MICRORNAS COMBINED TO RADIOMIC FEATURES AS PREDICTOR TO CLINICAL COMPLETE RESPONSE AFTER NEOADJUVANT RADIO-CHEMOTHERAPY FOR LOCALLY ADVANCED RECTAL CANCER
EAES Academy. Losurdo P. 07/05/22; 363020; P064
CLICK HERE TO LOGIN
REGULAR CONTENT
REGULAR CONTENT
Login now to access Regular content available to all registered users.
Abstract
Discussion Forum (0)
Rate & Comment (0)
Background and Aims
Artificial intelligence (AI) associate to radiomic features and microRNAs are promising Locally Advanced Rectal Cancer (LARC) biomarkers. Establishing the association between microRNAs and MRI-based radiomic measures, will help define their significance and potential impact in prediction of complete clinical response (cCR) to neoadjuvant radio-chemotherapy (nRCT).
Study Design
40 patients with LARC where retrospectively analyzed. An MRI of the pelvis before and after nRCT was performed. In the diagnostic biopsy, we measured the expression levels of 7 microRNAs, selected from literature, and correlated their expression levels with the tumor response rate (TRG), assessed on the surgical sample. We compared the accuracy of cCR prediction for: i) clinical predictors (age and stage); ii) radiomic features; iii) microRNAs levels); iv) combination of radiomics and microRNAs.
Statistical analysis
Clinical characteristics were compared between the two groups using t-test on continuous variables and using the Chi-square test for categorical ones. Radiomic data underwent a 3-steps processing: 1) pairwise correlation; 2) univariate association test between features and TRG class; 3) Principal Component Analysis (PCA) was performed on the remaining features and the first component (PC1) was extracted. The association between PC1 and TRG class was investigated with a logistic regression (LR) model.
For micro-RNA data, correlation analysis was conducted using Pearson method. Shapiro-Wilk test was performed for normality of data and Levene’s test for homogeneity of variance. Comparison of means between RESP and NO-RESP groups was performed using the t-test.
Variables of the different domains were used as predictors to estimate LR models and the predicted probabilities were used to calculate the Area Under the ROC curve (AUC-ROC).
Results:
Clinical predictors showed the lowest accuracy. The best performing model was based on the integration of radiomic features with mirR-145 expression level (AUC-ROC=0.90, 95%CI: 0.81-1.0). AI algorithm based on radiomics features, showed association with the TRG-class, and demonstrate a significant impact on the outcome. The overexpression of miR-145, seem to positively correlate with mjor response to nRCT, making a possible reliable biomarker.
Discussion
Currently, there are no biomarkers that accurately predict response to nCRT and thus the decision to begin this therapy is essentially clinical. Recent evidence points to an emerging role of microRNAs as predictive biomarkers for the identification of responder patients.
In addition to biomarkers, recent studies point to the use of radiomics as a powerful tool to predict the severity of the lesions. Radiomic represent quantitative and objective measures; could reflect tumor heterogeneity and sub-regional habitats and could improve diagnostic accuracy.
Conclusions:
Divide patients in non-responders and responders to nRCT could, in fact, spare inefficient and possibly harming treatments and design ad-hoc therapeutic approaches. Poor responders could be leaded directly to upfront surgery, reducing the risk related to nRCT, including the hypothesis of tumor progression. Integration of radiomics features and miR-145 expression, seems to be a positive predictive factor in order to recognize good responders to nRCT.
Artificial intelligence (AI) associate to radiomic features and microRNAs are promising Locally Advanced Rectal Cancer (LARC) biomarkers. Establishing the association between microRNAs and MRI-based radiomic measures, will help define their significance and potential impact in prediction of complete clinical response (cCR) to neoadjuvant radio-chemotherapy (nRCT).
Study Design
40 patients with LARC where retrospectively analyzed. An MRI of the pelvis before and after nRCT was performed. In the diagnostic biopsy, we measured the expression levels of 7 microRNAs, selected from literature, and correlated their expression levels with the tumor response rate (TRG), assessed on the surgical sample. We compared the accuracy of cCR prediction for: i) clinical predictors (age and stage); ii) radiomic features; iii) microRNAs levels); iv) combination of radiomics and microRNAs.
Statistical analysis
Clinical characteristics were compared between the two groups using t-test on continuous variables and using the Chi-square test for categorical ones. Radiomic data underwent a 3-steps processing: 1) pairwise correlation; 2) univariate association test between features and TRG class; 3) Principal Component Analysis (PCA) was performed on the remaining features and the first component (PC1) was extracted. The association between PC1 and TRG class was investigated with a logistic regression (LR) model.
For micro-RNA data, correlation analysis was conducted using Pearson method. Shapiro-Wilk test was performed for normality of data and Levene’s test for homogeneity of variance. Comparison of means between RESP and NO-RESP groups was performed using the t-test.
Variables of the different domains were used as predictors to estimate LR models and the predicted probabilities were used to calculate the Area Under the ROC curve (AUC-ROC).
Results:
Clinical predictors showed the lowest accuracy. The best performing model was based on the integration of radiomic features with mirR-145 expression level (AUC-ROC=0.90, 95%CI: 0.81-1.0). AI algorithm based on radiomics features, showed association with the TRG-class, and demonstrate a significant impact on the outcome. The overexpression of miR-145, seem to positively correlate with mjor response to nRCT, making a possible reliable biomarker.
Discussion
Currently, there are no biomarkers that accurately predict response to nCRT and thus the decision to begin this therapy is essentially clinical. Recent evidence points to an emerging role of microRNAs as predictive biomarkers for the identification of responder patients.
In addition to biomarkers, recent studies point to the use of radiomics as a powerful tool to predict the severity of the lesions. Radiomic represent quantitative and objective measures; could reflect tumor heterogeneity and sub-regional habitats and could improve diagnostic accuracy.
Conclusions:
Divide patients in non-responders and responders to nRCT could, in fact, spare inefficient and possibly harming treatments and design ad-hoc therapeutic approaches. Poor responders could be leaded directly to upfront surgery, reducing the risk related to nRCT, including the hypothesis of tumor progression. Integration of radiomics features and miR-145 expression, seems to be a positive predictive factor in order to recognize good responders to nRCT.
Background and Aims
Artificial intelligence (AI) associate to radiomic features and microRNAs are promising Locally Advanced Rectal Cancer (LARC) biomarkers. Establishing the association between microRNAs and MRI-based radiomic measures, will help define their significance and potential impact in prediction of complete clinical response (cCR) to neoadjuvant radio-chemotherapy (nRCT).
Study Design
40 patients with LARC where retrospectively analyzed. An MRI of the pelvis before and after nRCT was performed. In the diagnostic biopsy, we measured the expression levels of 7 microRNAs, selected from literature, and correlated their expression levels with the tumor response rate (TRG), assessed on the surgical sample. We compared the accuracy of cCR prediction for: i) clinical predictors (age and stage); ii) radiomic features; iii) microRNAs levels); iv) combination of radiomics and microRNAs.
Statistical analysis
Clinical characteristics were compared between the two groups using t-test on continuous variables and using the Chi-square test for categorical ones. Radiomic data underwent a 3-steps processing: 1) pairwise correlation; 2) univariate association test between features and TRG class; 3) Principal Component Analysis (PCA) was performed on the remaining features and the first component (PC1) was extracted. The association between PC1 and TRG class was investigated with a logistic regression (LR) model.
For micro-RNA data, correlation analysis was conducted using Pearson method. Shapiro-Wilk test was performed for normality of data and Levene’s test for homogeneity of variance. Comparison of means between RESP and NO-RESP groups was performed using the t-test.
Variables of the different domains were used as predictors to estimate LR models and the predicted probabilities were used to calculate the Area Under the ROC curve (AUC-ROC).
Results:
Clinical predictors showed the lowest accuracy. The best performing model was based on the integration of radiomic features with mirR-145 expression level (AUC-ROC=0.90, 95%CI: 0.81-1.0). AI algorithm based on radiomics features, showed association with the TRG-class, and demonstrate a significant impact on the outcome. The overexpression of miR-145, seem to positively correlate with mjor response to nRCT, making a possible reliable biomarker.
Discussion
Currently, there are no biomarkers that accurately predict response to nCRT and thus the decision to begin this therapy is essentially clinical. Recent evidence points to an emerging role of microRNAs as predictive biomarkers for the identification of responder patients.
In addition to biomarkers, recent studies point to the use of radiomics as a powerful tool to predict the severity of the lesions. Radiomic represent quantitative and objective measures; could reflect tumor heterogeneity and sub-regional habitats and could improve diagnostic accuracy.
Conclusions:
Divide patients in non-responders and responders to nRCT could, in fact, spare inefficient and possibly harming treatments and design ad-hoc therapeutic approaches. Poor responders could be leaded directly to upfront surgery, reducing the risk related to nRCT, including the hypothesis of tumor progression. Integration of radiomics features and miR-145 expression, seems to be a positive predictive factor in order to recognize good responders to nRCT.
Artificial intelligence (AI) associate to radiomic features and microRNAs are promising Locally Advanced Rectal Cancer (LARC) biomarkers. Establishing the association between microRNAs and MRI-based radiomic measures, will help define their significance and potential impact in prediction of complete clinical response (cCR) to neoadjuvant radio-chemotherapy (nRCT).
Study Design
40 patients with LARC where retrospectively analyzed. An MRI of the pelvis before and after nRCT was performed. In the diagnostic biopsy, we measured the expression levels of 7 microRNAs, selected from literature, and correlated their expression levels with the tumor response rate (TRG), assessed on the surgical sample. We compared the accuracy of cCR prediction for: i) clinical predictors (age and stage); ii) radiomic features; iii) microRNAs levels); iv) combination of radiomics and microRNAs.
Statistical analysis
Clinical characteristics were compared between the two groups using t-test on continuous variables and using the Chi-square test for categorical ones. Radiomic data underwent a 3-steps processing: 1) pairwise correlation; 2) univariate association test between features and TRG class; 3) Principal Component Analysis (PCA) was performed on the remaining features and the first component (PC1) was extracted. The association between PC1 and TRG class was investigated with a logistic regression (LR) model.
For micro-RNA data, correlation analysis was conducted using Pearson method. Shapiro-Wilk test was performed for normality of data and Levene’s test for homogeneity of variance. Comparison of means between RESP and NO-RESP groups was performed using the t-test.
Variables of the different domains were used as predictors to estimate LR models and the predicted probabilities were used to calculate the Area Under the ROC curve (AUC-ROC).
Results:
Clinical predictors showed the lowest accuracy. The best performing model was based on the integration of radiomic features with mirR-145 expression level (AUC-ROC=0.90, 95%CI: 0.81-1.0). AI algorithm based on radiomics features, showed association with the TRG-class, and demonstrate a significant impact on the outcome. The overexpression of miR-145, seem to positively correlate with mjor response to nRCT, making a possible reliable biomarker.
Discussion
Currently, there are no biomarkers that accurately predict response to nCRT and thus the decision to begin this therapy is essentially clinical. Recent evidence points to an emerging role of microRNAs as predictive biomarkers for the identification of responder patients.
In addition to biomarkers, recent studies point to the use of radiomics as a powerful tool to predict the severity of the lesions. Radiomic represent quantitative and objective measures; could reflect tumor heterogeneity and sub-regional habitats and could improve diagnostic accuracy.
Conclusions:
Divide patients in non-responders and responders to nRCT could, in fact, spare inefficient and possibly harming treatments and design ad-hoc therapeutic approaches. Poor responders could be leaded directly to upfront surgery, reducing the risk related to nRCT, including the hypothesis of tumor progression. Integration of radiomics features and miR-145 expression, seems to be a positive predictive factor in order to recognize good responders to nRCT.
Code of conduct/disclaimer available in General Terms & Conditions
{{ help_message }}
{{filter}}