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.

The Combination of Whole-Brain Features and Local-Lesion Features in DSC-PWI May Improve Ischemic Stroke Outcome Prediction

“…Medical images, such as CTP and PWI, have been widely used in stroke detection, lesion segmentation, final infarct prediction, and prognosis assessment [27,43,[47][48][49], proving that these imaging features are closely related to the current and future rehabilitation of stroke patients. Although the ability of features in stroke lesions to reflect functional recovery has been revealed [23][24][25], few studies are related to the role of whole-brain features in stroke outcome prediction. Under the inspiration of Ref.…”

“…Next, this study used neuroimaging software package FSL [31] to segment the skull from the average 3D DSC-PWI image, then obtain the brain tissue region of DSC-PWI images. Since the time period of the contrast agent passed through is in the range of about 17 to 22 and the end of the reaction is located at a time value greater than 30 [25], the average 3D image was computed from the first 15th and the last 15th 3D images to reduce the influence of the contrast agent. The computation of the average 3D image is expressed as Equation (1).…”

“…Then, the radiomics Moreover, we used the data smoothing filter with a 1 × 3 filtering kernel introduced in Ref. [25] to smooth DSC-PWI images and decrease their noise. Then, the time-intensity curve of each voxel in the DSC-PWI image was smoothed.…”

“…Feature selection plays a crucial role in data mining and improving algorithms performance, which is necessary work in feature processing. The Lasso algorithm has been recognized as one of the most effective selection methods for selecting relevant features with target variables [25,35,36]. Therefore, this study used Lasso to select outstanding radiomics features for 90-day mRS, and the feature with a non-zero coefficient was selected as one of the outstanding DRFs.…”

“…Ref. [25] extracted radiomics features from 3D images in time series of dynamic susceptibility contrast-enhanced perfusionweighted imaging (DSC-PWI) to generate dynamic radiomics features (DRFs), and the performance of DRFs, CTI, generated survival features, and the combination of them were compared and evaluated. The results showed that the CTI performed better than DRFs of ischemic lesions, and the additional survival features can improve the predictive ability.…”