IC‐P1‐003: Registration of Serial MRI Improves Diagnosis of Dementia

Co-registration of serial MRI has proved valuable in understanding neurodegenerative diseases and quantifying atrophy progression for trial outcomes. However, its diagnostic utility in a clinical setting has not been evaluated. A retrospective study was conducted using subjects referred to a cognitive disorders clinic. Subjects who visited clinic three or more times with a follow-up of at least one year were selected resulting in 120 subjects with two scans and a clinical/pathological diagnosis (Table 1). Subjects' scans were reviewed as a single first scan (method 1), the two unregistered presented side-by-side (method 2), and the registered pair (method 3). Separate assessments were made by a neuro-radiologist and a clinician. For each method each reviewer had to make both a main and a specific (subtype) diagnosis. Table 2 shows that there was some evidence that results from the three methods differed between reviewers. However a test of interaction was significant only for subtype diagnosis (p=0.042). For the clinician, distinction of normal/non-progressors from the disease categories was better using both methods 2 and 3 than method 1, however this was not significant (p=0.12). Both for distinction of main diagnosis and of subtype there was some evidence that the clinician's performance was worse with method 2 than with the other methods. For subtype diagnosis a global test for differences between the methods was of borderline significance (p=0.053), and a pair-wise comparison of method 3 with method 2 yielded p=0.025. For the radiologist, serial imaging presentations aided distinction of normal/non-progressors from patients (p=0.001), with no difference between methods 2 and 3 (p=0.77). Both for distinction of main disease diagnosis and of subtype there was some evidence that the radiologist's performance was better with methods 2 and 3 than with method 1. The differences were not significant for main diagnosis, but for subtype diagnosis both the global test of differences between the three groups (p=0.007) and the pair-wise comparison between methods 1 and 3 (p=0.002) was significant. Serial MRI and its registration may improve visual diagnosis in dementia, although the benefits may differ according to the experience and training of the reviewer.