Radiomics-guided deep neural networks stratify lung adenocarcinoma prognosis from CT scans

AbstractDeep learning (DL) is a breakthrough technology for medical imaging with high sample size requirements and interpretability issues. Using a pretrained DL model through a radiomics-guided approach, we propose a methodology for stratifying the prognosis of lung adenocarcinomas based on pretreatment CT. Our approach allows us to apply DL with smaller sample size requirements and enhanced interpretability. Baseline radiomics and DL models for the prognosis of lung adenocarcinomas were developed and tested using local (n = 617) cohort. The DL models were further tested in an external validation (n = 70) cohort. The local cohort was divided into training and test cohorts. A radiomics risk score (RRS) was developed using Cox-LASSO. Three pretrained DL networks derived from natural images were used to extract the DL features. The features were further guided using radiomics by retaining those DL features whose correlations with the radiomics features were high and Bonferroni-corrected p-values were low. The retained DL features were subject to a Cox-LASSO when constructing DL risk scores (DRS). The risk groups stratified by the RRS and DRS showed a significant difference in training, testing, and validation cohorts. The DL features were interpreted using existing radiomics features, and the texture features explained the DL features well.

Download Full-text

Association of radiomic features from prostate bi-parametric MRI with Decipher risk categories to predict risk for biochemical recurrence post-prostatectomy.

Journal of Clinical Oncology ◽

10.1200/jco.2019.37.15_suppl.e16561 ◽

2019 ◽

Vol 37 (15_suppl) ◽

pp. e16561-e16561 ◽

Cited By ~ 1

Author(s):

Lin Li ◽

Rakesh Shiradkar ◽

Patrick Leo ◽

Andrei Purysko ◽

Ahmad Algohary ◽

...

Keyword(s):

Biochemical Recurrence ◽

Characteristic Curve ◽

Texture Features ◽

Risk Groups ◽

Multivariate Logistic Regression Model ◽

Gleason Grade ◽

Grade Group ◽

Free Survival ◽

Significant Difference ◽

Model C

e16561 Background: Decipher test examines expression levels of 22 metastasis-related RNA sequences and provides a risk score (DS) for assessing metastasis risk within 5 years after radical prostatectomy (RP). Our previous work has shown that radiomics or computer extracted texture features are associated with DS on multiparametric MRI (mpMRI). DS associated radiomics (RDS) may capture signatures related to poor prognosis of prostate cancer (PCa) following RP. Thus, we sought to evaluate whether RDS were also able to predict the risk of biochemical recurrence (BCR), a surrogate clinical endpoint of metastasis, after RP. Methods: Two patient cohorts from two institutions who underwent 3T mpMRI prior to RP were used in this study - I1: 70 PCa patients who had a Decipher test following RP; I2: 120 PCa patients with at least 5-year follow-up after RP. I1 was investigated as a training set to identify RDS and develop a multivariate logistic regression model (CL). I2 was used as a testing set to validate CL for predicting BCR. A total of 150 radiomics were extracted within each lesion from T2-weighted MRI (T2WI) and apparent diffusion coefficient (ADC) maps. Area under the receiver operating characteristic curve (AUC) and Kaplan-Meier (KM) analysis were used to evaluate performance of CL. Results: Radiomics quantifying lesion spot and ripple texture patterns on ADC maps and T2WI intensity gradient changes were able to differentiate Decipher high and low/intermediate groups in I1 (AUC = 0.92). The same model CL was then applied to I2 resulting in a significant difference in BCR-free survival time between the predicted risk groups (p < 0.05, hazard ratio (HR) = 2.54) independent of Gleason Grade Group (GGG, p < 0.05, HR = 7.61) and PIRADS-v2 (p < 0.05, HR = 5.64) in multivariate testing. Especially, for patients with GGG > 2 (n = 57), CL predicted BCR risk groups shown HR = 3.1 (p < 0.05) while GGG alone was not prognostic (p > 0.05, HR = 1.93). Conclusions: Our radiomic model was able to predict DS and was found to also be prognostic of BCR-free survival. Further work will involve evaluating whether these radiomic features are predictive of PCa metastasis. [Table: see text]

Download Full-text

Prognostic Ferroptosis-Related lncRNA Signatures Associated With Immunotherapy and Chemotherapy Responses in Patients With Stomach Cancer

Frontiers in Genetics ◽

10.3389/fgene.2021.798612 ◽

2022 ◽

Vol 12 ◽

Author(s):

Donlin Lai ◽

Lin Tan ◽

Xiaojia Zuo ◽

DingSheng Liu ◽

Deyi Jiao ◽

...

Keyword(s):

Stomach Cancer ◽

Risk Group ◽

Survival Rates ◽

Clinical Information ◽

Predictive Performance ◽

Risk Groups ◽

Low Risk ◽

Risk Scores ◽

Immune Checkpoints ◽

Significant Difference

Ferroptosis is associated with the prognosis and therapeutic responses of patients with various cancers. LncRNAs are reported to exhibit antitumor or oncogenic functions. Currently, few studies have assessed the combined effects of ferroptosis and lncRNAs on the prognosis and therapy of stomach cancer. In this study, transcriptomic and clinical data were downloaded from TCGA database, and ferroptosis-related genes were obtained from the FerrDb database. Through correlation analysis, Cox analysis, and the Lasso algorithm, 10 prognostic ferroptosis-related lncRNAs (AC009299.2, AC012020.1, AC092723.2, AC093642.1, AC243829.4, AL121748.1, FLNB-AS1, LINC01614, LINC02485, LINC02728) were screened to construct a prognostic model, which was verified in two test cohorts. Risk scores for patients with stomach cancer were calculated, and patients were divided into two risk groups. The low-risk group, based on the median value, had a longer overall survival time in the KM curve, and a lower proportion of dead patients in the survival distribution curve. Potential mechanisms and possible functions were revealed using GSEA and the ceRNA network. By integrating clinical information, the association between lncRNAs and clinical features was analyzed and several features affecting prognosis were identified. Then, a nomogram was developed to predict survival rates, and its good predictive performance was indicated by a relatively high C-index (0.67118161) and a good match in calibration curves. Next, the association between these lncRNAs and therapy was explored. Patients in the low-risk group had an immune-activating environment, higher immune scores, higher TMB, lower TIDE scores, and higher expression of immune checkpoints, suggesting they might receive a greater benefit from immune checkpoint inhibitor therapy. In addition, a significant difference in the sensitivity to mitomycin. C, cisplatin, and docetaxel, but not etoposide and paclitaxel, was observed. In summary, this model had guiding significance for prognosis and personalized therapy. It helped screen patients with stomach cancer who might benefit from immunotherapy and guided the selection of personalized chemotherapeutic drugs.

Download Full-text

Prognostic Signature of Hepatocellular Carcinoma and Analysis of Immune Infiltration Based on m6A-Related lncRNAs

Frontiers in Oncology ◽

10.3389/fonc.2021.691372 ◽

2021 ◽

Vol 11 ◽

Author(s):

Ting Guo ◽

Kun He ◽

Yifei Wang ◽

Jingjing Sun ◽

Yong Chen ◽

...

Keyword(s):

Hepatocellular Carcinoma ◽

Univariate Analysis ◽

Risk Groups ◽

Risk Difference ◽

Risk Scores ◽

Lasso Regression ◽

Prognostic Signature ◽

Significant Difference ◽

Infiltrating Cells ◽

The Relationship

The relationship between m6A-related lncRNAs and prognosis in hepatocellular carcinoma (HCC) is not yet clear. We used Lasso regression to establish a prognostic signature based on m6A-related lncRNAs using a training set from TCGA, and then verified the signature efficacy in a test set. Fluorescence quantitative real-time PCR (qPCR), Survival analysis, clinical risk difference analysis, immune-related analysis, and drug-sensitivity analysis were conducted. The results revealed that 1,651 lncRNAs were differentially expressed in HCC tissues, among which, 163 were m6A-related. Univariate analysis showed that 87 lncRNAs were associated with the overall survival. Six differential m6A-related lncRNAs were validated and selected via Lasso regression to construct a prognostic signature which demonstrated a satisfactory predictive efficacy. In the clinically relevant pathologic stage, histologic grade, and T stage, the risk scores obtained based on this signature showed a statistically significant difference. The high- and low-risk groups exhibited a difference in the tumor immune infiltrating cells, immune checkpoint gene expression, and sensitivity to chemotherapy. In summary, the prognostic signature based on the m6A-related lncRNAs can effectively predict the prognosis of patients and might provide a new vista for the chemotherapy and immunotherapy of HCC.

Download Full-text

The Trade-Off of Post-Mastectomy Radiotherapy Usage for the Breast Cancer Patients Aged 70 Years or Older: A Study Based on SEER Database

10.21203/rs.3.rs-966806/v1 ◽

2021 ◽

Author(s):

Jingyi Lin ◽

Shiping Luo ◽

Jie Zhang ◽

Chuangui Song

Keyword(s):

Breast Cancer ◽

Elderly Patients ◽

Cox Regression ◽

Predictive Accuracy ◽

Risk Groups ◽

Low Risk ◽

Risk Scores ◽

Seer Database ◽

Significant Difference ◽

Her 2

Abstract Introduction: This study aimed to investigate the role of post-mastectomy radiotherapy (PMRT) in the female aged 70 years or older diagnosed with breast cancer, which is still controversial.Methods: This retrospective study enrolled women aged 70+ years diagnosed with breast cancer between 2004 and 2016 following mastectomy from the Surveillance, Epidemiology, and End Results (SEER) database. Propensity score matching (PSM) was performed to reduce covariable imbalance. A nomogram was created to predict the 1,3,5-years overall survival (OS) of elderly patients based on the the univariable and multivariable COX regression model results. The X-tile software determined the optimal cutoff value of risk scores from the nomogram and divided patients into three risk groups. Results: Of the 27,636 females were eligible, 17.2% (n=4,747) received PMRT while 82,8% (n=22889) not. After 1:1 matching, PMRT were associated with significant improvement in breast cause-specific survival (BCSS) and OS (p <0.001). By contrast, the BCSS and OS benefit from PMRT were not significant in patients with T1N1 tumor (BCSS：HR = 0.716，p = 0.249；OS：HR = 0.908，p = 0.572), and T2N1 tumor (BCSS：HR = 0.866，p = 0.289；OS：HR = 0.879，p = 0.166). Stratified by subtype, the HR+/HER-2- subtype and the HR-/HER-2- subtype (all p<0.001) have a significant prolonged survival, yet not significant difference are shown in the HER-2+ tumor. The nomogram has high predictive accuracy and discrimination, and well distinguish three risk groups. In the low-risk group, PMRT didn't significantly better OS (p=0.203).Conclusions: This study demonstrated that post-mastectomy radiotherapy improves the survival of females with elderly breast cancer. After a comprehensive assessment of the side effects and the quality of life, the omission of PMRT could be considered in patients with T1-2N1 breast cancer. Furthermore, the nomogram we constructed could be used as a decision tool for the omission of PMRT in low-risk elderly patients.

Download Full-text

Regional Cervical Plexus Blockage for Carotid Endarterectomy in Patients with Cardiovascular Risk Factors

The Heart Surgery Forum ◽

10.1532/hsf.1385 ◽

2015 ◽

Vol 18 (4) ◽

pp. 140 ◽

Cited By ~ 2

Author(s):

Mehmet Taşar ◽

Mehmet Kalender ◽

Okay Güven Karaca ◽

Ata Niyazi Ecevit ◽

Salih Salihi ◽

...

Keyword(s):

Risk Factors ◽

Cardiovascular Risk ◽

Carotid Endarterectomy ◽

Regional Anesthesia ◽

Carotid Artery Disease ◽

Risk Groups ◽

Cardiac Risk ◽

Cervical Plexus ◽

Cardiac Risk Factors ◽

Significant Difference

Background: Carotid artery disease is not rare in cardiac patients. Patients with cardiac risk factors and carotid stenosis are prone to neurological and cardiovascular complications. With cardiac risk factors, carotid endarterectomy operation becomes challenging. Regional anesthesia is an alternative option, so we aimed to investigate the operative results of carotid endarterectomy operations under regional anesthesia in patients with cardiac risk factors. Methods: We aimed to analyze and compare outcomes of carotid endarterectomy under regional anesthesia with cardiovascular risk groups retrospectively. Between 2006 and 2014, we applied 129 carotid endarterectomy ± patch plasty to 126 patients under combined cervical plexus block anesthesia. Patients were divided into three groups (high, moderate, low) according to their cardiovascular risks. Neurological and cardiovascular events after carotid endarterectomy were compared. Results: Cerebrovascular accident was seen in 7 patients (5.55%) but there was no significant difference between groups (P > .05). Mortality rate was 4.76% (n = 6); it was higher in the high risk group and was not statistically significant (P = .180). Four patients required revision for bleeding (3.17%). We did not observe any postoperative surgical infection. Conclusion: Carotid endarterectomy can be safely performed with regional cervical anesthesia in all cardiovascular risk groups. Comprehensive studies comparing general anesthesia and regional anesthesia are needed.

Download Full-text

Genetic co-expression networks contribute to creating predictive model and exploring novel biomarkers for the prognosis of breast cancer

Scientific Reports ◽

10.1038/s41598-021-84995-z ◽

2021 ◽

Vol 11 (1) ◽

Author(s):

Chi-Ming Chu ◽

Huan-Ming Hsu ◽

Chi-Wen Chang ◽

Yuan-Kuei Li ◽

Yu-Jia Chang ◽

...

Keyword(s):

Breast Cancer ◽

Prediction Models ◽

Reference Model ◽

External Validation ◽

Risk Scores ◽

Consistent Finding ◽

Hazard Ratios ◽

Genetic Predictors ◽

Novel Biomarkers ◽

Cox Proportional Hazard Regression

AbstractGenetic co-expression network (GCN) analysis augments the understanding of breast cancer (BC). We aimed to propose GCN-based modeling for BC relapse-free survival (RFS) prediction and to discover novel biomarkers. We used GCN and Cox proportional hazard regression to create various prediction models using mRNA microarray of 920 tumors and conduct external validation using independent data of 1056 tumors. GCNs of 34 identified candidate genes were plotted in various sizes. Compared to the reference model, the genetic predictors selected from bigger GCNs composed better prediction models. The prediction accuracy and AUC of 3 ~ 15-year RFS are 71.0–81.4% and 74.6–78% respectively (rfm, ACC 63.2–65.5%, AUC 61.9–74.9%). The hazard ratios of risk scores of developing relapse ranged from 1.89 ~ 3.32 (p < 10–8) over all models under the control of the node status. External validation showed the consistent finding. We found top 12 co-expressed genes are relative new or novel biomarkers that have not been explored in BC prognosis or other cancers until this decade. GCN-based modeling creates better prediction models and facilitates novel genes exploration on BC prognosis.

Download Full-text

A TTP-incorporated scoring model for predicting mortality of solid tumor patients with bloodstream infection caused by Escherichia coli

Supportive Care in Cancer ◽

10.1007/s00520-021-06442-z ◽

2021 ◽

Author(s):

Qing Zhang ◽

Hao-Yang Gao ◽

Ding Li ◽

Chang-Sen Bai ◽

Zheng Li ◽

...

Keyword(s):

Escherichia Coli ◽

High Risk ◽

Bloodstream Infection ◽

Solid Tumor ◽

Prognostic Scores ◽

Risk Scores ◽

Youden Index ◽

Tumor Patients ◽

Scoring Model ◽

Significant Difference

Abstract Background Few mortality-scoring models are available for solid tumor patients who are predisposed to develop Escherichia coli–caused bloodstream infection (ECBSI). We aimed to develop a mortality-scoring model by using information from blood culture time to positivity (TTP) and other clinical variables. Methods A cohort of solid tumor patients who were admitted to hospital with ECBSI and received empirical antimicrobial therapy was enrolled. Survivors and non-survivors were compared to identify the risk factors of in-hospital mortality. Univariable and multivariable regression analyses were adopted to identify the mortality-associated predictors. Risk scores were assigned by weighting the regression coefficients with corresponding natural logarithm of the odds ratio for each predictor. Results Solid tumor patients with ECBSI were distributed in the development and validation groups, respectively. Six mortality-associated predictors were identified and included in the scoring model: acute respiratory distress (ARDS), TTP ≤ 8 h, inappropriate antibiotic therapy, blood transfusion, fever ≥ 39 °C, and metastasis. Prognostic scores were categorized into three groups that predicted mortality: low risk (< 10% mortality, 0–1 points), medium risk (10–20% mortality, 2 points), and high risk (> 20% mortality, ≥ 3 points). The TTP-incorporated scoring model showed excellent discrimination and calibration for both groups, with AUC being 0.833 vs 0.844, respectively, and no significant difference in the Hosmer–Lemeshow test (6.709, P = 0.48) and the chi-square test (6.993, P = 0.46). Youden index showed the best cutoff value of ≥ 3 with 76.11% sensitivity and 79.29% specificity. TTP-incorporated scoring model had higher AUC than no TTP-incorporated model (0.837 vs 0.817, P < 0.01). Conclusions Our TTP-incorporated scoring model was associated with improving capability in predicting ECBSI-related mortality. It can be a practical tool for clinicians to identify and manage bacteremic solid tumor patients with high risk of mortality.

Download Full-text

Minimum sample size for external validation of a clinical prediction model with a binary outcome

Statistics in Medicine ◽

10.1002/sim.9025 ◽

2021 ◽

Author(s):

Richard D. Riley ◽

Thomas P. A. Debray ◽

Gary S. Collins ◽

Lucinda Archer ◽

Joie Ensor ◽

...

Keyword(s):

Prediction Model ◽

Sample Size ◽

External Validation ◽

Minimum Sample Size ◽

Binary Outcome ◽

Clinical Prediction ◽

Clinical Prediction Model ◽

Minimum Sample

Download Full-text

Determination of Sample Size and Influencing Factors to Characterize Faulting in Jointed Concrete Pavements

Transportation Research Record Journal of the Transportation Research Board ◽

10.1177/0361198121996704 ◽

2021 ◽

pp. 036119812199670

Author(s):

Ruohan Li ◽

Jorge A. Prozzi

Keyword(s):

Sample Size ◽

Plain Concrete ◽

Concrete Pavements ◽

Portland Cement Concrete ◽

Federal Highway ◽

Significant Difference ◽

Jointed Concrete Pavement ◽

Functional Classes ◽

High Precipitation ◽

Departments Of Transportation

The objective of this study is to evaluate the field variability of jointed concrete pavement (JCP) faulting and its effects on pavement performance. The standard deviation of faulting along both the longitudinal and transverse directions are calculated. Based on these, the overall variability is determined, and the required sample sizes needed for a given precision at a certain confidence level are calculated and presented. This calculation is very important as state departments of transportation are required to report faulting every 0.1 mi to the Federal Highway Administration as required by the 2015 FAST Act. On average, twice the number of measurements are needed on jointed reinforced concrete pavements (JRCP) to achieve the same confidence and precision as on jointed plain concrete pavements (JPCP). For example, a sample size of 13 is needed to achieve a 95% confidence interval with a precision of 1.0 mm for average faulting of JPCP, while 26 measurements are required for JRCP ones. Average faulting was found to correlate with several climatic, structural, and traffic variables, while no significant difference was found between edge and outer wheelpath measurements. The application of Portland cement concrete overlay and the use of dowel bars (rather than aggregate interlock) are found to significantly reduce faulting. Older sections located on higher functional classes, and in regions of high precipitation or where the daily temperature change is larger, tend to have higher faulting, and might require larger samples sizes as compared with the rest when faulting surveys are to be conducted.

Download Full-text

Modification and Validation of the Phosphate Removal Model: A Multicenter Study

Kidney & Blood Pressure Research ◽

10.1159/000511375 ◽

2021 ◽

pp. 1-10

Author(s):

Weichen Zhang ◽

Qiuna Du ◽

Jing Xiao ◽

Zhaori Bi ◽

Chen Yu ◽

...

Keyword(s):

Polynomial Regression ◽

External Validation ◽

Original Model ◽

Phosphate Concentration ◽

Phosphate Removal ◽

Internal Validation ◽

Modified Model ◽

New Model ◽

Significant Difference ◽

Removal Model

Background: Our research group has previously reported a noninvasive model that estimates phosphate removal within a 4-h hemodialysis (HD) treatment. The aim of this study was to modify the original model and validate the accuracy of the new model of phosphate removal for HD and hemodiafiltration (HDF) treatment. Methods: A total of 109 HD patients from 3 HD centers were enrolled. The actual phosphate removal amount was calculated using the area under the dialysate phosphate concentration time curve. Model modification was executed using second-order multivariable polynomial regression analysis to obtain a new parameter for dialyzer phosphate clearance. Bias, precision, and accuracy were measured in the internal and external validation to determine the performance of the modified model. Results: Mean age of the enrolled patients was 63 ± 12 years, and 67 (61.5%) were male. Phosphate removal was 19.06 ± 8.12 mmol and 17.38 ± 6.75 mmol in 4-h HD and HDF treatments, respectively, with no significant difference. The modified phosphate removal model was expressed as Tpo4 = 80.3 × C45 − 0.024 × age + 0.07 × weight + β × clearance − 8.14 (β = 6.231 × 10−3 × clearance − 1.886 × 10−5 × clearance2 – 0.467), where C45 was the phosphate concentration in the spent dialysate measured at the 45th minute of HD and clearance was the phosphate clearance of the dialyzer. Internal validation indicated that the new model was superior to the original model with a significantly smaller bias and higher accuracy. External validation showed that R2, bias, and accuracy were not significantly different than those of internal validation. Conclusions: A new model was generated to quantify phosphate removal by 4-h HD and HDF with a dialyzer surface area of 1.3–1.8 m2. This modified model would contribute to the evaluation of phosphate balance and individualized therapy of hyperphosphatemia.

Download Full-text