Studies suggest natural language processing (NLP) is an effective tool in identifying candidates for epilepsy surgery.
A review of 6 studies found natural language processing (NLP) showed moderate-to-high performance levels in identifying suitable candidates to undergo epilepsy surgery, an effective, but oftentimes underutilized treatment, in which approximately 50% to 60% of patients became seizure-free after surgery.
“This study has found that there is evidence, using multiple algorithms, that NLP may aid in the identification of candidates who may benefit from referral for epilepsy surgery evaluation,” wrote the researchers of the study. “It is noteworthy that these studies have shown that the NLP approaches may identify suitable candidates prior to the time that treating neurologists refer their patients.”
The systemic review is published in the Journal of Clinical Neuroscience.
Similar to machine learning, NLP uses computers to analyze or interact with human language and has various uses in health care, including information extraction, information retrieval, document categorization, and text summarization. Furthermore, NLP can aid in generating meaningful information, such as diagnosis or prognosis from electric health record data.
Previous research has suggested that NLP may be useful in identifying patients with drug-resistant focal epilepsy, who account for about 30% of individuals with epilepsy. In the current review, researchers aimed to examine previous studies using NLP to identify patients for epilepsy surgery.
A data search identified 1369 publication results from PubMed (n = 324), EMBASE (n = 94) and Cochrane library (n = 951), in which 58 full-text articles were identified for review.
After exclusion, 6 studies were selected for analysis. Most studies were conducted in a single study center, with 1 study utilizing data from 2 centers, and 1 study from 6 centers. Study characteristics included were the number of participants, age, gender, and NLP information, such as task assigned, ground truth (gold standard), and the type of NLP algorithms used.
Five of the 6 studies used support vector machines and 1 study used NLP strategies, such as random forest models and gradient boosted machines. Furthermore, all studies showed moderate-to-to-high levels of performance.
Some of the studies showed that NLP could identify patients 1 to 2 years prior to the treating clinicians initial referral. However, none of the studies identified evaluated the influence of implementing these algorithms on health care systems or patient outcomes.
“NLP is a promising technology for the identification of patients who may benefit from epilepsy surgery referral,” wrote the researchers.
The researchers acknowledge some limitations to the study, including a small sample size, most studies being in a single center, and only including English-language studies. Additionally, the researchers believe that further research needs to be done to trial novel algorithms using NLP to identify patients for epilepsy surgery.
Despite these limitations, the researchers believe the study was able to reach moderate-to-high performance levels in selecting patients who may benefit from epilepsy surgery, when other treatments are no longer an appropriate option for them.
The researchers noted that future interventional studies that implement NLP algorithms and evalute their impact on health system and patient outcomes should focus on more diverse center.
"Further research may aim to trial novel algorithms in the identification of patients who may be suitable for epilepsy surgery evaluation, such as deep learning methods and bidirectional encoder representations from transformers. Additionally, future studies may seek to evaluate the generalisability of machine learning NLP for these tasks across diverse sites."
Resources
Tan S, Tang C, Ng JS, et al. Identifying epilepsy surgery candidates with natural language processing: a systematic review. J Clin Neurosci. 2023;114:104-109. doi:10.1016/j.jocn.2023.06.010
Etonogestrel Contraceptive Implants Reduce Pain Crises in Women With Sickle Cell Disease
October 23rd 2024Etonogestrel-releasing contraceptive implants in women with sickle cell disease significantly reduced pain intensity and frequency of pain crises over 12 months, with no adverse changes in metabolic or liver function markers.
Read More