Radiomics-based Artificial Intelligence System to Predict Neoadjuvant Treatment Response in Rectal Cancer

• Conditions: Rectal Cancer

• Registration Number: NCT04273477
• Lead Sponsor: Sixth Affiliated Hospital, Sun Yat-sen University

Brief Summary

In this study, investigators utilize a radiomics prediction model to predict the tumor response to neoadjuvant chemoradiotherapy (nCRT) before the nCRT is administered for patients with locally advanced rectal cancer (LARC). Previously, the radiomics prediction model has been constructed based on the radiomics features extracted from pretreatment Magnetic Resonance Imaging (MRI) in the training set, and optimized in the external validation set. The predictive power of this radiomics prediction model to discriminate the pathologic complete response (pCR) patients from non-pCR individuals, will be further verified in this prospective, multicenter clinical study.

Detailed Description

This is a multicenter, prospective, observational clinical study for validation of a radiomics-based artificial intelligence (AI) prediction model. Patients who have been pathologically diagnosed as rectal adenocarcinoma and defined as clinical II-III staging without distant metastasis will be enrolled from the Sixth Affiliated Hospital of Sun Yat-sen University, the Third Affiliated Hospital of Kunming Medical College and Sir Run Run Shaw Hospital Affiliated by Zhejiang University School of Medicine. All participants should follow a standard treatment protocol, including concurrent neoadjuvant chemoradiotherapy (nCRT), total mesorectum excision (TME) surgery and adjuvant chemotherapy. Enhanced Magnetic Resonance Imaging (MRI) examination should be completed before the administration of nCRT treatment. The tumor volumes at high solution T2-weighted, contrast-enhanced T1-weighted and diffusion weighted images will be manually delineated, respectively. The outlined MRI images will be captured by the radiomics prediction model to generate a predicted response ("predicted pCR" vs. "predicted non-pCR") of each patient, whereas the true response ("confirmed pCR" vs. "confirmed non-pCR") is derived from pathologic reports after TME surgery serving as the gold standard for evaluation. The prediction accuracy, specificity, sensitivity and Area Under Curve (AUC) of Receiver Operating Characteristic (ROC) curves will be calculated. This study is aimed to provide a reliable and accurate AI system to predict the pathologic tumor response to nCRT before its administration, which might facilitate the identification of pCR candidates for further precision therapy among patients with locally advanced rectal cancer.

Study Timeline

• Study Start: 2020-01-10
• Status Verified: 2020-02
• Study First Submitted: 2020-02-15
• Study First Submitted QC: 2020-02-15
• Last Update Submitted: 2020-02-15
• Study First Posted: 2020-02-18
• Last Update Posted: 2020-02-18
• Primary Completion: 2020-07
• Study Completion: 2020-12

Recruitment & Eligibility

• Status: UNKNOWN
• Sex: All
• Target Recruitment: 100

Inclusion Criteria

• pathologically diagnosed as rectal adenocarcinoma
• defined as clinical II-III staging (≥T3, and/or positive nodal status) without distant metastasis by enhanced Magnetic Resonance Imaging (MRI)
• intending to receive or undergoing neoadjuvant concurrent chemoradiotherapy (5-fluorouracil based chemotherapy, given orally or intravenously; Intensity-Modulated Radiotherapy or Volume-Modulated Radiotherapy delivered at 50 gray (Gy) in gross tumor volume (GTV) and 45 Gy in clinical target volume (CTV) by 25 fractions)
• intending to receive total mesorectum excision (TME) surgery after neoadjuvant therapy (not completed at the enrollment), and adjuvant chemotherapy
• MRI (high-solution T2-weighted imaging, contrast-enhanced T1-weighted imaging, and diffusion-weighted imaging are required) examination is completed before the neoadjuvant chemoradiotherapy

Exclusion Criteria

• with history of other cancer
• insufficient imaging quality of MRI to delineate tumor volume or obtain measurements (e.g., lack of sequence, motion artifacts)
• incomplete neoadjuvant chemoradiotherapy
• no surgery after neoadjuvant chemoradiotherapy resulting in lack of pathologic assessment of tumor response
• tumor recurrence or distant metastasis during neoadjuvant chemoradiotherapy

Study & Design

• Study Type: OBSERVATIONAL
• Study Design: Not specified

Primary Outcome Measures

Name	Time	Method
The prediction accuracy of the radiomics prediction model	baseline	The prediction accuracy of the MRI radiomics-based artificial intelligence prediction system for identifying pCR candidates from non-pCR individuals among nCRT treated LARC patients will be calculated.

Secondary Outcome Measures

Name	Time	Method
The area under curve (AUC) of Receiver Operating Characteristic (ROC) curves of the radiomics prediction model	baseline	The area under curve (AUC) of Receiver Operating Characteristic (ROC) curves of the MRI radiomics-based artificial intelligence prediction system for identifying pCR candidates from non-pCR individuals among nCRT treated LARC patients will be calculated.
The specificity of the radiomics prediction model	baseline	The specificity of the MRI radiomics-based artificial intelligence prediction system for identifying pCR candidates from non-pCR individuals among nCRT treated LARC patients will be calculated.
The sensitivity of the radiomics prediction model	baseline	The sensitivity of the MRI radiomics-based artificial intelligence prediction system for identifying pCR candidates from non-pCR individuals among nCRT treated LARC patients will be calculated.

Trial Locations

Locations (3)

the Sixth Affiliated Hospital of Sun Yat-sen University; 🇨🇳 Guangzhou, Guangdong, China

The Third Affiliated Hospital of Kunming Medical College; 🇨🇳 Kunming, Yunnan, China

Sir Run Run Shaw Hospital; 🇨🇳 Hangzhou, Zhejiang, China