1722

Validation of Radiomics Signature for Chemoradiotherapy Prediction of Advanced Cervical Cancer Based on a High Resolution T2WI Images
Defeng Liu1, Qinglei Shi2, Xu Yan2, Lanxiang Liu1, Yujie Cui1, Xiaohang Zhang3, and Juan Du4

1Qinhuangdao Municipal No. 1 Hospital, Qinhuangdao, China, 2MR Scientific Marketing, Siemens Healthcare, Qinhuangdao, China, 3University of Chinese Academy of Sciences, Beijing, China, 4Graduate School of Hebei Medical University, Shijiazhuang, China

Synopsis

This study performed a radiomics signature analysis based on a high resolution T2WI images, and evaluate the value of these quantitative features in prediction the treatment effect of neoadjuvant chemotherapy-radiation therapy for advanced cervical cancer (>IIb) . And found that Shape and first-order features seems can provide valuable information and showed potential in prediction treatment effect of this disease.

Purpose

To study the predictive ability of radiomics signature for advanced cervical cancer (>IIb) treated with neoadjuvant chemotherapy-radiation therapy based on a high resolution T2WI images.

Materials and Methods

This retrospective study included 100 patients with locally advanced cervical cancer scanned from March 2013 to May 2018. Baseline and posttherapy MRI and follow-up data were retrieved for all patients. During these cases, 86 cases were squamous carcinoma and 14 cases were adenocarcinoma, and the categories of pathological staging include: 26 cases in IIb, 37 cases in IIIa, 19 cases in IIIb, 11 cases in IVa, 7 cases in IVb. All these patients received concurrent chemoradiotherapy, and all MR examinations were performed before treatment within one month. According to the curative effect, the patients were divided into two groups: complete remission group and partial remission group. All patients were scanned at a 3 T scanner (MAGNETOM Verio, Siemens Healthcare, Erlangen , Germany). The segmentation of the cancer and the calculation of the cancer were performed using an in-house developed tool written in Python 3.5. The features about shape and first-order were extracted and analyzed by using an Independent-samples t test to test the difference, and with receiver operating characteristic curve (ROC) to evaluate the diagnostic performance with SPSS software 18.0 (SPSS, Chicago, IL ). P value<0.05 was considered statistically significant difference.

Results

Significant difference were found in shape derived features (VoxelNum, LeastAxis, Maximum2DDiameterRow, SurfaceArea, MinorAxis, Maximum2DDiameterColumn, Maximum3DDiameter, aximum2DDiameterSlice, Sphericity, Volume, MajorAxis) and in first-order features (10Percentile, Mean, TotalEnergy, Energy, 90Percentil, RootMeanSquared) (All P<0.05) (Table 1). After analysis of ROC, the features of shape, the SurfaceArea demonstrated a highest AUC (0.859), for the features of first-order, the TotalEnergy and Energy get the highest AUC (0.863) (Table 2).

Conclusions

Shape and first-order features derived from radiomics signature seems can provide valuable information and showed potential in prediction treatment effect for advanced cervical cancer (>IIb) treated with neoadjuvant chemotherapy-radiation therapy based on a high resolution T2WI images.

Acknowledgements

No acknowledgement found.

References

No reference found.

Figures

Fig.1. The AUC of features about shape.

Fig.2. The AUC of features about first order.

Fig.3. (a) partial respones (b) complete respones


Table.1 Shape and first-order features between complete remission group and partial remission groups.

Table.2 The area under curve (AUC) for features showed significant difference between two groups.

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)
1722