A comparison of machine learning classifiers for pediatric epilepsy using resting-state functional MRI latency data.

Epilepsy affects 1 in 150 children under the age of 10 and is the most common chronic pediatric neurological condition; poor seizure control can irreversibly disrupt normal brain development. The present study compared the ability of different machine learning algorithms trained with resting-state functional MRI (rfMRI) latency data to detect epilepsy. Preoperative rfMRI and anatomical MRI scans were obtained for 63 patients with epilepsy and 259 healthy controls. The normal distribution of latency z-scores from the epilepsy and healthy control cohorts were analyzed for overlap in 36 seed regions. In these seed regions, overlap between the study cohorts ranged from 0.44-0.58. Machine learning features were extracted from latency z-score maps using principal component analysis. Extreme Gradient Boosting (XGBoost), Support Vector Machines (SVM), and Random Forest algorithms were trained with these features. Area under the receiver operating characteristics curve (AUC), accuracy, sensitivity, specificity and F1-scores were used to evaluate model performance. The XGBoost model outperformed all other models with a test AUC of 0.79, accuracy of 74%, specificity of 73%, and a sensitivity of 77%. The Random Forest model performed comparably to XGBoost across multiple metrics, but it had a test sensitivity of 31%. The SVM model did not perform >70% in any of the test metrics. The XGBoost model had the highest sensitivity and accuracy for the detection of epilepsy. Development of machine learning algorithms trained with rfMRI latency data could provide an adjunctive method for the diagnosis and evaluation of epilepsy with the goal of enabling timely and appropriate care for patients.

Convolutional Neural Networks for Pediatric Refractory Epilepsy Classification Using Resting-State Functional Magnetic Resonance Imaging.

OBJECTIVE: This study aims to evaluate the performance of convolutional neural networks (CNNs) trained with resting-state functional magnetic resonance imaging (rfMRI) latency data in the classification of patients with pediatric epilepsy from healthy controls. METHODS: Preoperative rfMRI and anatomic magnetic resonance imaging scans were obtained from 63 pediatric patients with refractory epilepsy and 259 pediatric healthy controls. Latency maps of the temporal difference between rfMRI and the global mean signal were calculated using voxel-wise cross-covariance. Healthy control and epilepsy latency z score maps were pseudorandomized and partitioned into training data (60%), validation data (20%), and test data (20%). Healthy control individuals and patients with epilepsy were labeled as negative and positive, respectively. CNN models were then trained with the designated training data. Model hyperparameters were evaluated with a grid-search method. The model with the highest sensitivity was evaluated using unseen test data. Accuracy, sensitivity, specificity, F1 score, and area under the receiver operating characteristic curve were used to evaluate the ability of the model to classify epilepsy in the test data set. RESULTS: The model with the highest validation sensitivity correctly classified 74% of unseen test patients with 85% sensitivity, 71% specificity, F1 score of 0.56, and an area under the receiver operating characteristic curve of 0.86. CONCLUSIONS: Using rfMRI latency data, we trained a CNN model to classify patients with pediatric epilepsy from healthy controls with good performance. CNN could serve as an adjunct in the diagnosis of pediatric epilepsy. Identification of pediatric epilepsy earlier in the disease course could decrease time to referral to specialized epilepsy centers and thus improve prognosis in this population.

Reducing opioid usage: a pilot study comparing postoperative selective dorsal rhizotomy protocols.

OBJECTIVE: Selective dorsal rhizotomy (SDR) is a surgical procedure used to treat spasticity in children with spastic cerebral palsy. Currently, there is a lack of work examining the efficacy of optimizing pain management protocols after single-level laminectomy for SDR. This pilot study aimed to compare the clinical outcomes of SDR completed with a traditional pain management protocol versus one designed for opioid dosage reduction. METHODS: The Texas Comprehensive Spasticity Center prospective database was queried for all patients who underwent SDR between 2015 and 2018. Demographic, surgical, and postoperative data for all patients who underwent SDR were collected from medical records. The study was designed as a retrospective study between the patient-controlled analgesia (PCA) and dexmedetomidine infusion (INF) groups with 80% power to detect a 50% difference at a significance level of 0.05. Patients in the INF group received perioperative gabapentin, intraoperative dexmedetomidine infusion, and scheduled acetaminophen and NSAIDs postoperatively. RESULTS: Medication administration records, pain scores, and therapy notes were collected for 30 patients. Patients who underwent SDR between June 2015 and the end of December 2017 received traditional pain management (PCA group, n = 14). Patients who underwent SDR between January 2018 and the end of December 2018 received modified pain management (INF group, n = 16). No patients were lost to follow-up. Differences in age, weight, height, preoperative Gross Motor Function Classification System scores, operative duration, hospital length of stay, and sex distribution were not statistically different between the 2 groups (p > 0.05). Analysis of analgesic medication doses demonstrated that the INF group required fewer doses and lower amounts of opioids overall, and also fewer NSAIDs than the PCA group. When converted to the morphine milligram equivalent, the patients in the INF group used fewer doses and lower amounts of opioids overall than the PCA group. These differences were either statistically significant (p < 0.05) or trending toward significance (p < 0.10). Both groups participated in physical and occupational therapy similarly postoperatively (p > 0.05). Pain scores were comparable between the groups (p > 0.05) despite patients in the INF group requiring fewer opioids. CONCLUSIONS: Infusion with dexmedetomidine during SDR surgery combined with perioperative gabapentin and scheduled acetaminophen and NSAIDs postoperatively resulted in similar pain scores to traditional pain management with opioids. In addition, this pilot study demonstrated that patients who received the INF pain management protocol required reduced opioid dosages and were able to participate in therapy similarly to the control PCA group.

Role of resting state MRI temporal latency in refractory pediatric extratemporal epilepsy lateralization.

BACKGROUND: Pediatric epilepsy affects 0.5-1% of children, with 10-30% of these children refractory to medical anticonvulsant therapy and potentially requiring surgical intervention. Analysis of resting state functional MRI (rsMRI) signal temporal differences (latency) has been proposed to study the pathological cognitive processes. PURPOSE/HYPOTHESIS: To quantitatively and qualitatively analyze the correlation of rsMRI signal latency to pediatric refractory extratemporal epilepsy seizure foci lateralization. STUDY TYPE: Retrospective review. POPULATION: With Institutional Review Board approval, rsMRI and anatomical MRI scans were obtained from 38 registered pediatric epilepsy surgery patients from Washington University and 259 healthy control patients from the ADHD-200 dataset. FIELD STRENGTH/SEQUENCE: 3 T echo planar imaging (EPI) blood oxygenation level-dependent (BOLD) sequence. ASSESSMENT: The images were transformed to pediatric atlases in Talairach space. Preoperative voxelwise latency maps were generated with parabolic interpolation of the rsMRI signal lateness or earliness when compared with the global mean signal (GMS) using cross-covariance analysis. STATISTICAL TESTS: Latency z-score maps were created for each epilepsy patient by voxelwise calculation using healthy control mean and standard deviation maps. Voxelwise hypothesis testing was performed via multiple comparisons corrected (false discovery and familywise error rate) and uncorrected methods to determine significantly late and early voxels. Significantly late and/or early voxels were counted for the right and left hemisphere separately. The hemisphere with the greater proportion of significantly late and/or early voxels was hypothesized to contain the seizure focus. Preoperative rsMRI latency analysis hypotheses were compared with postoperative seizure foci lateralization determined by resection images. RESULTS: Preoperative rsMRI latency analysis correctly identified seizure foci lateralization of 64-85% of postoperative epilepsy resections with the proposed methods. RsMRI latency lateralization analysis was 77-100% sensitive and 58-79% specific. In some patients, qualitative analysis yielded preoperative rsMRI latency patterns specific to procedure performed. DATA CONCLUSION: Preoperative rsMRI signal latency of pediatric epilepsy patients was correlated with seizure foci lateralization. J. Magn. Reson. Imaging 2019;49:1347-1355.

Synthesis and characterization of pH-responsive nanoscale hydrogels for oral delivery of hydrophobic therapeutics.

pH-responsive, polyanionic nanoscale hydrogels were developed for the oral delivery of hydrophobic therapeutics, such as common chemotherapeutic agents. Nanoscale hydrogels were designed to overcome physicochemical and biological barriers associated with oral delivery of hydrophobic therapeutics such as low solubility and poor permeability due to P-glycoprotein related drug efflux. Synthesis of these nanoscale materials was achieved by a robust photoemulsion polymerization method. By varying hydrophobic monomer components, four formulations were synthesized and screened for optimal physicochemical properties and in vitro biocompatibility. All of the responsive nanoscale hydrogels were capable of undergoing a pH-dependent transition in size. Depending on the selection of the hydrophobic monomer, the sizes of the nanoparticles vary widely from 120nm to about 500nm at pH 7.4. Polymer composition was verified using Fourier transform infrared spectroscopy and 1H-nuclear magnetic resonance spectroscopy. Polymer biocompatibility was assessed in vitro with an intestinal epithelial cell model. All formulations were found to have no appreciable cytotoxicity, defined as greater than 80% viability after polymer incubation. We demonstrate that these nanoscale hydrogels possess desirable physicochemical properties and exhibit agreeable in vitro biocompatibility for oral delivery applications.