TY - JOUR
T1 - Feasibility of UTE-MRI-based radiomics model for prediction of histopathologic subtype of lung adenocarcinoma
T2 - in comparison with CT-based radiomics model
AU - Lee, Suji
AU - Lee, Chang Young
AU - Kim, Na Young
AU - Suh, Yong Joo
AU - Lee, Hye Jeong
AU - Yong, Hwan Seok
AU - Kim, Hye Ryun
AU - Kim, Young Jin
N1 - Publisher Copyright:
© The Author(s), under exclusive licence to European Society of Radiology 2023.
PY - 2024/5
Y1 - 2024/5
N2 - Objectives: To assess the feasibility of the UTE-MRI radiomic model in predicting the micropapillary and/or solid (MP/S) patterns of surgically resected lung adenocarcinoma. Materials and methods: We prospectively enrolled 74 lesions from 71 patients who underwent UTE-MRI and CT before curative surgery for early lung adenocarcinoma. For conventional radiologic analysis, we analyzed the longest lesion diameter and lesion characteristics at both UTE-MRI and CT. Radiomic features were extracted from the volume of interest of the lesions and Rad-scores were generated using the least absolute shrinkage and selection operator with fivefold cross-validation. Six models were constructed by combining the conventional radiologic model, UTE-MRI Rad-score, and CT Rad-score. The areas under the curves (AUCs) of each model were compared using the DeLong method. Early recurrence after curative surgery was analyzed, and Kaplan–Meier survival analysis was performed. Results: Twenty-four lesions were MP/S-positive, and 50 were MP/S-negative. The longitudinal size showed a small systematic difference between UTE-MRI and CT, with fair intermodality agreement of lesion characteristic (kappa = 0.535). The Rad-scores of the UTE-MRI and CT demonstrated AUCs of 0.84 and 0.841, respectively (p = 0.98). Among the six models, mixed conventional, UTE-MRI, and CT Rad-score model showed the highest diagnostic performance (AUC = 0.879). In the survival analysis, the high- and low-risk groups were successfully divided by the Rad-score in UTE-MRI (p = 0.01) and CT (p < 0.01). Conclusion: UTE-MRI radiomic model predicting MP/S positivity is feasible compared with the CT radiomic model. Also, it was associated with early recurrence in the survival analysis. Clinical relevance statement: A radiomic model utilizing UTE-MRI, which does not present a radiation hazard, was able to successfully predict the histopathologic subtype of lung adenocarcinoma, and it was associated with the patient’s recurrence-free survival. Key Points: • No studies have reported the ultrashort echo time (UTE)-MRI-based radiomic model for lung adenocarcinoma. • The UTE-MRI Rad-score showed comparable diagnostic performance with CT Rad-score for predicting micropapillary and/or solid histopathologic pattern. • UTE-MRI is feasible not only for conventional radiologic analysis, but also for radiomics analysis.
AB - Objectives: To assess the feasibility of the UTE-MRI radiomic model in predicting the micropapillary and/or solid (MP/S) patterns of surgically resected lung adenocarcinoma. Materials and methods: We prospectively enrolled 74 lesions from 71 patients who underwent UTE-MRI and CT before curative surgery for early lung adenocarcinoma. For conventional radiologic analysis, we analyzed the longest lesion diameter and lesion characteristics at both UTE-MRI and CT. Radiomic features were extracted from the volume of interest of the lesions and Rad-scores were generated using the least absolute shrinkage and selection operator with fivefold cross-validation. Six models were constructed by combining the conventional radiologic model, UTE-MRI Rad-score, and CT Rad-score. The areas under the curves (AUCs) of each model were compared using the DeLong method. Early recurrence after curative surgery was analyzed, and Kaplan–Meier survival analysis was performed. Results: Twenty-four lesions were MP/S-positive, and 50 were MP/S-negative. The longitudinal size showed a small systematic difference between UTE-MRI and CT, with fair intermodality agreement of lesion characteristic (kappa = 0.535). The Rad-scores of the UTE-MRI and CT demonstrated AUCs of 0.84 and 0.841, respectively (p = 0.98). Among the six models, mixed conventional, UTE-MRI, and CT Rad-score model showed the highest diagnostic performance (AUC = 0.879). In the survival analysis, the high- and low-risk groups were successfully divided by the Rad-score in UTE-MRI (p = 0.01) and CT (p < 0.01). Conclusion: UTE-MRI radiomic model predicting MP/S positivity is feasible compared with the CT radiomic model. Also, it was associated with early recurrence in the survival analysis. Clinical relevance statement: A radiomic model utilizing UTE-MRI, which does not present a radiation hazard, was able to successfully predict the histopathologic subtype of lung adenocarcinoma, and it was associated with the patient’s recurrence-free survival. Key Points: • No studies have reported the ultrashort echo time (UTE)-MRI-based radiomic model for lung adenocarcinoma. • The UTE-MRI Rad-score showed comparable diagnostic performance with CT Rad-score for predicting micropapillary and/or solid histopathologic pattern. • UTE-MRI is feasible not only for conventional radiologic analysis, but also for radiomics analysis.
KW - Lung neoplasm
KW - Magnetic resonance imaging
KW - Pathology
KW - Progression-free survival
KW - Radiomics
UR - http://www.scopus.com/inward/record.url?scp=85174178946&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85174178946&partnerID=8YFLogxK
U2 - 10.1007/s00330-023-10302-1
DO - 10.1007/s00330-023-10302-1
M3 - Article
C2 - 37840100
AN - SCOPUS:85174178946
SN - 0938-7994
VL - 34
SP - 3422
EP - 3430
JO - European Radiology
JF - European Radiology
IS - 5
ER -