often when high magnification was used than were the remaining patients (100% [10 of 10] vs 47% [7 of 15], P<.01). MR imaging and MR spectroscopic data of homogeneous livers were significantly more likely than those of heterogeneous livers to be concordant with steatosis grades when high magnification was used (81% [13 of 16] vs 47% [8 of 17], P<.05). For all patients, percentage of fat in tissue was lower than that in hepatocytes, which affected individual patients, but not the overall correlation.

            Conclusion

            MR imaging and MR spectroscopic data were generally concordant with histologic measures of steatosis. Discordance between them may reflect differences in magnification at histologic examination and in liver heterogeneity.
              Cover page: Liver Steatosis: Concordance of MR Imaging and MR Spectroscopic Data with Histologic Grade
              • Article
              • Peer Reviewed
              UC San Francisco Previously Published Works (2016)
              The use of multiparametric MRI scans for the evaluation of men with prostate cancer has increased dramatically and is likely to continue expanding as new developments come to practice. However, it has not yet gained the same level of acceptance of other imaging tests. Partly, this is because of the use of suboptimal protocols, lack of standardization, and inadequate patient preparation. In this manuscript, we describe several practical aspects of prostate MRI that may facilitate the implementation of new prostate imaging programs or the expansion of existing ones.
                Cover page: Practical aspects of prostate MRI: hardware and software considerations, protocols, and patient preparation.
                • Article
                • Peer Reviewed
                UC San Francisco Previously Published Works (2022)
                In this study, the objective was to characterize the MR signatures of the various benign prostate tissues and to differentiate them from cancer. Data was from seventy prostate cancer patients who underwent multiparametric MRI (mpMRI) and subsequent prostatectomy. The scans included T2-weighted imaging (T2W), diffusion weighted imaging, dynamic contrast-enhanced MRI (DCE MRI), and MR spectroscopic imaging. Histopathology tissue information was translated to MRI images. The mpMRI parameters were characterized separately per zone and by tissue type. The tissues were ordered according to trends in tissue parameter means. The peripheral zone tissue order was cystic atrophy, high grade prostatic intraepithelial neoplasia (HGPIN), normal, atrophy, inflammation, and cancer. Decreasing values for tissue order were exhibited by ADC (1.8 10-3 mm2/s to 1.2 10-3 mm2/s) and T2W intensity (3447 to 2576). Increasing values occurred for DCE MRI peak (143% to 157%), DCE MRI slope (101%/min to 169%/min), fractional anisotropy (FA) (0.16 to 0.19), choline (7.2 to 12.2), and choline / citrate (0.3 to 0.9). The transition zone tissue order was cystic atrophy, mixed benign prostatic hyperplasia (BPH), normal, atrophy, inflammation, stroma, anterior fibromuscular stroma, and cancer. Decreasing values occurred for ADC (1.6 10-3 mm2/s to 1.1 10-3 mm2/s) and T2W intensity (2863 to 2001). Increasing values occurred for DCE MRI peak (143% to 150%), DCE MRI slope (101%/min to 137%/min), FA (0.18 to 0.25), choline (7.9 to 11.7), and choline / citrate (0.3 to 0.7). Logistic regression was used to create parameter model fits to differentiate cancer from benign prostate tissues. The fits achieved AUCs ≥0.91. This study quantified the mpMRI characteristics of benign prostate tissues and demonstrated the capability of mpMRI to discriminate among benign as well as cancer tissues, potentially aiding future discrimination of cancer from benign confounders.
                  Cover page: Identification of prostate cancer using multiparametric MR imaging characteristics of prostate tissues referenced to whole mount histopathology
                  • Article
                  • Peer Reviewed
                  UC San Francisco Previously Published Works (2023)
                  Multi-parametric MRI (mpMRI) has proven itself a clinically useful tool to assess prostate cancer (PCa). Our objective was to generate PCa risk maps to quantify the volume and location of both all PCa and high grade (Gleason grade group ≥ 3) PCa. Such capabilities would aid physicians and patients in treatment decisions, targeting biopsy, and planning focal therapy. A cohort of men with biopsy proven prostate cancer and pre-prostatectomy mpMRI were studied. PCa and benign ROIs (1524) were identified on mpMRI and histopathology with histopathology serving as the reference standard. Logistic regression models were created to differentiate PCa from benign tissues. The MRI images were registered to ensure correct overlay. The cancer models were applied to each image voxel within prostates to create probability maps of cancer and of high-grade cancer. Use of an optimum probability threshold quantified PCa volume for all lesions >0.1 cc. Accuracies were calculated using area under the curve (AUC) for the receiver operating characteristic (ROC). The PCa models utilized apparent diffusion coefficient (ADC), T2 weighted (T2W), dynamic contrast-enhanced MRI (DCE MRI) enhancement slope, and DCE MRI washout as the statistically significant MRI scans. Application of the PCa maps method provided total PCa volume and individual lesion volumes. The AUCs derived from lesion analysis were 0.91 for all PCa and 0.73 for high-grade PCa. At the optimum threshold, the PCa maps detected 135 / 150 (90%) histopathological lesions >0.1 cc. This study showed the feasibility of cancer risk maps, created from pre-prostatectomy, mpMR images validated with histopathology, to detect PCa lesions >0.1 cc. The method quantified the volume of cancer within the prostate. Method improvements were identified by determining root causes for over and underestimation of cancer volumes. The maps have the potential for improved non-invasive capability in quantitative detection, localization, volume estimation, and MRI characterization of PCa.
                    Cover page: Prostate cancer lesion detection, volume quantification and high-grade cancer differentiation using cancer risk maps derived from multiparametric MRI with histopathology as the reference standard
                    • Article
                    • Peer Reviewed
                    UC San Francisco Previously Published Works (2019)

                    Objective

                    To investigate the link between dietary sugar consumption and two separate pathogenetic mechanisms associated with metabolic syndrome: de novo lipogenesis (DNL) and nonenzymatic glycation.

                    Design and participants

                    We assessed changes in serum d-lactate (the detoxification end-product of methylglyoxal) concentration in response to 9 days of isocaloric fructose restriction in 20 children with obesity and metabolic syndrome, and examined correlations with changes in DNL, liver fat, insulin sensitivity, and other metrics of hepatic metabolism.

                    Interventions

                    Nine days of dietary sugar restriction, with substitution of equal amounts of refined starch.

                    Main outcome measures

                    On days 0 and 10, children had laboratory evaluation of d-lactate levels and other analytes, and underwent oral glucose tolerance testing, magnetic resonance spectroscopy to quantify fat depots, and 13C-acetate incorporation into triglyceride (TG) to measure DNL.

                    Results

                    d-Lactate was associated with baseline liver fat fraction (P < 0.001) and visceral adipose tissue (P < 0.001) but not with subcutaneous adipose tissue. At baseline, d-lactate was positively correlated with DNL-area under the curve (AUC) (P = 0.003), liver fat fraction (P = 0.02), TG (P = 0.004), and TG/high-density lipoprotein ratio (P = 0.002). After 9 days of isocaloric fructose restriction, serum d-lactate levels reduced by 50% (P < 0.0001), and changes in d-lactate correlated with both changes in DNL-AUC and measures of insulin sensitivity.

                    Conclusion

                    Baseline correlation of d-lactate with DNL and measures of insulin sensitivity and reduction in d-lactate after 9 days of isocaloric fructose restriction suggest that DNL and nonenzymatic glycation are functionally linked via intermediary glycolysis in the pathogenesis of metabolic syndrome and point to fructose as a key dietary substrate that drives both pathways.
                      Cover page: Isocaloric Fructose Restriction Reduces Serum d-Lactate Concentration in Children With Obesity and Metabolic Syndrome
                      Top