Indirect Volume Estimation for Acute Ischemic Stroke from Diffusion Weighted Image Using Slice Image Segmentation

Abstract

The accurate estimation of acute ischemic stroke (AIS) using diffusion-weighted imaging (DWI) is crucial for assessing patients and guiding treatment options. This study aimed to propose a method that estimates AIS volume in DWI objectively, quickly, and accurately. We used a dataset of DWI with AIS, including 2159 participants (1179 for internal validation and 980 for external validation) with various types of AIS. We constructed algorithms using 3D segmentation (direct estimation) and 2D segmentation (indirect estimation) and compared their performances with those annotated by neurologists. The proposed pretrained indirect model demonstrated higher segmentation performance than the direct model, with a sensitivity, specificity, F1-score, and Jaccard index of 75.0%, 77.9%, 76.0, and 62.1%, respectively, for internal validation, and 72.8%, 84.3%, 77.2, and 63.8%, respectively, for external validation. Volume estimation was more reliable for the indirect model, with 93.3% volume similarity (VS), 0.797 mean absolute error (MAE) for internal validation, VS of 89.2% and a MAE of 2.5% for external validation. These results suggest that the indirect model using 2D segmentation developed in this study can provide an accurate estimation of volume from DWI of AIS and may serve as a supporting tool to help physicians make crucial clinical decisions.

The Role of Artificial Intelligence (Ai) on Mri Brain Examination With Clinical Ischemic Stroke

“…The next AI application in MRI brain examination with clinical ischemic stroke is the prediction of infarct volume (Figure 4). This is supported by the research conducted by Nielsen et al (2018), Benzakoun et al (2021), Yu et al (2020), and Lee et al (2022). In a study conducted by Yu et al (2020), the accuracy of infarct volume prediction is quite high with a median DSC (DSC, (0 ≤ DSC ≤ 1)) of 0.53 (IQR (interquartile range), 0.31 -0.68)), and obtained a volume error of 9 ml with IQR -14 -29 ml, absolute volume error 24 ml with IQR 11 -50 ml.…”

“…In a study conducted by Yu et al (2020), the accuracy of infarct volume prediction is quite high with a median DSC (DSC, (0 ≤ DSC ≤ 1)) of 0.53 (IQR (interquartile range), 0.31 -0.68)), and obtained a volume error of 9 ml with IQR -14 -29 ml, absolute volume error 24 ml with IQR 11 -50 ml. While the research conducted by Lee et al (2022) obtained F1 scores of 0.76 and 0.772, and obtained volume similarity (VS) of 93.3% and 89.2% respectively. According to the theory of Yu et al (2020) and Lee et al (2022), objective, accurate, and rapid prediction of infarct volume helps determine medical decisions and patient prognosis.…”

“…While the research conducted by Lee et al (2022) obtained F1 scores of 0.76 and 0.772, and obtained volume similarity (VS) of 93.3% and 89.2% respectively. According to the theory of Yu et al (2020) and Lee et al (2022), objective, accurate, and rapid prediction of infarct volume helps determine medical decisions and patient prognosis.…”

“…Based on the research of Woo et al (2019) combining two types of CNNs, namely U-Net and DenseNet, can increase the level of diagnostic accuracy in segmenting stroke lesions using an AI model. Meanwhile, based on the research by Lee et al (2022), the indirect segmentation (2D) CNN model has more stable performance. Lastly, Ho et al (2019) state that the Logistic Regression (LR) type machine learning model has an Area Under Curve (AUC) value of 0.765 in the TSS classification test.…”

“…Ho et al (2019) stated that the performance of the AI model was low due to limited training data, while Wei et al (2021) stated that data limitations made it difficult to generalize the results of the AI model. Zhu et al (2021) and Lee et al(2022)…”