Lesion defined by Multiparametric MRI ISODATA correlates well with neurological deficit and PWI
OBJECTIVE: To show that the lesion defined by ISODATA (Iterative Self Organizing Data Analysis Technique) has a greater correlation with the deficit on perfusion weighted image (PWI) and National Institute of Health Stroke Scale (NIHSS) compared to the lesion defined by diffusion weighted image (DWI) or T2 weighted image (T2WI). BACKGROUND: Early in the course of acute ischemic stroke, the PWI lesion area is usually larger than the DWI or T2WI lesion areas, and correlates best with clinical neurological deficit. The computer segmentation algorithm ISODATA integrates different MRI parameters (DWI, T2WI, T1WI) and provides complimentary information about status of the tissue. METHODS: We obtained MRI parameters of DWI, T2WI and T1WI from 11 patients at two time points after stroke, the acute (<12 hrs) and chronic (3 months). The clinical neurological deficit was graded by the NIHSS at each time point. PWI was obtained at the acute time point. Relative mean transit time images (rMTT) were used to define the perfusion deficit. We compared PWI deficit area with DWI, ISODATA and T2WI lesion of the same location. The lesion areas were also correlated with NIHSS at acute time point. At chronic time point T2WI and ISODATA lesion volumes were compared with NIHSS. RESULTS: In acute phase PWI lesion area correlated best with NIHSS (r=0.72, p<0.05), followed by ISODATA (r=0.62, p<0.05), DWI (r =0.51, p<0.05) & T2WI (r=0.21, p<0.05) lesion areas. ISODATA correlated best with PWI lesion (r=0.90, p<0.05), and outperformed DWI (r=0.63, p<0.05) and T2WI (r= 0.48, p<0.05). We also noticed at chronic time point that there is a good correlation between ISODATA and T2WI(r= 0.97, p<0.05), ISODATA and NIHSS (r= 0.70, p<0.0.5), T2WI and NIHSS (r=0.68, p<0.05). CONCLUSION: This preliminary study shows that the integrated ISODATA approach to tissue segmentation discriminates abnormal from normal tissue at each time point after stroke. The area of the ISODATA defined lesion correlates well with PWI at the acute and T2WI at the chronic time point and the clinical neurological status of the patient.