Publications by authors named "Jennifer Xiong"

4 Publications

  • Page 1 of 1

Improving robustness of a deep learning-based lung-nodule classification model of CT images with respect to image noise.

Phys Med Biol 2021 Nov 24. Epub 2021 Nov 24.

Department of Radiation Oncology, UT Southwestern Medical Center, 6363 Forest Park Rd. BL10.202G, MC9315, Dallas, Texas, 75390-9315, UNITED STATES.

Objective: Robustness is an important aspect to consider, when developing methods for medical image analysis. This study investigated robustness properties of deep neural networks (DNNs) for a lung nodule classification problem based on CT images and proposed a solution to improve robustness.

Approach: We firstly constructed a class of four DNNs with different widths, each predicting an output label (benign or malignant) for an input CT image cube containing a lung nodule. These networks were trained to achieve Area Under the Curve of 0.891-0.914 on a testing dataset. We then added to the input CT image cubes noise signals generated randomly using a realistic CT image noise model based on a noise power spectrum at 100 mAs, and monitored the DNN's output change. We defined $SAR_{5} (\%)$ to quantify the robustness of the trained DNN model, indicating that for $5\%$ of CT image cubes, the noise can change the prediction results with a chance of at least $SAR_{5} (\%)$. To understand robustness, we viewed the information processing pipeline by the DNN as a two-step process, with the first step using all but the last layers to extract representations of the input CT image cubes in a latent space, and the second step employing the last fully-connected layer as a linear classifier to determine the position of the sample representations relative to a decision plane. To improve robustness, we proposed to retrain the last layer of the DNN with a Supporting Vector Machine (SVM) hinge loss function to enforce the desired position of the decision plane.

Main Results: $SAR_{5}$ ranged in $47.0\sim 62.0\%$ in different DNNs. The unrobustness behavior may be ascribed to the unfavorable placement of the decision plane in the latent representation space, which made some samples be perturbed to across the decision plane and hence susceptible to noise. The DNN-SVM model improved robustness over the DNN model and reduced $SAR_{5}$ by $8.8\sim 21.0\%$.

Significance: This study provided insights about the potential reason for the unrobustness behavior of DNNs and the proposed DNN-SVM model improved model robustness.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1088/1361-6560/ac3d16DOI Listing
November 2021

Anesthetic and obstetric outcomes in pregnant women undergoing cesarean delivery according to body mass index: Retrospective analysis of a single-center experience.

Ann Med Surg (Lond) 2018 Dec 2;36:129-134. Epub 2018 Nov 2.

Department of Anesthesiology and Perioperative Medicine, Medical College of Georgia at Augusta University, USA.

Aim: To evaluate maternal, neonatal and anesthetic outcomes according to BMI in women undergoing cesarean section.

Background: Increased incidence rates of obesity and morbid obesity have been reported in the United States. Pregnant obese patients are at increased risk of maternal and fetal complications, and obstetric and anesthetic management of these patients is especially challenging.

Methods: A retrospective chart review of patients who underwent cesarean section in a single center between 2015 and 2016 was conducted. Anesthetic, obstetric and neonatal outcomes were analyzed in relation to levels of BMI.

Results: Seven hundred and seventy one patients underwent cesarean section during the study period. The number of patients with normal BMI, obesity and morbid obesity was 213 (27.6%), 365 (47.3%) and 193 (25%), respectively. Sixty-one percent of the patients in morbidly obese group had at least one comorbidity (p < 0.01). We found no significant differences with respect to perioperative obstetric complications. Intraoperative blood loss was significantly higher in the morbidly obese group.

Conclusion: Increasing BMI is associated with comorbidities such as hypertension and diabetes mellitus, and with increased intraoperative blood loss. We were unable to detect differences in other obstetric, anesthetic and neonatal outcomes.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1016/j.amsu.2018.10.023DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC6234280PMC
December 2018

Growth factor priming differentially modulates components of the extracellular matrix proteome in chondrocytes and synovium-derived stem cells.

PLoS One 2014 7;9(2):e88053. Epub 2014 Feb 7.

Department of Biomedical Engineering, Columbia University, New York, New York, United States of America.

To make progress in cartilage repair it is essential to optimize protocols for two-dimensional cell expansion. Chondrocytes and SDSCs are promising cell sources for cartilage repair. We previously observed that priming with a specific growth factor cocktail (1 ng/mL transforming growth factor-β1, 5 ng/mL basic fibroblast growth factor, and 10 ng/mL platelet-derived growth factor-BB) in two-dimensional culture, led to significant improvement in mechanical and biochemical properties of synovium-derived stem cell (SDSC)-seeded constructs. The current study assessed the effect of growth factor priming on the proteome of canine chondrocytes and SDSCs. In particular, growth factor priming modulated the proteins associated with the extracellular matrix in two-dimensional cultures of chondrocytes and SDSCs, inducing a partial dedifferentiation of chondrocytes (most proteins associated with cartilage were down-regulated in primed chondrocytes) and a partial differentiation of SDSCs (some collagen-related proteins were up-regulated in primed SDSCs). However, when chondrocytes and SDSCs were grown in pellet culture, growth factor-primed cells maintained their chondrogenic potential with respect to glycosaminoglycan and collagen production. In conclusion, the strength of the label-free proteomics technique is that it allows for the determination of changes in components of the extracellular matrix proteome in chondrocytes and SDSCs in response to growth factor priming, which could help in future tissue engineering strategies.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0088053PLOS
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3917883PMC
May 2015

Deciphering the MSG controversy.

Int J Clin Exp Med 2009 Nov 15;2(4):329-36. Epub 2009 Nov 15.

Robert S. Dow Neurobiology Laboratories, Legacy Clinical Research Center Portland, Oregon, USA.

Monosodium glutamate (MSG), a common flavor enhancer in various canned food and stereotypically associated with food in Chinese restaurants, has been claimed and tested to have side effects including headache and dizziness. However, the mechanism behind MSG-induced headache was not clear. Using dissociated mouse neuronal culture and cell injury assays, we determined whether incubation of neurons with clinically relevant concentrations of MSG induces cell swelling or death, and whether any measure can be taken to prevent or reduce MSG effects. We demonstrated that (1) Treatment with MSG induces a dose-dependent swelling and death of mature neurons (12-14 days in culture) with little effect on young immature neurons (<1 week in culture). The threshold concentration of MSG for neuronal injury is 3 microM; (2) MSG only injures neurons with little effect on glial cells; (3) Boiling MSG does not affect its toxicity but the addition of Vitamin C provides significant protection against MSG toxicity; (4) Pretreatment of neurons with a low dose of MSG reduces subsequent injury by a large dose of MSG. Together, our studies suggest that the side effect of MSG may be mediated, at least in part, by its toxic effect on brain neurons. Pre-exposure to low doses of MSG or the use of Vitamin C may prevent or reduce the side effects of MSG.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2802046PMC
November 2009
-->