Deep Learning-Based Multi-Modality Segmentation of Primary Gross Tumor Volume in CT and MRI for Nasopharyngeal Carcinoma

PURPOSE/OBJECTIVE(S)

The delineation of primary gross tumor volume (GTV) of nasopharyngeal carcinoma (NPC) is an essential step for radiotherapy planning. In clinical practice, radiation oncologists manually delineate the GTV in planning CT with the help of diagnostic MRI. This is because NPC tumors are closely adjacent to many important anatomic structures, and CT and MRI provide complementary strength to accurately determine the tumor extension boundary. Manual delineation is time-consuming with the potential registration errors between MRI and CT decreasing the delineation accuracy. In this study, we propose a fully automated GTV segmentation method based on CT and MRI by first aligning MRI to CT, and then, segmenting the GTV using a multi-modality deep learning model.

MATERIALS/METHODS

We collected 104 nasopharyngeal carcinoma patients with both planning CT and diagnostic MRI scans (T1 & T2 phases). An experienced radiation oncologists manually delineated the GTV, which was further examined by another senior radiation oncologist. Then, a coarse to fine cross-modality registration from MRI to CT was conducted as follows: (1) A rigid transformation was performed on MRI to roughly align MRI to CT with similar anatomic position. (2) Then, the region of interest (RoI) on both CT and rigid-transformed MRI were cropped. (3) A leading cross-modality deformable registration algorithm, named DEEDS, was applied on the cropped MRI and CT RoIs to find an accurate local alignment. Next, using CT and registered MRI as the combined input, a multi-modality deep segmentation network based on nnUNet was trained to generate the GTV prediction. 20% patients were randomly selected as the unseen testing set to quantitatively evaluate the performance.

RESULTS

The quantitative NPC GTV segmentation performance is summarized in Table 1. The deep segmentation model using CT alone achieved reasonable high performance with 76.6% Dice score and 1.34mm average surface distance (ASD). When both CT and registered MRI were used, the segmentation model further improved the performance by 0.9% Dice score increase and 11% relative ASD error reduction, demonstrating the complementary strength of CT and MRI in determining NPC GTV. Notably, the achieved 77.5% Dice score and 1.19mm ASD by the multimodality model is among the top performing results reported in recent automatic NPC GTV segmentation using either CT or MRI modality.

CONCLUSION

We developed a fully automated multi-modal deep-learning model for NPC GTV segmentation. The developed model can segment the NPC GTV in high accuracy. With further optimization and validation, this automated model has potential to standardize the NPC GTV segmentation and significantly decrease the workload of radiation oncologists in clinical practice.

Deep Learning for Automatic Prediction of Lymph Node Station Metastasis in Esophageal Cancer Patients from Contrast-Enhanced CT

PURPOSE/OBJECTIVE(S)

The diagnosis of lymph node (LN) metastasis in computed tomography (CT) is an essential yet challenging task in esophageal cancer staging and treatment planning. Although criteria (e.g., RECIST, morphological/texture features) are proposed to predict LN metastasis, the diagnostic accuracy remains low with sensitivity <50% and specificity <75%, as reported in previous studies. Deep learning (DL) has the potential to address this issue by learning from large-scale labeled data. However, due to the practical surgery procedure in lymph node dissection, it is difficult to pair the metastasis of individual LN reported in the pathology report to the LN instance found in the CT image. Hence, in this study, we first use pathology reports to determine the LNS metastasis, then develop a multiple instance deep learning (MIDL) model to predict lymph node station (LNS) metastasis.

MATERIALS/METHODS

We collected 1200 esophageal cancer patients with preoperative contrast-enhanced CT before surgery. A recently developed automatic mediastinal LNS segmentation model was first applied to segment LNS of 1 to 8 based on the IASLC protocol. For each LNS, the local CT region of interest (ROI) was cropped to generate a station-wise CT patch, where the LNS was labeled as metastatic if at least one metastatic LN was indicated in the pathology report. Using the station-wise CT patch and LNS label, we train a 3D MIDL model, MobileNetV3, to predict LNS metastasis. To better provide the LN position priors in MIDL, LN instances (with a short axis >4mm) were also segmented using an automatic LN detection algorithm and were added to the MIDL model as an auxiliary input. Five-fold cross-validation was conducted to evaluate the MIDL performance.

RESULTS

The MIDL model's performance is summarized in Table 1. The MIDL model incorporating an additional LN instance mask demonstrated a superior overall AUC of 0.7539, surpassing the model without the LN mask input by 2.93%. The specificity was evaluated at a threshold resulting in a recall of 0.7, and the best model outperformed the CT input model in terms of specificity by 2.11%. This highlights the value of including the LN position prior to the MIDL model. Notably, when a threshold was set to result in a specificity of 75%, the best MIDL model demonstrated a significantly higher recall compared to the previously reported clinical diagnostic recall (39.7% vs. 63.21%).

CONCLUSION

We developed a MIDL classification model to predict LNS metastasis using CT scans of 1200 patients. Our findings suggest that the MIDL model can substantially improve LNS metastasis prediction and has the potential to play an essential role in cancer staging, treatment planning, and prognostic analysis.

Anatomy-Guided Deep Learning Model for Accurate and Robust Gross Tumor Volume Segmentation in Lung Cancer Radiation Therapy

PURPOSE/OBJECTIVE(S)

In lung cancer radiation therapy, clinicians must outline the gross tumor volume (GTV) precisely on the planning computed tomography (pCT) for accurate radiation dose delivery. However, due to the limited contrast between tumor and normal tissues in lung parenchyma, accurate delineation of tumor boundaries is difficult leading to large inter-observer variation. In this study, we develop an anatomy-guided lung GTV deep segmentation model using a training cohort of multi-center datasets. The quantitative segmentation performance is evaluated on an independent dataset, where the inter-observer delineation variation is also assessed.

MATERIALS/METHODS

We collected and curated four publicly available lung datasets with GTV annotations (Lung-PET-CT-Dx, LIDC-IDRI, NSCLC-Radiogenomics and RIDER-CT) for deep learning model development. A total of 871 CT scans of patients, who were diagnosed with T1-T4 NSCLC, were available for training after data curation. The GTV annotations of primary tumor were examined and edited by two experienced radiation oncologists following the RTOG 1106 protocol. An anatomy-guided deep learning model was proposed, which consisted two deep networks. The first deep network used CT scan as input and segmented 4 anatomic organs (airway, heart, pulmonary artery and pulmonary vein), while the second deep network took both CT scan and these pre-segmented 4 organs as input and segmented the lung GTV. With the help of anatomic priors from 4 pre-segmented organs, the second deep network could more easily locate the GTV. We used nnUNet as the deep segmentation network. For evaluation, we used NSCLC-Radiomics as the testing dataset, which contains 20 CT scans each annotated by 5 radiation oncologists. The auto-segmented GTV were compared against each of the manual GTV reference. Inter-observer variation was also assessed using the 5 manual GTV references.

RESULTS

The proposed anatomic-guided lung GTV segmentation model achieved a mean Dice score of 82.4% and 95% Hausdorff distance (HD95) of 6.9mm when averaged cross 20 patients and 5 GTV references (Table 1), which outperformed the basic deep GTV segmentation model by markedly reducing 19.4% HD95 error. The performance of proposed model was also comparable to the inter-observer variation (Dice score: 82.4% vs. 81.9%, HD95 6.9 vs. 6.4mm), indicating that our model had similar reproducibility as human observers.

CONCLUSION

We developed and tested an anatomy-guided deep learning model for segmenting GTV in NSCLC patients. The model achieves high quantitative segmentation performance, which is comparable to the human observer variation. It can be potentially used in radiotherapy practice to improve GTV delineation consistency and reduce workloads of radiation oncologists.

[Impact of different diagnostic criteria for assessing mild micro-hepatic encephalopathy in liver cirrhosis: an analysis based on a prospective, multicenter, real-world study]

Objective: To compare the differences in the prevalence of mild micro-hepatic encephalopathy (MHE) among patients with cirrhosis by using the psychometric hepatic encephalopathy score (PHES) and the Stroop smartphone application (Encephal App) test. Methods: This prospective, multi-center, real-world study was initiated by the National Clinical Medical Research Center for Infectious Diseases and the Portal Hypertension Alliance and registered with International ClinicalTrials.gov (NCT05140837). 354 cases of cirrhosis were enrolled in 19 hospitals across the country. PHES (including digital connection tests A and B, digital symbol tests, trajectory drawing tests, and serial management tests) and the Stroop test were conducted in all of them. PHES was differentiated using standard diagnostic criteria established by the two studies in China and South Korea. The Stroop test was evaluated based on the criteria of the research and development team. The impact of different diagnostic standards or methods on the incidence of MHE in patients with cirrhosis was analyzed. Data between groups were differentiated using the t-test, Mann-Whitney U test, and χ (2) test. A kappa test was used to compare the consistency between groups. Results: After PHES, the prevalence of MHE among 354 cases of cirrhosis was 78.53% and 15.25%, respectively, based on Chinese research standards and Korean research normal value standards. However, the prevalence of MHE was 56.78% based on the Stroop test, and the differences in pairwise comparisons among the three groups were statistically significant (kappa = -0.064, P < 0.001). Stratified analysis revealed that the MHE prevalence in three groups of patients with Child-Pugh classes A, B, and C was 74.14%, 83.33%, and 88.24%, respectively, according to the normal value standards of Chinese researchers, while the MHE prevalence rates in three groups of patients with Child-Pugh classes A, B, and C were 8.29%, 23.53%, and 38.24%, respectively, according to the normal value standards of Korean researchers. Furthermore, the prevalence rates of MHE in the three groups of patients with Child-Pugh grades A, B, and C were 52.68%, 58.82%, and 73.53%, respectively, according to the Stroop test standard. However, among the results of each diagnostic standard, the prevalence of MHE showed an increasing trend with an increasing Child-Pugh grade. Further comparison demonstrated that the scores obtained by the number connection test A and the number symbol test were consistent according to the normal value standards of the two studies in China and South Korea (Z = -0.982, -1.702; P = 0.326, 0.089), while the other three sub-tests had significant differences (P < 0.001). Conclusion: The prevalence rate of MHE in the cirrhotic population is high, but the prevalence of MHE obtained by using different diagnostic criteria or methods varies greatly. Therefore, in line with the current changes in demographics and disease spectrum, it is necessary to enroll a larger sample size of a healthy population as a control. Moreover, the establishment of more reliable diagnostic scoring criteria will serve as a basis for obtaining accurate MHE incidence and formulating diagnosis and treatment strategies in cirrhotic populations.

Correlation between capsule endoscopy classification and CT lymphangiography of primary intestinal lymphangiectasia

AIM

To investigate the correlation between capsule endoscopy (CE) classification of primary intestinal lymphangiectasia (PIL) and computed tomography (CT) lymphangiography (CTL).

MATERIALS AND METHODS

A total of 52 patients with diagnosed PIL were enrolled. All patients were examined using CTL and small intestinal CE before surgery. CE assessments included the morphology, scope, colour, and size of lesions. CTL assessments included intestinal wall, lymphatic vessel dilatation, lymph fluid reflux, and lymphatic fistula. Patients were divided into three groups according to type diagnosed by CE, and the CTL characteristics were analysed among the groups.

RESULTS

CE showed 15 patients with type I, 27 with II, and 10 with type III. Intestinal wall thickening was observed in 15 type I, 21 type II, and seven type III. Pericardial effusion was observed in only three type I patients; the difference among types was statistically significant (p=0.02). Abnormal contrast agent distribution in the intestinal wall and mesentery was observed in 15 type II patients, and the difference was significantly greater than that of types I and III (p=0.02). Abnormal contrast agent distribution in the abdominal cavity was observed in 12 type II, and the difference was statistically significant (p=0.03).

CONCLUSION

The CE PIL classification reflects the extent and scope of intestinal mucosa lesions; CTL more systematically demonstrates abnormal lymphatic vessels or reflux, and its manifestations of PIL are related to the CE classification. The combination of CTL with CE is useful for accurately evaluating PIL, and provides guidance for preoperative assessment and treatment management of PIL patients.

Correlation between glenoid bone structure and recurrent anterior dislocation of the shoulder joint

BACKGROUND

The aim of the study was to investigate the anatomical characteristics and symmetry of the bilateral glenoid structures of Chinese people and to explore the relationship between the glenoid bone structure and recurrent anterior dislocation.

MATERIALS AND METHODS

The control group included 131 individuals with no history of shoulder dislocation. The dislocation group consisted of 131 patients with a history of unilateral shoulder dislocation. All subjects underwent computed tomography scans. Glenoid shape (pear-shaped, inverted comma-shaped, oval-shaped), width, height, depth, version angle, area, maximum fitting circle area and volume were measured.

RESULTS

There was no significant difference in normal bilateral glenoid of Chinese people (p > 0.05). There were statistically significant differences in depth, height to width ratio, maximum fitting circle area and shape between the dislocation and control groups (p < 0.05). Regression analyses showed that the glenoid depth (odds ratio [OR] 0.48; p < 0.01), the glenoid height to width ratio (OR 28.61; p < 0.01), the glenoid maximum fitting circle area (OR 1.01; p < 0.01) and the glenoid shape (p <0.05; pear-shaped OR 0.432; inverted comma-shaped OR 0.954) were associated with anterior shoulder instability. Pear-shaped and inverted comma-shaped glenoid had lower risk of recurrent anterior shoulder dislocation compared to oval glenoid. Receiver operating characteristic curve analysis showed that individuals with anterior shoulder instability had smaller glenoid depth and larger height to width ratio and the glenoid maximum fitting circle area compared with the control group.

CONCLUSIONS

The normal bilateral glenoids of Chinese people are basically symmetrical. The glenoid shape, depth, height to width ratio and maximum fitting circle area are risk factors for recurrent anterior shoulder dislocation. Evaluation of the glenoid bone structure enables more accurate prediction of the risk of recurrent shoulder dislocation.

A functional genomics screen identifying blood cell development genes in Drosophila by undergraduates participating in a course-based research experience

Undergraduate students participating in the UCLA Undergraduate Research Consortium for Functional Genomics (URCFG) have conducted a two-phased screen using RNA interference (RNAi) in combination with fluorescent reporter proteins to identify genes important for hematopoiesis in Drosophila. This screen disrupted the function of approximately 3500 genes and identified 137 candidate genes for which loss of function leads to observable changes in the hematopoietic development. Targeting RNAi to maturing, progenitor, and regulatory cell types identified key subsets that either limit or promote blood cell maturation. Bioinformatic analysis reveals gene enrichment in several previously uncharacterized areas, including RNA processing and export and vesicular trafficking. Lastly, the participation of students in this course-based undergraduate research experience (CURE) correlated with increased learning gains across several areas, as well as increased STEM retention, indicating that authentic, student-driven research in the form of a CURE represents an impactful and enriching pedagogical approach.

Curcumin decreases dibutyl phthalate-induced renal dysfunction in Kunming mice via inhibiting oxidative stress and apoptosis

Curcumin (Cur) has been used extensively in dietary supplement with antioxidant and anti-apoptotic properties. Although dibutyl phthalate (DBP) has adverse effects on the kidney, any association between DBP exposure and the role of Cur is unclear. We tested the hypothesis that exposure to DBP has adverse consequences on renal dysfunction in mice and the potential protective role of Cur in decreasing DBP-induced renal dysfunction via inhibiting oxidative stress and apoptosis. Kidney function, oxidative stress biomarkers, and apoptosis factors as well as Bcl-2 and Bax were investigated. The results showed a marked increase of renal dysfunction, oxidative stress and apoptosis level after DBP exposure compared to the control. While administration of Cur to DBP-treated mice may reduce these adverse biochemical changes compared with DBP-alone group. Overall, these results suggest that oxidative stress and apoptosis are involved in DBP-induced renal disorder, whereas Cur plays a protective role in inhibiting these two pathways.

Deep medical image analysis with representation learning and neuromorphic computing

Deep learning is increasingly used in medical imaging, improving many steps of the processing chain, from acquisition to segmentation and anomaly detection to outcome prediction. Yet significant challenges remain: (i) image-based diagnosis depends on the spatial relationships between local patterns, something convolution and pooling often do not capture adequately; (ii) data augmentation, the de facto method for learning three-dimensional pose invariance, requires exponentially many points to achieve robust improvement; (iii) labelled medical images are much less abundant than unlabelled ones, especially for heterogeneous pathological cases; and (iv) scanning technologies such as magnetic resonance imaging can be slow and costly, generally without online learning abilities to focus on regions of clinical interest. To address these challenges, novel algorithmic and hardware approaches are needed for deep learning to reach its full potential in medical imaging.

[Trends analysis of vitamin D status among pregnant women in Hefei during 2015-2017]

From March 2015 to February 2018, 4 728 women aged 18 to 45 years old with single-pregnancy at the gestational age of 13 to 27 weeks in Hefei were recruited to analyze the trend of vitamin D status. The average levels of serum 25(OH)D in 2015, 2016 and 2017 were (43.22±18.41) nmol/L, (39.3±15.1) nmol/L and (36.6±17.0) nmol/L, and the prevalence of vitamin D deficiency were 69.5%, 77.6% and 81.4%, respectively. Compared with 2015, the levels of serum 25(OH)D in pregnant women in 2016 and 2017 decreased by 5.23 (95%CI: 4.10-6.35) nmol/L and 7.98 (95%CI: 6.77-9.19) nmol/L. The OR (95%CI) values for the risk of vitamin D deficiency were 1.88 (95%CI: 1.57-2.24) and 2.41 (95%CI: 1.98-2.93).

[Phosphatase and tensin homolog deleted on chromosome ten deficiency sensitizes tumor cells to lithium chloride treatment]

Objective: To identify the therapeutic efficacy of lithium chloride (LiCl) on phosphatase and tensin homolog deleted on chromosome ten (PTEN)-deficient tumors. Methods: First, the Catalogue of Somatic Mutations in Cancer for mutation spectrum of human endometrial carcinoma samples was analyzed. Second, the relationship between PTEN abundance and LiCl inhibition of endometrial cancer cell lines using Pten(+/+) and Pten(-/-) mouse embryonic fibroblast (MEF) lines was investigated. Moreover, potential alterations of mammalian target of rapamycin (mTOR) signaling pathway after treatment with LiCl were checked.Last,LiCl's efficacy on PTEN null tumors was studied. Results: PTEN was mutated in 39% of endometrial carcinomas. LiCl preferentially inhibited the proliferation of PTEN-deficient endometrial carcinoma cells and MEFs. Furthermore, LiCl blocked PTEN-deficient tumor development. Mechanistically, LiCl down-regulated mTOR signaling. Conclusions: PTEN is the most frequently mutated gene in endometrial carcinoma.By targeting mTOR signaling pathway,LiCl is a promising regimen for the treatment of tumors with PTEN deficiency.

[Prospective cohort study on association between peri-conceptional air pollution exposure and gestational diabetes mellitus]

Objective: To explore the association between the exposure to major air pollutants in pre-pregnancy and early pregnancy (peri-conceptional period) and gestational diabetes mellitus (GDM). Methods: From March 2015 to April 2018, 4 817 pregnancies were recruited at three prenatal check-ups hospital in Hefei (Hefei First People's Hospital, Hefei. Maternal and Child Care Hospital and the First Affiliated Hospital of Anhui Medical University), China. Questionnaire was used to collect the demographic data, the health status and lifestyle of pregnant women. GDM was diagnosed according to the Chinese Guidelines for the Prevention and Treatment of Type 2 Diabetes (2017 Edition). Logistic regression was used to investigate the association of exposure to major air pollutants (PM(2.5), PM(10), SO(2), CO and NO(2)) during different periods of pre-pregnancy (12 weeks before pregnancy) and first trimester (12 weeks after last menstruation) and duration of exposure to high levels of pollutants with GDM. Results: The mean±SD of the age of subjects was (29.14±4.19) years old and the prevalence of GDM was 21.4% (n=1 030). The results of multivariate logistic regression analysis showed that after adjusting for confounding factors, the risk of GDM increased gradually with the prolonged exposure time of high-concentration pollutants compared with pregnant women who were not exposed to high pollution during the pre-pregnancy (χ(2)=61.28, P(trend)<0.001) with the OR (95%CI) values for exposure time of 1, 2, and 3 months about 1.42 (1.10-1.84), 1.73 (1.29-2.33), and 2.51 (1.75-3.59), respectively. In the pre-pregnancy period, in every 10 μg/m(3) increase of PM(2.5) and PM(10), the OR (95%CI) values of GDM were 1.14 (1.08-1.20) and 1.13 (1.08-1.19), respectively; for each increase of 1 μg/m(3) and 0.10 mg/m(3) of SO(2) and CO, the OR (95% CI) values of GDM were 1.03 (1.01-1.05) and 1.07 (1.01-1.13), respectively. For every 1 μg/m(3) increase in the average concentration of SO(2) in the first trimester, the OR (95%CI) value of GDM was 1.02 (1.01-1.05). Conclusion: PM(2.5), PM(10), SO(2) and CO exposure during the pre-pregnancy and SO(2) exposure in first trimester were positively correlated with the risk of GDM.

Tumor mutational burden is predictive of response to immune checkpoint inhibitors in MSI-high metastatic colorectal cancer

BACKGROUND

Microsatellite instability (MSI) is a biomarker for response to immune checkpoint inhibitors (ICPIs). PD-1 inhibitors in metastatic colorectal carcinoma (mCRC) with MSI-high (MSI-H) have demonstrated a high disease control rate and favorable progression-free survival (PFS); however, reported response rates to pembrolizumab and nivolumab are variable and often <50%, suggesting that additional predictive biomarkers are needed.

METHODS

Clinicopathologic data were collected from patients with MSI-H mCRC confirmed by hybrid capture-based next-generation sequencing (NGS) treated with PD-1/L1 inhibitors at five institutes. Tumor mutational burden (TMB) was determined on 0.8-1.1 Mb of sequenced DNA and reported as mutations/Mb. Potential biomarkers of response and time to progression were analyzed by univariate and multivariate analyses. Once TMB was confirmed as a predictive biomarker, a larger dataset of 18 140 unique CRC patients was analyzed to define the relevance of the identified TMB cut-point.

RESULTS

A total of 22 patients were treated with PD-1/L1 inhibitors including 19 with pembrolizumab monotherapy. Among tested variables, TMB showed the strongest association with objective response (OR; P < 0.001) and PFS, by univariate (P < 0.001) and multivariate analysis (P < 0.01). Using log-rank statistics, the optimal predictive cut-point for TMB was estimated between 37 and 41 mutations/Mb. All 13 TMBhigh cases responded, while 6/9 TMBlow cases had progressive disease. The median PFS for TMBhigh has not been reached (median follow-up >18 months) while the median PFS for TMBlow was 2 months. A TMB of 37.4 mutations/Mb in a large MSI-H mCRC population (821/18, 140 cases; 4.5%) evaluated by NGS corresponded to the 35th percentile cut-point.

CONCLUSIONS

TMB appears to be an important independent biomarker within MSI-H mCRC to stratify patients for likelihood of response to ICPIs. If validated in prospective studies, TMB may play an important role in guiding the sequencing and/or combinations of ICPIs in MSI-H mCRC.

A meta-analysis of external fixation and flexible intramedullary nails for femoral fractures in children

BACKGROUND

The purpose of this meta-analysis was to compare the outcomes of external fixation and flexible intramedullary nails for femoral fractures in children between 5 and 15 years of age based on the current evidence.

MATERIALS AND METHODS

We searched relevent studies in the following database: Cochrane library, PubMed and EMABASE up to May 2014. All randomized controlled trials, Clinical controlled trials and retrospective controlled studies comparing external fixation and flexible intramedullary nails in femoral fractures of children were included. Data was extracted independently for meta-analysis.

RESULTS

Seven trials altogether involving 338 cases of femoral fractures of children treated by external fixation (128 cases) and flexible intramedullary nails (210 cases) were included in the meta-analysis. Results showed that flexible intramedullary nails was superior to external fixation in less time to union , lower postoperative infection rate and refracture rate . It may not increase delayed union, Limb-length discrepancy , pain and bursitis . Both fixations obtained a similar patient satisfaction.

CONCLUSION

Flexible intramedullary nail had greater advantages for the treatment of femoral fractures in children aged 5-15 years, compared to external fixation based on current meta-analysis. This conclusion will ultimately require rigorous and adequately powered randomized controlled trials to be proved.

Lipopolysaccharide derived from the digestive tract provokes oxidative stress in the liver of dairy cows fed a high-grain diet

The aims of this study were to measure oxidative stress parameters and to investigate the molecular mechanism triggered by grain-induced subacute ruminal acidosis in mid-lactation cows. Twelve Holstein-Friesian cows with an average weight of 455±28kg were divided into 2 groups and subjected to 2 diets over 18wk: either a low-grain (forage-to-concentrate ratio=6:4) or a high-grain (forage-to-concentrate ratio=4:6) diet based on dry matter. Being fed a long-term high-grain diet resulted in a significant decrease in rumen pH and a significant increase in ruminal lipopolysaccharide (LPS) at 4 h postfeeding in the morning. The increase was also observed in LPS concentrations in the portal vein, hepatic vein, and jugular vein blood plasma as well as reduced milk yield in a high-grain diet. Cows fed a high-grain diet had lower levels of catalase and glutathione peroxidase (GPx) activity and total antioxidant capacity than cows fed a low-grain diet; however, super oxide dismutase (SOD) activity and malondialdehyde (MDA) levels were higher in both the liver and the plasma of high-grain than in low-grain cows. Positive correlations were observed between plasma LPS versus hepatic MDA, plasma MDA, and hepatic SOD activity, whereas hepatic GPx and plasma GPx were negatively correlated with plasma LPS. The relative mRNA abundances of GPX1 and CAT were significantly lower in the liver of cows fed a high-grain diet than those fed a low-grain diet, whereas SOD1 was significantly higher in cows fed a high-grain diet than cows fed a low-grain diet. The expression levels of Nrf2, NQO1, MT1E, UGT1A1, MGST3, and MT1A were downregulated, whereas NF-kB was upregulated, in cows fed a high-grain diet. Furthermore, nuclear factor E2-related factor 2 (Nrf2) total protein and mRNA levels were significantly lower than in low-grains. Our results demonstrate the relationship between the translocated LPS and the suppression of cellular antioxidant defense capacity, which lead to increased oxidative stress and suggests that the Nrf2-dependent antioxidant response may be affected by higher levels of LPS translocated to the bloodstream.