An image processing algorithm to aid diagnosis of mesial temporal sclerosis in children: a case-control study

• Published in: Pediatric radiology 2020, Vol.50 (1), p.98-106
• Main Authors: Strnad, Benjamin S., Orlowski, Hilary L. P., Parsons, Matthew S., Salter, Amber, Dahiya, Sonika, Sharma, Aseem
• Format: Article
• Language: English
• Subjects: Adolescent; Adult; Algorithms; Case-Control Studies; Child; Child, Preschool; Children; Diagnosis; Epilepsy; Epilepsy, Temporal Lobe - diagnostic imaging; Epilepsy, Temporal Lobe - pathology; Female; Females; Hippocampus; Humans; Image contrast; Image detection; Image enhancement; Image Interpretation, Computer-Assisted - methods; Image processing; Imaging; Magnetic resonance imaging; Magnetic Resonance Imaging - methods; Male; Males; Medical diagnosis; Medical imaging; Medicine; Medicine & Public Health; Neuroimaging; Neuroradiology; NMR; Noise; Noise control; Noise intensity; Nuclear magnetic resonance; Nuclear Medicine; Oncology; Original Article; Pediatrics; Radiology; Reproducibility of Results; Retrospective Studies; Sclerosis; Seizures; Sensitivity; Sensitivity and Specificity; Signal processing; Temporal Lobe - diagnostic imaging; Temporal Lobe - pathology; Ultrasound; Young Adult
• ISSN: 0301-0449; 1432-1998
• DOI: 10.1007/s00247-019-04518-x

Background Mesial temporal sclerosis (MTS) is an important cause of intractable epilepsy. Early and accurate diagnosis of MTS is essential to providing curative and life-changing therapy but can be challenging in children in whom the impact of diagnosis is particularly high. Magnetic resonance imaging (MRI) plays an important role in the diagnosis of MTS, and image processing of MRI is a recently studied strategy to improve its accuracy. Objective In a retrospective case-control study, we assessed the performance of an image processing algorithm (Correlative Image Enhancement [CIE]) for detecting MTS-related hippocampal signal abnormality in children. Materials and methods Twenty-seven pediatric MTS cases (9 males, 18 females; mean age: 16±standard deviation [SD] 6.7 years) were identified from a pathology database of amygdylo-hippocampectomies performed in children with epilepsy. Twenty-seven children with no seizure history (9 males, 18 females; mean age: 13.8±SD 2.8 years), and with normal brain MRI, were identified for the control group. Blinded investigators processed the MRI coronal FLAIR (fluid-attenuated inversion recovery) images with CIE, saved the processed images as a separate series, and made equivalent region of interest measurements on the processed and unprocessed series to calculate contrast-to-noise ratio. Six blinded reviewers then rated the randomized series for hippocampal signal abnormality and MTS disease status. Results CIE increased signal intensity and contrast-to-noise ratio in 26/27 hippocampi with pathologically confirmed MTS (96.3%) with the mean (SD) contrast-to-noise ratio of cases increasing from 14.9 (11.1) to 77.7 (58.7) after processing ( P