Novel Survival Features Generated by Clinical Text Information and Radiomics Features May Improve the Prediction of Ischemic Stroke Outcome

Abstract

Background: Accurate outcome prediction is of great clinical significance in customizing personalized treatment plans, reducing the situation of poor recovery, and objectively and accurately evaluating the treatment effect. This study intended to evaluate the performance of clinical text information (CTI), radiomics features, and survival features (SurvF) for predicting functional outcomes of patients with ischemic stroke.

Methods: SurvF was constructed based on CTI and mRS radiomics features (mRSRF) to improve the prediction of the functional outcome in 3 months (90-day mRS). Ten machine learning models predicted functional outcomes in three situations (2-category, 4-category, and 7-category) using seven feature groups constructed by CTI, mRSRF, and SurvF.

Results: For 2-category, ALL (CTI + mRSRF+ SurvF) performed best, with an mAUC of 0.884, mAcc of 0.864, mPre of 0.877, mF1 of 0.86, and mRecall of 0.864. For 4-category, ALL also achieved the best mAuc of 0.787, while CTI + SurvF achieved the best score with mAcc = 0.611, mPre = 0.622, mF1 = 0.595, and mRe-call = 0.611. For 7-category, CTI + SurvF performed best, with an mAuc of 0.788, mPre of 0.519, mAcc of 0.529, mF1 of 0.495, and mRecall of 0.47.

Conclusions: The above results indicate that mRSRF + CTI can accurately predict functional outcomes in ischemic stroke patients with proper machine learning models. Moreover, combining SurvF will improve the prediction effect compared with the original features. However, limited by the small sample size, further validation on larger and more varied datasets is necessary.

Keywords: clinical text information; ischemic stroke outcome; machine learning; radiomics features; survival features.

Figure 1

Flowchart of this study. (a) Process of the preprocessing of DSC-PWI datasets and making ROIs; (b) Calculating radiomics features, where the value of the feature is represented by color; (c–e) The process of feature selection, feature fusion, and stroke outcome prediction.

Figure 2

Flowchart of selecting significant features. $$F_{i0}^{*}$$ and $$F_{i1}^{*}$$ are the ith feature in LT and NT groups, respectively.

Figure 3

Flowchart of multidimensional feature fusion.

Figure 4

Significant features in the radiomics features group. (a,b) Ratio and p-value of significant features in the eight radiomics feature groups, respectively.

Figure 5

Performance of the 13 feature sets on the ten learning models. (a) Five mean indexes of 13 methods on the ten models and (b) CS of 13 methods.

Figure 6

Selected mRSRF and statistics in mRS_2, mRS_4, and mRS_7. (a) mRSRF in three situations, and green color represents selected items from 128 outstanding features in F_method. (b) Box plot among the three groups of mRSRF. (c) C-index of the Deepsurv model in training. (d,e) Pearson correlation coefficients and p-values among the three groups of mRSRF.

Figure 7

Performance of seven feature groups in the situation of mRS_2. (a–e) Auc, Pre, Acc, F1, and Recall on the ten models. (f) ROC curves of seven feature groups on the RF model.

Figure 8

Performance of seven feature groups in the situation of mRS_4. (a–e) Auc, Pre, Acc, F1, and Recall on the ten models. (f) ROC curves of seven feature groups on the RF model.

Figure 9

Performance of seven feature groups in the situation of mRS_7. (a–e) Auc, Pre, Acc, F1, and Recall on the ten models. (f) ROC curves of seven feature groups on the RF model.