Sequence Labeling Research Articles

A Neural Span-Based Continual Named Entity Recognition Model

Named Entity Recognition (NER) models capable of Continual Learning (CL) are realistically valuable in areas where entity types continuously increase (e.g., personal assistants). Meanwhile the learning paradigm of NER advances to new patterns such as the span-based methods. However, its potential to CL has not been fully explored. In this paper, we propose SpanKL, a simple yet effective Span-based model with Knowledge distillation (KD) to preserve memories and multi-Label prediction to prevent conflicts in CL-NER. Unlike prior sequence labeling approaches, the inherently independent modeling in span and entity level with the designed coherent optimization on SpanKL promotes its learning at each incremental step and mitigates the forgetting. Experiments on synthetic CL datasets derived from OntoNotes and Few-NERD show that SpanKL significantly outperforms previous SoTA in many aspects, and obtains the smallest gap from CL to the upper bound revealing its high practiced value. The code is available at https://github.com/Qznan/SpanKL.

Nested Named Entity Recognition as Building Local Hypergraphs

Named entity recognition is a fundamental task in natural language processing. Based on the sequence labeling paradigm for flat named entity recognition, multiple methods have been developed to handle the nested structures. However, they either require fixed recognition order or introduce complex hypergraphs. To tackle this problem, we propose a novel model named Local Hypergraph Builder Network (LHBN) that builds multiple simpler local hypergraphs to capture named entities instead of a single complex full-size hypergraph. The proposed model has three main properties: (1) The named entities that share boundaries are captured in the same local hypergraph. (2) The boundary information is enhanced by building local hypergraphs. (3) The hypergraphs can be built bidirectionally to take advantage of the identification direction preference of different named entities. Experiments illustrate that our model outperforms previous state-of-the-art methods on four widely used nested named entity recognition datasets: ACE04, ACE05, GENIA, and KBP17. The code is available at https://github.com/yanyk13/local-hypergraph-building-network.git.

Hierarchical Text Classification as Sub-hierarchy Sequence Generation

Hierarchical text classification (HTC) is essential for various real applications. However, HTC models are challenging to develop because they often require processing a large volume of documents and labels with hierarchical taxonomy. Recent HTC models based on deep learning have attempted to incorporate hierarchy information into a model structure. Consequently, these models are challenging to implement when the model parameters increase for a large-scale hierarchy because the model structure depends on the hierarchy size. To solve this problem, we formulate HTC as a sub-hierarchy sequence generation to incorporate hierarchy information into a target label sequence instead of the model structure. Subsequently, we propose the Hierarchy DECoder (HiDEC), which decodes a text sequence into a sub-hierarchy sequence using recursive hierarchy decoding, classifying all parents at the same level into children at once. In addition, HiDEC is trained to use hierarchical path information from a root to each leaf in a sub-hierarchy composed of the labels of a target document via an attention mechanism and hierarchy-aware masking. HiDEC achieved state-of-the-art performance with significantly fewer model parameters than existing models on benchmark datasets, such as RCV1-v2, NYT, and EURLEX57K.

Abnormal cerebral blood flow and brain function in type 2 diabetes mellitus.

Type 2 diabetes mellitus (T2DM) lead to impaired cerebral blood perfusion, which leads to changes in brain function and affects the cognitive function of patients. In this study, cerebral blood flow (CBF) was used to evaluate the effect of T2DM on cerebral perfusion, and functional connectivity (FC) analysis was further used to explore whether the FC between the abnormal CBF region and the whole brain was changed. In addition, amplitude of low-frequency fluctuation (ALFF) and degree centrality (DC) were used to investigate the changes in spontaneous activity and connectivity strength of the brain network. We recruited 40 T2DM patients and 55 healthy controls (HCs). They underwent 3D-T1WI, rs-fMRI, arterial spin labeling (ASL) sequence scans and a series of cognitive tests. Cognitive test scores and brain imaging indicators were compared between the two groups, and the relationships among laboratory indicators, cognitive test scores, and brain imaging indicators were explored in the T2DM group. Compared to HCs, The CBF values of Calcarine_L and Precuneus_R in the T2DM group were lower. The DC value of Paracentral_Lobule_L and Precuneus_L, and the ALFF value of Hippocampus_L in the T2DM group were higher. In addition, the CBF values of Calcarine_L was negatively correlated with fasting insulin and HOMA_IR. This study found that there were regions of cerebral hypoperfusion in T2DM patients, which are associated with insulin resistance. In addition, we found abnormally elevated brain activity and enhanced functional connectivity in T2DM patients, which we speculated was the compensatory mechanism of brain neural activity.

Part-of-Speech Tagging of Odia Language Using Statistical and Deep Learning Based Approaches

Automatic part-of-speech (POS) tagging is a preprocessing step of many natural language processing tasks, such as named entity recognition, speech processing, information extraction, word sense disambiguation, and machine translation. It has already gained promising results in English and European languages. However, in Indian languages, particularly in the Odia language, it is not yet well explored because of the lack of supporting tools, resources, and morphological richness of the language. Unfortunately, we were unable to locate an open source POS tagger for the Odia language, and only a handful of attempts have been made to develop POS taggers for the Odia language. The main contribution of this research work is to present statistical approaches such as the maximum entropy Markov model and conditional random field (CRF), as well as deep learning based approaches, including the convolutional neural network (CNN) and bidirectional long short-term memory (Bi-LSTM) to develop the Odia POS tagger. A publicly accessible corpus annotated with the Bureau of Indian Standards (BIS) tagset is used in our work. However, most of the languages around the globe have used the dataset annotated with the Universal Dependencies (UD) tagset. Hence, to maintain uniformity, the Odia dataset should use the same tagset. Thus, following the BIS and UD guidelines, we constructed a mapping from the BIS tagset to the UD tagset. The maximum entropy Markov model, CRF, Bi-LSTM, and CNN models are trained using the Indian Languages Corpora Initiative corpus with the BIS and UD tagsets. We have experimented with various feature sets as input to the statistical models to prepare a baseline system and observed the impact of constructed feature sets. The deep learning based model includes the Bi-LSTM network, the CNN network, the CRF layer, character sequence information, and a pre-trained word vector. Seven different combinations of neural sequence labeling models are implemented, and their performance measures are investigated. It has been observed that the Bi-LSTM model with the character sequence feature and pre-trained word vector achieved a result with 94.58% accuracy.

IMG-03. HARNESSING COMPUTATIONAL IMAGING, SCALABLE CLOUD-BASED WORKFLOWS, AND MULTI-MODAL DATA ANALYTICS WITH INTEROPERABLE SOFTWARE PLATFORMS IN PEDIATRIC NEURO-ONCOLOGY; THE ROAD AHEAD FOR CHILDREN’S BRAIN TUMOR NETWORK (CBTN)

Abstract Neuroimaging is acquired as part of the clinical standard of care for treatment of brain tumor patients. Capitalizing on this data and integration of imaging with other data types to perform predictive analytics is an increasing need. Here, we review the efforts at CBTN in the past couple of years to establish user-friendly workflows to support the entire imaging data lifecycle and to bridge access to multi-modal datasets with scalable analytics scale, to empower researchers to make breakthrough discoveries that will advance patient care. An automated deep learning-based brain extraction and tumor subregion segmentation model based on a multi-institutional dataset was developed that can reliably segment the MRI scans of children across a variety of brain tumor histologies. This model was subsequently integrated into an end-to-end imaging pipeline for collecting, managing, and analyzing clinically acquired radiology exams of the CBTN consortium with processing of 12,500 exams to date. A public website was also created to allow users to search and explore the dataset based on sequence labels, and clinical and imaging attributes. In addition, we established a scalable workflow and processes for de-identification, ingestion, quality control, and management of digital pathology slides on CBTN (over 8,000 slides to date). Finally, we established interoperability between imaging (Flywheel) and molecular (CAVATICA) platforms deployed in cloud ecosystems which allows streamlined, user-friendly workflows to ingest and harmonize files, perform cohort selection, prepare data using standard processing packages and cloud resources, and conduct analysis on extracted multi-modal feature sets. The described workflow lays the foundation for broad use of imaging studies and access to multi-modal datasets and analytics, with the goal of empowering researchers to make breakthroughs in patient care. The next step is to apply these tools in the context of clinical trials.

Learning Optimal Solutions via an LSTM-Optimization Framework

In this study, we present a deep learning-optimization framework to tackle dynamic mixed-integer programs. Specifically, we develop a bidirectional Long Short Term Memory (LSTM) framework that can process information forward and backward in time to learn optimal solutions to sequential decision-making problems. We demonstrate our approach in predicting the optimal decisions for the single-item capacitated lot-sizing problem (CLSP), where a binary variable denotes whether to produce in a period or not. Due to the dynamic nature of the production and inventory levels that must meet the periodic demand, the CLSP can be treated as a sequence labeling task where a recurrent neural network can capture the problem’s temporal dynamics. Computational results show that our LSTM-Optimization (LSTM-Opt) framework significantly reduces the solution time of benchmark CLSP problems without much loss in feasibility and optimality. For example, the predictions at the 85% level reduce the CPLEX solution time by a factor of 9 on average for over 240000 test instances with an optimality gap of less than 0.05% and 0.4% infeasibility in the test set. Also, models trained using shorter planning horizons can successfully predict the optimal solution of the instances with longer planning horizons. For the hardest data set, the LSTM predictions at the 25% level reduce the solution time of 70 CPU hours to less than 2 CPU minutes with an optimality gap of 0.8% and without any infeasibility. The LSTM-Opt framework outperforms classical ML algorithms, such as the logistic regression and random forest, in terms of the solution quality, and exact approaches, such as the ( $$\ell$$ , S) and dynamic programming-based inequalities, with respect to the solution time improvement. Our machine learning approach could be beneficial in tackling sequential decision-making problems similar to CLSP, which need to be solved repetitively, frequently, and in a fast manner.

An empirical study of extracting embedded text from digital images

The utilization of images as a means of transferring information is a widespread technique employed to circumvent simple detection functions that primarily focus on analyzing textual content rather than conducting thorough file examinations. This study investigates the efficacy of deep learning models in detecting embedded information within digital images. The data used for analysis was acquired from a secondary source and underwent comprehensive preprocessing. Feature extraction, sequence labeling, and predictive model training were performed using CRNN, CNN, and RNN models. Two specific models were trained and tested in this research: 1) CNN, RNN-LSTM with the Adam optimizer, and 2) CNN, RNN-GRU with the RAdam optimizer for text detection. The findings reveal that Model #1 achieved the highest F1-score during testing, with a score of 98.37% for text detection and 96.73% for word detection. The second model obtained an F1-score of 94.84% and 93.05% for text and word detection, respectively. Model #1 exhibited a word detection accuracy of 98.38% and a text detection accuracy of 96.47%. These findings indicate that the first model outperformed the second model, suggesting that employing RNN-LSTM and the Adam optimizer made a positive impact. Therefore, utilizing deep learning tools and emerging technologies is crucial for extracting textual information and analyzing visual data. In summary, this study concludes that deep learning models can be relied upon to effectively detect textual information embedded within digital images.

The named entity recognition of vessel power equipment fault using the multi-details embedding model

The fault diagnosis of vessel power equipment is established by the manual work with low efficiency. The knowledge graph(KG) usually is applied to extract the experience and operation logic of controllers into knowledge, which can enrich the means of fault judgment and recovery decision. As an important part of KG building, the performance of named entity recognition (NER) is critical to the following tasks. Due to the challenges of information insufficiency and polysemous words in the entities of vessel power equipment fault, this study adopts the fusion model of Bidirectional Encoder Representations from Transformers (BERT), revised Convolutional neural network (CNN), bidirectional long short-term memory (BiLSTM), and conditional random field (CRF). Firstly, the adjusted BERT and revised CNN are respectively adopted to acquire the multiple embeddings including semantic information and contextual glyph features. Secondly, the local context features are effectively extracted by adopting the channel-wised fusion structures. Finally, BiLSTM and CRF are respectively adopted to obtain the semantic information of the long sequences and the prediction sequence labels. The experimental results show that the performance of NER by the proposed model outperforms other mainstream models. Furthermore, this work provides the foundation of the tasks of intelligent diagnosis and NER in other fields.

MTMG: A multi-task model with multi-granularity information for drug-drug interaction extraction

Drug-drug interactions (DDIs) extraction includes identifying drug entities and interactions between drug pairs from the biomedical corpus. The discovery of potential DDIs aids in our understanding of the mechanisms underlying adverse reactions or combination therapy to improve patient safety. The manual extraction of DDIs is very time-consuming and expensive; therefore, computer-aided extraction of DDIs is vital. Many neural network-based methods have been proposed and achieved good efficiency in the extraction of DDIs over the years. However, most studies improved the performance of DDIs extraction with various external drug features while directly using golden drug entities, leading to error propagation and low universality in practical application. In this paper, we propose a new multi-task framework called MTMG, which changes DDIs extraction from a sentence-level classification task to a sequence labeling task named Drug-Specified Token Classification (DSTC). The proposed approach, MTMG, jointly trains DSTC with drug named entity recognition (DNER) and two sentence-level auxiliary tasks we designed. We aim to improve the performance of the entire DDIs extraction pipeline by better using the correlation between entities and relationships and, to the extent possible, using the information of varying granularity implied in the dataset. Experimental results show that MTMG can both improve the accuracy of DNER and DDIs extraction and outperforms state-of-the-art technique.

Performance analysis of sentiment classification based neural network

Deep learning has more significant advantages for word embedding technology than sentiment analysis. This paper studies the application of deep learning on the word embedding problem in context, mainly discusses the RNN model with Word2Vec and without Word2Vec, then compares and analyzes their performance in the experiment, mainly evaluating the accuracy and test loss of seven models. This paper compares and illustrates the model which gets the different results in experiments, complementing the model and re-running the model, and analyzing the reasons for the difference in the performance of each model. The seven models are a single-layer neural network, multiple-layer (two and three) feed-forward neural networks, Convolutional Neural Network (CNN)- A feedforward neural network, which consists of single or multiple convolutional layers, pooling layers, and a fully connected layer on top, so this model is good at image processing. Long Short Term Memory (LSTM)- A temporal recurrent neural network, the advantage of the model is it could solve the gradient disappearance and explosion problem when it handles the long-sequence problem. Bi-directional Long Short Term Memory (Bi-LSTM)-Composed of forwarding LSTM and backward LSTM, it is very common for sequence labelling tasks that are related to the top and bottom, which are often used to model context information in NLP. Bi-directional Encoder Representation from Transformers (BERT)- A bidirectional language model. Finally, this paper analyses and evaluates these models with a specific illustration and research.

Recognition of fine-grained sow nursing behavior based on the SlowFast and hidden Markov models

Information on sow nursing behavior plays an important role in improving the survival rate of weaned piglets. To obtain more information on sow nursing processing, it is necessary to introduce beginning- and end-stage nursing behavior recognition. This is defined as fine-grained nursing behavior recognition. In this paper, we attempt to automate fine-grained nursing behavior recognition with surveillance video. First, nursing behavior is divided into three subbehaviors: before piglet sucking, piglet sucking and end piglet sucking. Then, a deep neural network model with a dual-stream structure, named SlowFast, is introduced to realize preliminary sow nursing behavior recognition by extracting spatiotemporal features. Finally, for some erroneous results in behavior recognition and the recognition ambiguity in behavior transition, a hidden Markov model (HMM) is introduced to modify the behavioral label sequence by using the Viterbi algorithm. To verify the effectiveness of our method, all experiments are carried out on sow nursing videos collected in an actual breeding environment. The results show that our method achieves a 90.50% sequence coincidence degree for the recognition results on long videos containing the whole nursing behavior, and the accuracy of behavior transition times reaches 87.05%. In summary, this study refines the automatic sow nursing behavior monitoring content to provide more information for precision livestock farming.

Temporal convolutional network for a Fast DNA mutation detection in breast cancer data.

Early detection of breast cancer can be achieved through mutation detection in DNA sequences, which can be acquired through patient blood samples. Mutation detection can be performed using alignment and machine learning techniques. However, alignment techniques require reference sequences, and machine learning techniques still cannot predict index mutation and require supporting tools. Therefore, in this research, a Temporal Convolutional Network (TCN) model was proposed to detect the type and index mutation faster and without reference sequences and supporting tools. The architecture of the proposed TCN model is specifically designed for sequential labeling tasks on DNA sequence data. This allows for the detection of the mutation type of each nucleotide in the sequence, and if the nucleotide has a mutation, the index mutation can be obtained. The proposed model also uses 2-mers and 3-mers mapping techniques to improve detection performance. Based on the tests that have been carried out, the proposed TCN model can achieve the highest F1-score of 0.9443 for COSMIC dataset and 0.9629 for RSCM dataset, Additionally, the proposed TCN model can detect index mutation six times faster than BiLSTM model. Furthermore, the proposed model can detect type and index mutations based on the patient's DNA sequence, without the need for reference sequences or other additional tools.

Extracting social determinants of health from clinical note text with classification and sequence-to-sequence approaches.

Social determinants of health (SDOH) are nonmedical factors that can influence health outcomes. This paper seeks to extract SDOH from clinical texts in the context of the National NLP Clinical Challenges (n2c2) 2022 Track 2 Task. Annotated and unannotated data from the Medical Information Mart for Intensive Care III (MIMIC-III) corpus, the Social History Annotation Corpus, and an in-house corpus were used to develop 2 deep learning models that used classification and sequence-to-sequence (seq2seq) approaches. The seq2seq approach had the highest overall F1 scores in the challenge's 3 subtasks: 0.901 on the extraction subtask, 0.774 on the generalizability subtask, and 0.889 on the learning transfer subtask. Both approaches rely on SDOH event representations that were designed to be compatible with transformer-based pretrained models, with the seq2seq representation supporting an arbitrary number of overlapping and sentence-spanning events. Models with adequate performance could be produced quickly, and the remaining mismatch between representation and task requirements was then addressed in postprocessing. The classification approach used rules to generate entity relationships from its sequence of token labels, while the seq2seq approach used constrained decoding and a constraint solver to recover entity text spans from its sequence of potentially ambiguous tokens. We proposed 2 different approaches to extract SDOH from clinical texts with high accuracy. However, accuracy suffers on text from new healthcare institutions not present in the training data, and thus generalization remains an important topic for future study.

Decoding murine cytomegalovirus.

The genomes of both human cytomegalovirus (HCMV) and murine cytomegalovirus (MCMV) were first sequenced over 20 years ago. Similar to HCMV, the MCMV genome had initially been proposed to harbor ≈170 open reading frames (ORFs). More recently, omics approaches revealed HCMV gene expression to be substantially more complex comprising several hundred viral ORFs. Here, we provide a state-of-the art reannotation of lytic MCMV gene expression based on integrative analysis of a large set of omics data. Our data reveal 365 viral transcription start sites (TiSS) that give rise to 380 and 454 viral transcripts and ORFs, respectively. The latter include >200 small ORFs, some of which represented the most highly expressed viral gene products. By combining TiSS profiling with metabolic RNA labelling and chemical nucleotide conversion sequencing (dSLAM-seq), we provide a detailed picture of the expression kinetics of viral transcription. This not only resulted in the identification of a novel MCMV immediate early transcript encoding the m166.5 ORF, which we termed ie4, but also revealed a group of well-expressed viral transcripts that are induced later than canonical true late genes and contain an initiator element (Inr) but no TATA- or TATT-box in their core promoters. We show that viral upstream ORFs (uORFs) tune gene expression of longer viral ORFs expressed in cis at translational level. Finally, we identify a truncated isoform of the viral NK-cell immune evasin m145 arising from a viral TiSS downstream of the canonical m145 mRNA. Despite being ≈5-fold more abundantly expressed than the canonical m145 protein it was not required for downregulating the NK cell ligand, MULT-I. In summary, our work will pave the way for future mechanistic studies on previously unknown cytomegalovirus gene products in an important virus animal model.

Clinical and radiological description of 120 pediatric stroke-like episodes.

Stroke-like episodes (SLEs) are defined as acute onset of neurological symptoms mimicking a stroke and radiological lesions non-congruent to vascular territory. We aimed to analyze the acute clinical and radiological features of SLEs to determine their pathophysiology. We performed a monocenter retrospective analysis of 120 SLEs in 60 children over a 20-year period. Inclusion criteria were compatible clinical symptoms and stroke-like lesions on brain magnetic resonance imaging (MRI; performed for all 120 events) with focal hyperintensity on diffusion-weighted imaging in a non-vascular territory. Three groups were identified: children with mitochondrial diseases (n = 22) involving mitochondrial DNA mutations (55%) or nuclear DNA mutations (45%); those with other metabolic diseases or epilepsy disorders (n = 22); and those in whom no etiology was found despite extensive investigations (n = 16). Age at first SLE was younger in the group with metabolic or epilepsy disorders (18 months vs. 128 months; p < 0.0001) and an infectious trigger was more frequent (69% vs. 20%; p = 0.0001). Seizures occurred in 75% of episodes, revealing 50% episodes of SLEs and mainly leading to status epilepticus (90%). Of the 120 MRI scans confirming the diagnosis, 28 were performed within a short and strict 48-h period and were further analyzed to better understand the underlying mechanisms. The scans showed primary cortical hyperintensity (n = 28/28) with decreased apparent diffusion coefficient in 52% of cases. Systematic hyperperfusion was found on spin labeling sequences when available (n = 18/18). Clinical and radiological results support the existence of a vicious circle based on two main mechanisms: energy deficit and neuronal hyperexcitability at the origin of SLE.

THT: Transfer Head and Tail for Joint Entity Relation Extraction

Extracting entities and relations from unstructured text is the basic task of building knowledge graphs. Various methods based on sequence labeling have been proposed recently, but most of them use uniform token representations to identify entities. It is clear that the first and last words of the head and tail entities should have different representations. For this reason, we propose a method based on word interaction that provides different representations for the first and last words of entities. Specifically, initial token representation is obtained by BERT. Each token interacts with the next k tokens for the entity’s first token recognition, interacts with the previous k tokens for the entity’s last token recognition, and merges them to obtain the head entity. Then recognize the tail entities based on the head entities and form entity pairs. Finally, get the triples based on a simple multi-classification for relation determination. Extensive experimental results on four datasets show that THT performs better than state-of-the-art baselines.

A Semi-shared Hierarchical Joint Model for Sequence Labeling

A Semi-shared Hierarchical Joint Model for Sequence Labeling

Agreement of Resting Brain Perfusion using Two Different Pseudo-Continuous Arterial Spin Labeling Sequences and the Associations with Lipid Biomarkers

Objective: We developed a modified pseudo-continuous arterial spin labeling (pCASL) sequence for measuring cerebrovascular reactivity (CVR) to hypercapnia in humans; however, the method has yet to be validated against an established pCASL sequence for assessing brain perfusion at rest. The purpose of this project was to compare resting brain perfusion from our modified sequence (Rest-CVR) to the pCASL sequence developed by the Human Connectome Project (HCP). The secondary purpose was to determine the association of these pCASL sequences with lipid biomarkers. Methods: Resting brain perfusion was assessed using both sequences in 7 middle-aged adult participants (56 ± 5 years old) using a 64-channel head coil inside a Siemens Prisma 3T MRI scanner. The modified Rest-CVR sequence included a shorter scan time (~3 minutes) with a single post-labeling delay and duration of 1800ms, FOV= 215mm2, slice thickness 2.27mm, TR=4510ms, and TE=20ms. Blood samples were also collected to determine the association of each sequence with circulating lipid biomarkers to test the sensitivity of each sequence to clinically relevant risk factors of cerebrovascular disease. Results: Outcomes from the two pCASL sequences were significantly correlated [R2 = 0.88; F (1,5) = 35.06 p = 0.002]. However, resting brain perfusion from Rest-CVR was significantly lower than HCP [Rest-CVR: 58.84 ± 15.22 vs. HCP: 81.12 ± 12.54 ml/min/100g; p = 0.0008]. Resting brain perfusion measurements from both sequences were negatively associated with low density lipoprotein (LDL) cholesterol (HCP: r = -0.97; p = 0.0001 and Rest-CVR: r = -0.87; p = 0.008) suggesting sensitivity of both to a clinically relevant risk factor for cerebrovascular disease. However, neither sequence was associated with total cholesterol, triglycerides, or high-density lipoprotein (HDL) cholesterol (p&gt;0.05 for all). Conclusion: Rest-CVR was determined to be a reliable measure of resting brain perfusion in that it correlated with the HCP protocol; however, the lower magnitude relative to HCP makes it a less valid approach for accurately determining brain perfusion. Resting brain perfusion from both sequences was negatively associated with LDL suggesting both are sensitive to clinically relevant outcomes. The shorter Rest-CVR sequence may be most suitable for studies in which an exact estimate of brain perfusion is not required such as when assessing a relative change to hypercapnia. Grant Support: NIH/NIGMS - P20 GM113125 This is the full abstract presented at the American Physiology Summit 2023 meeting and is only available in HTML format. There are no additional versions or additional content available for this abstract. Physiology was not involved in the peer review process.

Uncommon Presentation of an Uncommon Disease: Stroke-like Migraine Attacks after Radiation Therapy in a Patient with Sudden Onset Expressive Aphasia and Cryptogenic Fevers (P3-5.029)

<h3>Objective:</h3> To present a case of stroke-like migraine attacks after radiation therapy (SMART) syndrome accompanied by fever. <h3>Background:</h3> SMART syndrome is an uncommon condition that presents as migraines with focal neurologic deficits in patients who have undergone brain radiation. This syndrome can present decades after completion of radiation therapy and therefore represents a unique diagnostic challenge. <h3>Design/Methods:</h3> Case report. <h3>Results:</h3> A 26-year-old man with a history of myxopapillary ependymoma status post whole brain radiation therapy 15 years prior presented with 3 days of headaches, sudden onset aphasia, and fevers. Initial exam was significant for severe expressive aphasia with only 1–2 words of speech production. Head CT was unremarkable. CT angiography demonstrated slightly decreased arborization of the distal MCA territory vessels in the left cerebral hemisphere but was otherwise unremarkable. Patient was empirically treated for CNS infection with antimicrobial therapy. CSF analysis was unremarkable, as was the remainder of his infectious workup. Long term EEG monitoring showed focal slowing in the left hemisphere but did not show any epileptiform activity. MRI was completed with arterial spin labeling (ASL) sequence, which showed increased perfusion in the left cerebral hemisphere, consistent with SMART syndrome. Patient’s exam gradually improved over the following week. He followed up in clinic 6 weeks later and his symptoms had significantly improved. He was almost back to baseline. Fevers and headaches had completed resolved. <h3>Conclusions:</h3> This case of SMART syndrome has the typical findings of sudden onset migraines and focal neurologic deficits in a patient who previously underwent brain radiation therapy. Our case was unique in that the patient had fevers of unknown origin; a phenomenon that has only been reported in one prior case. The use of MRI perfusion or ASL can aid in the diagnosis; in this case showing increased perfusion over the affected cerebral hemisphere. <b>Disclosure:</b> Dr. Gill has nothing to disclose. Dr. Gill has nothing to disclose. Dr. Avila has received personal compensation in the range of $500-$4,999 for serving on a Scientific Advisory or Data Safety Monitoring board for Alnylam Pharmaceuticals. Dr. Avila has received personal compensation in the range of $10,000-$49,999 for serving on a Speakers Bureau for Alnylam Pharmaceuticals. Dr. Avila has received personal compensation in the range of $500-$4,999 for serving on a Speakers Bureau for Akcea Therapeutics. Gino Mongelluzzo has nothing to disclose.