1
|
Cai L, Lambregts DMJ, Beets GL, Mass M, Pooch EHP, Guérendel C, Beets-Tan RGH, Benson S. An automated deep learning pipeline for EMVI classification and response prediction of rectal cancer using baseline MRI: a multi-centre study. NPJ Precis Oncol 2024; 8:17. [PMID: 38253770 PMCID: PMC10803303 DOI: 10.1038/s41698-024-00516-x] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 08/16/2023] [Accepted: 12/14/2023] [Indexed: 01/24/2024] Open
Abstract
The classification of extramural vascular invasion status using baseline magnetic resonance imaging in rectal cancer has gained significant attention as it is an important prognostic marker. Also, the accurate prediction of patients achieving complete response with primary staging MRI assists clinicians in determining subsequent treatment plans. Most studies utilised radiomics-based methods, requiring manually annotated segmentation and handcrafted features, which tend to generalise poorly. We retrospectively collected 509 patients from 9 centres, and proposed a fully automated pipeline for EMVI status classification and CR prediction with diffusion weighted imaging and T2-weighted imaging. We applied nnUNet, a self-configuring deep learning model, for tumour segmentation and employed learned multiple-level image features to train classification models, named MLNet. This ensures a more comprehensive representation of the tumour features, in terms of both fine-grained detail and global context. On external validation, MLNet, yielding similar AUCs as internal validation, outperformed 3D ResNet10, a deep neural network with ten layers designed for analysing spatiotemporal data, in both CR and EMVI tasks. For CR prediction, MLNet showed better results than the current state-of-the-art model using imaging and clinical features in the same external cohort. Our study demonstrated that incorporating multi-level image representations learned by a deep learning based tumour segmentation model on primary MRI improves the results of EMVI classification and CR prediction with good generalisation to external data. We observed variations in the contributions of individual feature maps to different classification tasks. This pipeline has the potential to be applied in clinical settings, particularly for EMVI classification.
Collapse
Affiliation(s)
- Lishan Cai
- Department of Radiology, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX, Amsterdam, The Netherlands
- GROW School for Oncology and Developmental Biology, Maastricht University Medical Centre, P. Debyelaan 25, 66202 AZ, Maastricht, The Netherlands
| | - Doenja M J Lambregts
- Department of Radiology, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX, Amsterdam, The Netherlands
- GROW School for Oncology and Developmental Biology, Maastricht University Medical Centre, P. Debyelaan 25, 66202 AZ, Maastricht, The Netherlands
| | - Geerard L Beets
- GROW School for Oncology and Developmental Biology, Maastricht University Medical Centre, P. Debyelaan 25, 66202 AZ, Maastricht, The Netherlands
- Department of Surgery, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX, Amsterdam, The Netherlands
| | - Monique Mass
- Department of Radiology, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX, Amsterdam, The Netherlands
- GROW School for Oncology and Developmental Biology, Maastricht University Medical Centre, P. Debyelaan 25, 66202 AZ, Maastricht, The Netherlands
| | - Eduardo H P Pooch
- Department of Radiology, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX, Amsterdam, The Netherlands
- GROW School for Oncology and Developmental Biology, Maastricht University Medical Centre, P. Debyelaan 25, 66202 AZ, Maastricht, The Netherlands
| | - Corentin Guérendel
- Department of Radiology, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX, Amsterdam, The Netherlands
- GROW School for Oncology and Developmental Biology, Maastricht University Medical Centre, P. Debyelaan 25, 66202 AZ, Maastricht, The Netherlands
| | - Regina G H Beets-Tan
- Department of Radiology, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX, Amsterdam, The Netherlands
- GROW School for Oncology and Developmental Biology, Maastricht University Medical Centre, P. Debyelaan 25, 66202 AZ, Maastricht, The Netherlands
| | - Sean Benson
- Department of Radiology, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX, Amsterdam, The Netherlands.
| |
Collapse
|
2
|
Zhu K, Chen Z, Cui L, Zhao J, Liu Y, Cao J. The Preoperative Diagnostic Performance of Multi-Parametric Quantitative Assessment in Rectal Carcinoma: A Preliminary Study Using Synthetic Magnetic Resonance Imaging. Front Oncol 2022; 12:682003. [PMID: 35707367 PMCID: PMC9190242 DOI: 10.3389/fonc.2022.682003] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 03/17/2021] [Accepted: 04/19/2022] [Indexed: 12/24/2022] Open
Abstract
Objective Synthetic MRI (SyMRI) can reconstruct different contrast-weighted images(T1, T2, PD) and has shorter scan time, easier post-processing and better reproducibility. Some studies have shown splendid correlation with conventional mapping techniques and no degradation in the quality of syMRI images compared with conventional MRI. It is crucial to select an individualized treatment plan based on the preoperative images of rectal carcinoma (RC). We tried to explore the feasibility of syMRI on T, N stage and extramural vascular invasion (EMVI) of rectal cancer. Materials and Methods A total of 100 patients (37 females and 63 males) diagnosed with rectal carcinoma were enrolled. All the patients underwent preoperative pelvic MR examinations including conventional MR sequence and synthetic MRI. Two radiologists evaluated the MRI findings of each rectal carcinoma and EMVI score in consensus. The values for T1, T2 relaxation times and PD value were measured in tumor(ROI-1) and pararectal fat space(ROI-2) and analyzed independently. A receiver operating characteristic (ROC) analysis was performed. Correlations between the T1, T2 and PD values and EMVI score were also evaluated. Results Compared with the normal rectal wall, the values of T1 and T2 relaxation times of the tumor were significantly higher (P <0.001). There was no statistically significant difference in the PD value (P >0.05). As for ROI, the ROI of pararectal fat space(ROI-2) had better significance than rectal cancer lesion (ROI-1). T2 value of ROI-1 and T1 value of ROI-2 were higher in the pEMVI positive group than in the negative group (P=0.002 and 0.001) and T1 value of ROI-2 had better performance with an AUC of 0.787, (95% CI:0.693- 0.882). T1 value, T2 value and PD value from ROI-2 were effective for both T and N stage of rectal cancer. High-grade pathological stage had showed higher T1 value (PT stage=0.013,PN stage=0.035), lower T2 value (PT stage=0.025,PN stage=0.034) and lower PD value (PT stage=0.017). We also enrolled the characteristics with P < 0.05 in the combined model which had better diagnostic efficacy. A significant positive correlation was found between the T1 value of pararectal fat space(ROI-2) and EMVI score (r value = 0.519, P<0.001). The T2 value(r=0.213,P=0.049) and PD value(r=0.354,P=0.001) from ROI-1 was correlated with EMVI score. Correlation analysis did not show any significant associations between T2 value of tumor, T2, PD values of pararectal fat space and EMVI scores. Conclusion Synthetic MRI can provide multi-parameter quantitative image maps with a easier measurement and slightly shorter acquisition time compared with conventional MRI. The measurement of multi-parametric quantitative values contributes to diagnosing the tumor and evaluating T stage, N stage and EMVI. It has the potential to be used as a preoperative diagnostic and grading technique in rectal carcinoma.
Collapse
Affiliation(s)
- Kexin Zhu
- Department of Radiology, The First Affiliated Hospital of China Medical University, Shenyang, China
| | - Zhicheng Chen
- Department of Radiology, Shengjing Hospital of China Medical University, Shenyang, China
| | - Lingling Cui
- Department of Radiology, The First Affiliated Hospital of China Medical University, Shenyang, China
| | - Jinli Zhao
- Department of Radiology, The First Affiliated Hospital of China Medical University, Shenyang, China
| | - Yi Liu
- Department of Radiology, The First Affiliated Hospital of China Medical University, Shenyang, China
| | - Jibin Cao
- Department of Radiology, The First Affiliated Hospital of China Medical University, Shenyang, China
| |
Collapse
|
3
|
Shu Z, Mao D, Song Q, Xu Y, Pang P, Zhang Y. Multiparameter MRI-based radiomics for preoperative prediction of extramural venous invasion in rectal cancer. Eur Radiol 2021; 32:1002-1013. [PMID: 34482429 DOI: 10.1007/s00330-021-08242-9] [Citation(s) in RCA: 27] [Impact Index Per Article: 9.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 04/13/2021] [Revised: 07/21/2021] [Accepted: 08/02/2021] [Indexed: 12/17/2022]
Abstract
OBJECTIVES To compare multiparameter MRI-based radiomics for preoperative prediction of extramural venous invasion (EMVI) in rectal cancer using different machine learning algorithms and to develop and validate the best diagnostic model. METHODS We retrospectively analyzed 317 patients with rectal cancer. Of these, 114 were EMVI positive and 203 were EMVI negative. Radiomics features were extracted from T2-weighted imaging, T1-weighted imaging, diffusion-weighted imaging, and enhanced T1-weighted imaging of rectal cancer, followed by the dimension reduction of the features. Logistic regression, support vector machine, Bayes, K-nearest neighbor, and random forests algorithms were trained to obtain the radiomics signatures. The area under the receiver operating characteristic curve (AUC) was used to evaluate the performance of each radiomics signature. The best radiomics signature was selected and combined with clinical and radiological characteristics to construct a joint model for predicting EMVI. Finally, the predictive performance of the joint model was assessed. RESULTS The Bayes-based radiomics signature performed well in both the training set and the test set, with the AUCs of 0.744 and 0.738, sensitivities of 0.754 and 0.728, and specificities of 0.887 and 0.918, respectively. The joint model performed best in both the training set and the test set, with the AUCs of 0.839 and 0.835, sensitivities of 0.633 and 0.714, and specificities of 0.901 and 0.885, respectively. CONCLUSIONS The joint model demonstrated the best diagnostic performance for the preoperative prediction of EMVI in patients with rectal cancer. Hence, it can be used as a key tool for clinical individualized EMVI prediction. KEY POINTS • Radiomics features from magnetic resonance imaging can be used to predict extramural venous invasion (EMVI) in rectal cancer. • Machine learning can improve the accuracy of predicting EMVI in rectal cancer. • Radiomics can serve as a noninvasive biomarker to monitor the status of EMVI.
Collapse
Affiliation(s)
- Zhenyu Shu
- Department of Radiology, Zhejiang Provincial People's Hospital, Affiliated People's Hospital, Hangzhou Medical College, Hangzhou, Zhejiang, China
| | - Dewang Mao
- Department of Radiology, Zhejiang Provincial People's Hospital, Affiliated People's Hospital, Hangzhou Medical College, Hangzhou, Zhejiang, China
| | - Qiaowei Song
- Department of Radiology, Zhejiang Provincial People's Hospital, Affiliated People's Hospital, Hangzhou Medical College, Hangzhou, Zhejiang, China
| | - Yuyun Xu
- Department of Radiology, Zhejiang Provincial People's Hospital, Affiliated People's Hospital, Hangzhou Medical College, Hangzhou, Zhejiang, China
| | - Peipei Pang
- Department of Pharmaceuticals Diagnosis, GE Healthcare, Hangzhou, China
| | - Yang Zhang
- Department of Radiology, Zhejiang Provincial People's Hospital, Affiliated People's Hospital, Hangzhou Medical College, Hangzhou, Zhejiang, China.
| |
Collapse
|
4
|
Junquera-Olay S, Baleato-González S, Canedo-Antelo M, Capeans-González L, Santiago-Pérez MI, Garcia-Figueiras R. "Rectal cancer survival: A retrospective analysis of MRI features and their association with prognosis". Curr Probl Diagn Radiol 2021; 51:30-37. [PMID: 33483190 DOI: 10.1067/j.cpradiol.2020.12.005] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 09/17/2020] [Revised: 12/05/2020] [Accepted: 12/31/2020] [Indexed: 11/22/2022]
Abstract
PURPOSE To assess rectal cancer aggressiveness using magnetic resonance (MR) imaging features and to investigate their relationship with patient prognosis. MATERIALS AND METHODS Clinical information and Pelvic MR scans of 106 consecutive patients with primary rectal cancer (RC) were analyzed. Clinical symptoms, age, sex, tumor location, and patient´s survival were recorded. The variables investigated by MR were: depth or mural/extramural tumor involvement, distance to mesorectal margin, lymph node involvement, vascular, peritoneal or sphincter complex infiltration. The association between imaging features and disease-free survival (DFS) was also assessed using a Kaplan-Meier model. Differences between survival curves were tested for significance using the Mantel-Cox LogRank test. RESULTS The final study population was 106 patients (65 males, 41 females). The median age was 69.5 years (range, 39-92 years). No significant differences were found between death risk and sex, age or tumor location (p>0,05). However, the relative risk (RR) of tumor mortality increased significantly with the presence of the variables: vascular infiltration (×5), T4 tumors (× 4.57), N2 lymph node involvement (more than 3 affected nodes × 4.11) and mesorectal fascia involvement (× 3,77). CONCLUSION Tumor extension, number of pathological lymph nodes, mesorectal fascia involvement and vascular infiltration values obtained on initial MR imaging staging showed a significant difference for disease-free survival in RC at six years of control.
Collapse
Affiliation(s)
- Sonsoles Junquera-Olay
- Department of Radiology, Santiago de Compostela University Hospital, Choupana Avenue, Santiago de Compostela, Coruña, 15706, Spain.
| | - Sandra Baleato-González
- Department of Radiology, Santiago de Compostela University Hospital, Choupana Avenue, Santiago de Compostela, Coruña, 15706, Spain
| | - María Canedo-Antelo
- Department of Radiology, Santiago de Compostela University Hospital, Choupana Avenue, Santiago de Compostela, Coruña, 15706, Spain
| | | | | | - Roberto Garcia-Figueiras
- Department of Radiology, Santiago de Compostela University Hospital, Choupana Avenue, Santiago de Compostela, Coruña, 15706, Spain
| |
Collapse
|