Review of Natural Language Processing in Radiology.

Natural language processing (NLP) is an interdisciplinary field, combining linguistics, computer science, and artificial intelligence to enable machines to read and understand human language for meaningful purposes. Recent advancements in deep learning have begun to offer significant improvements in NLP task performance. These techniques have the potential to create new automated tools that could improve clinical workflows and unlock unstructured textual information contained in radiology and clinical reports for the development of radiology and clinical artificial intelligence applications. These applications will combine the appropriate application of classic linguistic and NLP preprocessing techniques, modern NLP techniques, and modern deep learning techniques.

Do epileptic seizures predict outcome in patients with oligodendroglioma?

BACKGROUND: Many patients with an oligodendroglioma (OD) experience seizures, some of which become refractory to anti-epileptic drugs (AEDs). This study aims (1) to quantify the rate of seizures and medically refractory epilepsy in patients with ODs; and (2) to determine if there is any association between short-term and long-term survival, and the presence and drug-responsiveness of seizures. METHODS: A retrospective review was conducted of the medical records of patients who had been pathologically identified as having an OD at the London Health Sciences Centre or the London Regional Cancer Program in London, Ontario from January 1996 to July 2008. Deaths were ascertained by reviewing all hospital records. Survival analysis was performed. RESULTS: One-hundred sixty-six patients met inclusion criteria. Epileptic seizures were the presenting feature or occurred as part of the initial manifestation of the OD in 75.3% of patients, with 90.4% (n=150) experiencing at least one seizure and 76.5% developing epilepsy over the course of observation. Of the 150 patients with seizures, 23 experienced a single seizure (13.9% of the 166), whereas 127 patients experienced multiple seizures (76.5%). In those with multiple seizures, the epilepsy was refractory to drug treatment slightly more than half the time (54.3%). Survival analysis demonstrated consistently superior survival among those with a single seizure. Those without seizures had the worst survival rates over the first few years post-diagnosis; but then no further deaths occurred. Survival among those with refractory seizures tended to be better than among those whose seizures were drug responsive, over the first 10 years post-diagnosis. CONCLUSIONS: Seizures are common and may influence survival in patients with oligodendogliomas. Those who experience just one seizure appear to do best.

Prevalence of benign epileptiform variants.

OBJECTIVE: There are numerous distinctive benign electroencephalographic (EEG) patterns which are morphologically epileptiform but are non-epileptic. The aim of this study was to determine the prevalence of different benign epileptiform variants (BEVs) among subjects who underwent routine EEG recordings in a large EEG laboratory over 35 years. METHODS: We retrospectively studied the prevalence of BEVs among 35,249 individuals who underwent outpatient EEG recordings at London Health Sciences Centre in London, Ontario, Canada between January 1, 1972 and December 31, 2007. The definitions of the Committee on Terminology of the International Federation of Societies for EEG and Clinical Neurophysiology (IFSECN) were used to delineate epileptiform patterns (Chatrian et al. A glossary of terms most commonly used by clinical electroencephlographers. Electroenceph Clin Neurophysiol 1974;37:538-48) and the descriptions of Klass and Westmoreland [Klass DW, Westmoreland BF. Nonepileptogenic epileptiform electroenephalographic activity. Ann Neurol 1985;18:627-35] were used to categorize the BEVs. RESULTS: BEVs were identified in 1183 out of 35,249 subjects (3.4%). The distribution of individual BEVs were as follows: benign sporadic sleep spikes 1.85%, wicket waves 0.03%, 14 and 6 Hz positive spikes 0.52%, 6 Hz spike-and-waves 1.02%, rhythmic temporal theta bursts of drowsiness 0.12%, and subclinical rhythmic electrographic discharge of adults in 0.07%. CONCLUSION: The prevalence of six types of BEVs was relatively low among the Canadian subjects when compared to the reports from other countries. SIGNIFICANCE: BEVs are relatively uncommon incidental EEG findings. Unlike focal epileptic spikes and generalized spike-and-waves, BEVs do not predict the occurrence of epilepsy. Accurate identification of the BEVs can avoid misdiagnosis and unnecessary investigations.