Machine learning-based segmentation of ischemic penumbra by using diffusion tensor metrics in a rat model

Abstract

Background: Recent trials have shown promise in intra-arterial thrombectomy after the first 6-24 h of stroke onset. Quick and precise identification of the salvageable tissue is essential for successful stroke management. In this study, we examined the feasibility of machine learning (ML) approaches for differentiating the ischemic penumbra (IP) from the infarct core (IC) by using diffusion tensor imaging (DTI)-derived metrics. Methods: Fourteen male rats subjected to permanent middle cerebral artery occlusion (pMCAO) were included in this study. Using a 7 T magnetic resonance imaging, DTI metrics such as fractional anisotropy, pure anisotropy, diffusion magnitude, mean diffusivity (MD), axial diffusivity, and radial diffusivity were derived. The MD and relative cerebral blood flow maps were coregistered to define the IP and IC at 0.5 h after pMCAO. A 2-level classifier was proposed based on DTI-derived metrics to classify stroke hemispheres into the IP, IC, and normal tissue (NT). The classification performance was evaluated using leave-one-out cross validation. Results: The IC and non-IC can be accurately segmented by the proposed 2-level classifier with an area under the receiver operating characteristic curve (AUC) between 0.99 and 1.00, and with accuracies between 96.3 and 96.7%. For the training dataset, the non-IC can be further classified into the IP and NT with an AUC between 0.96 and 0.98, and with accuracies between 95.0 and 95.9%. For the testing dataset, the classification accuracy for IC and non-IC was 96.0 ± 2.3% whereas for IP and NT, it was 80.1 ± 8.0%. Overall, we achieved the accuracy of 88.1 ± 6.7% for classifying three tissue subtypes (IP, IC, and NT) in the stroke hemisphere and the estimated lesion volumes were not significantly different from those of the ground truth (p = .56, .94, and .78, respectively). Conclusions: Our method achieved comparable results to the conventional approach using perfusion-diffusion mismatch. We suggest that a single DTI sequence along with ML algorithms is capable of dichotomizing ischemic tissue into the IC and IP.

Estimating the volume of penumbra in rodents using DTI and stack-based ensemble machine learning framework

“…Similarly, we previously reported that the combination of DTI metrics and ML can effectively identify the penumbra without the need for contrast agents [38]. In the present study, we adopted a stack-based ensemble technique, substantially increasing the Pearson correlation coefficient between the ML-estimated PV and the PDM-defined PV from 0.61 [38] to our value of 0.93. Despite its optimal performance, this ensemble technique requires extended training time and high computational resources [39].…”

“…Despite FA being extensively examined as a potential diagnostic biomarker among DTI-derived metrics, its performance in the hyperacute phase remains controversial because of its definition as a ratio of q to L [59,60]. The increase, reduction, or no change in FA values is dependent on the simultaneous analysis of q and L. However, studies have consistently reported decreased MD values [8,38,60], which is reflected in our importance permutation results. MD, p, and L reflect the magnitudes of molecular motion of water, which changes and becomes detectable within minutes after stroke onset.…”

“…Deep learning models can help identify the penumbra without the use of any contrast agents [ 36 , 37 ]. Similarly, we previously reported that the combination of DTI metrics and ML can effectively identify the penumbra without the need for contrast agents [ 38 ]. In the present study, we adopted a stack-based ensemble technique, substantially increasing the Pearson correlation coefficient between the ML-estimated PV and the PDM-defined PV from 0.61 [ 38 ] to our value of 0.93.…”

“…In contrast to a previous study that used only a single ML algorithm [ 38 ], our study used a stack-based ensemble framework. Although the results of McNemar test indicated that the stacking model may not consistently outperform a base model because various factors influence the success of ensemble models [ 47 ], the major advantage of ensemble methods lies in their stability, which can substantially enhance performance and reduce bias compared with single model-based approaches [ 48 ].…”

Neuropilin-1 promotes mitochondrial structural repair and functional recovery in rats with cerebral ischemia

“…Stroke has been ranked as a primary cause of disability and death in the elderly populace worldwide [ 1 ] and in the early phases of ischemic stroke, severe cell death and mitochondrial dysfunction may result in long-term neurological impairment [ 2 ]. As a countermeasure, vascular recanalization is currently the most efficacious treatment for acute ischemic stroke.…”