Natural Language Processing Platform Research Articles

Concept Recognition and Characterization of Patients Undergoing Resection of Vestibular Schwannoma Using Natural Language Processing

Abstract Background Natural language processing (NLP), a subset of artificial intelligence (AI), aims to decipher unstructured human language. This study showcases NLP's application in surgical health care, focusing on vestibular schwannoma (VS). By employing an NLP platform, we identify prevalent text concepts in VS patients' electronic health care records (EHRs), creating concept panels covering symptomatology, comorbidities, and management. Through a case study, we illustrate NLP's potential in predicting postoperative cerebrospinal fluid (CSF) leaks. Methods An NLP model analyzed EHRs of surgically managed VS patients from 2008 to 2018 in a single center. The model underwent unsupervised (trained on one million documents from EHR) and supervised (300 documents annotated in duplicate) learning phases, extracting text concepts and generating concept panels related to symptoms, comorbidities, and management. Statistical analysis correlated concept occurrences with postoperative complications, notably CSF leaks. Results Analysis included 292 patients' records, yielding 6,901 unique concepts and 360,929 occurrences. Concept panels highlighted key associations with postoperative CSF leaks, including “antibiotics,” “sepsis,” and “intensive care unit admission.” The NLP model demonstrated high accuracy (precision 0.92, recall 0.96, macro F1 0.93). Conclusion Our NLP model effectively extracted concepts from VS patients' EHRs, facilitating personalized concept panels with diverse applications. NLP shows promise in surgical settings, aiding in early diagnosis, complication prediction, and patient care. Further validation of NLP's predictive capabilities is warranted.

Use of natural language processing to uncover racial bias in obstetrical documentation

Natural Language Processing (NLP), a form of Artificial Intelligence, allows free-text based clinical documentation to be integrated in ways that facilitate data analysis, data interpretation and formation of individualized medical and obstetrical care. In this cross-sectional study, we identified all births during the study period carrying the radiology-confirmed diagnosis of fibroid uterus in pregnancy (defined as size of largest diameter of >5 cm) by using an NLP platform and compared it to non-NLP derived data using ICD10 codes of the same diagnosis. We then compared the two sets of data and stratified documentation gaps by race. Using fibroid uterus in pregnancy as a marker, we found that Black patients were more likely to have the diagnosis entered late into the patient's chart or had missing documentation of the diagnosis.With appropriate algorithm definitions, cross referencing and thorough validation steps, NLP can contribute to identifying areas of documentation gaps and improve quality of care.

Lexicon-grammar tables standardization and implementation

<span>The lexicon-grammar approach is a very important linguistic approach in automatic natural language processing (NLP). It allows for the description of the lexicon of the language through readable and intuitive tables for human manual editing. However, the automatic use of the lexicon-grammar tables in the automatic NLP platforms remains difficult, given the incompatibility between the codes used to represent the properties in the lexicon-grammar tables and those used to represent the properties in the automatic NLP platforms. In this work, we present our method of standardizing the lexicon-grammar tables for the French language, since they constitute very rich lexical, syntactic, and semantic linguistic resources. First, we standardize their properties so that they can be compatible with those used in the NLP platforms. Then, to implement the standardized tables, we used a linguistic platform such as NooJ. For that, we describe the process of integrating these tables into this platform through the automatic generation of the dictionaries from these tables. Finally, to test the efficiency of the generated dictionaries, we create for some of them syntactic grammars that take into account all the grammatical, syntactic, and semantic information contained in the dictionaries.</span>

An Integrated Voice Recognition and Natural Language Processing Platform to Automatically Extract Thoracolumbar Injury Classification Score Features From Radiology Reports

Many predictive models for estimating clinical outcomes after spine surgery have been reported in the literature. However, implementation of predictive scores in practice is limited by the time-intensive nature of manually abstracting relevant predictors. In this study, we designed natural language processing (NLP) algorithms to automate data abstraction for the thoracolumbar injury classification score (TLICS). We retrieved the radiology reports of all Mayo Clinic patients with an International Classification of Diseases, 9th or 10th revision, code corresponding to a fracture of the thoracolumbar spine between January 2005 and October 2020. Annotated data were used to train an N-gram NLP model using machine learning methods, including random forest, stepwise linear discriminant analysis, k-nearest neighbors, and penalized logistic regression models. A total of 1085 spine radiology reports were included in our analysis. Our dataset included 483 compression, 401 burst, 103 translational/rotational, and 98 distraction fractures. A total of 103 reports had documented an injury of the posterior ligamentous complex. The overall accuracy of the random forest model for fracture morphology feature detection was 76.96% versus 65.90% in the stepwise linear discriminant analysis, 50.69% in the k-nearest neighbors, and 62.67% in the penalized logistic regression. The overall accuracy to detect posterior ligamentous complex integrity was highest in the random forest model at 83.41%. Our random forest model was implemented in the backend of a web application in which users can dictate reports and have TLICS features automatically extracted. We have developed a machine learning NLP model for extracting TLICS features from radiology reports, which we deployed in a web application that can be integrated into clinical practice.

Development and evaluation of a program based on a generative pre-trained transformer model from a public natural language processing platform for efficiency enhancement in post-procedural quality control of esophageal endoscopic submucosal dissection.

Post-procedural quality control of endoscopic submucosal dissection (ESD) is emphasized in guidelines. However, this process can be tedious and time-consuming. Recently, a pre-training model called generative pre-trained transformer (GPT) on a public natural language processing platform has emerged and garnered significant attention, whose capabilities align well with the post-procedural quality control process and have the potential to streamline it. Therefore, we developed a simple program utilizing this platform and evaluated its performance. Esophageal ESDs were retrospectively included. The manual quality control process was performed and act as reference standard. GPT's prompt was optimized through multiple iterations. A Python program was developed to automatically submit prompt with pathological report of each ESD procedure and collect quality control information provided by GPT. Its performance on quality control was evaluated with accuracy, precision, recall, and F-1 score. 165 cases were involved into the dataset, of which 5 were utilized as the prompt optimization dataset and 160 as the validation dataset. Definitive prompt was achieved through seven iterations. Time spent on the validation dataset by GPT was 13.47 ± 2.43min. Accuracies of pathological diagnosis, invasion depth, horizontal margin, vertical margin, vascular invasion, and lymphatic invasion of the quality control program were (0.940, 0.952) (95% CI), (0.925, 0.945) (95% CI), 0.931, 1.0, and 1.0, respectively. Precisions were (0.965, 0.969) (95% CI), (0.934, 0.954) (95% CI), and 0.957 for pathological diagnosis, invasion depth, and horizontal margin, respectively. Recalls were (0.940, 0.952) (95% CI), (0.925, 0.945) (95% CI), and 0.931 for factors as mentioned, respectively. F1-score were (0.945, 0.957) (95% CI), (0.928, 0.948) (95% CI), and 0.941 for factors as mentioned, respectively. This quality control program was qualified of post-procedural quality control of esophageal ESDs. GPT can be easily applied to this quality control process and reduce workload of the endoscopists.

Can natural language processing serve as a consultant in oral surgery?

ObjectiveIn this comprehensive evaluation, ten experienced oral surgeon experts posed a total of twenty oral surgery-related questions, including dental implant and tooth extractions, to three distinct Natural Language Processing (NLP)-based chatbot platforms: ChatGPT, Microsoft Bing, and Google Bard. The study aimed to assess the effectiveness of these chatbots in responding to specialized medical questions. Materials and methodsTwo primary evaluation metrics were employed: a Likert Scale (LS) for measuring the accuracy and completeness of responses and a Global Quality Scale (GQS) for evaluating the clarity of responses. Statistical analyses, including one-way analysis of variance (ANOVA) and Post Hoc Tukey, were conducted to assess and compare the performance of the chatbots as rated by the experts. ResultsThe results of the study revealed significant differences in the performance of the chatbots. ChatGPT statistically achieved a better mean LS score of 1.4000±0.15986 than Microsoft Bing (1.8750±0.18143) and Google Bards (2.0500±0.12472) (P < 0.001). Additionally, ChatGPT statistically achieved a higher GQS score of 4.4200±0.30111 than Microsoft Bing (3.7550±0.28621) and Google Bards (3.5250±0.22392) (P < 0.001). ConclusionsThese findings showed the substantial advantage of ChatGPT in effectively addressing oral surgery-related questions with superior accuracy, completeness, and clarity. The study highlights the potential of advanced NLP platforms to enhance information retrieval and communication within the field of oral surgery, reinforcing the utility of such technologies in medical and surgical domains.

Improving Neurology Clinical Care With Natural Language Processing Tools.

The integration of natural language processing (NLP) tools into neurology workflows has the potential to significantly enhance clinical care. However, it is important to address the limitations and risks associated with integrating this new technology. Recent advances in transformer-based NLP algorithms (e.g., GPT, BERT) could augment neurology clinical care by summarizing patient health information, suggesting care options, and assisting research involving large datasets. However, these NLP platforms have potential risks including fabricated facts and data security and substantial barriers for implementation. Although these risks and barriers need to be considered, the benefits for providers, patients, and communities are substantial. With these systems achieving greater functionality and the pace of medical need increasing, integrating these tools into clinical care may prove not only beneficial but necessary. Further investigation is needed to design implementation strategies, mitigate risks, and overcome barriers.

Large scale evaluation of pulmonary nodule workup: A real-world study of over 150,000 patients in New York State.

8559 Background: Around 1.6 million patients in the US annually have a pulmonary nodule (PN) found on imaging. Previous studies suggest 1-12% of PNs represent an underlying malignancy at a potentially curative stage. PNs ≥6mm are considered higher risk. Despite guideline availability, only about one-third of commercially insured patients with a PN receive recommended workup, which may include further investigation via imaging or surgery. In addition, patients and providers may delay workup of PNs identified during COVID-19. We aimed to describe variations in PN workup for different patient demographic and clinical characteristics at 7 healthcare institutions in New York (NY) state. Methods: We used a retrospective cross-sectional design in which structured demographic and unstructured clinical data were used to analyze patient documents from a NY state health information exchange from January 2018 to February 2022. Clinical features were abstracted using a clinical natural language processing platform, CLiX unlock. Results: 58 million documents from 5.5 million unique patients were analyzed. Of the overall 151,436 patients with PNs, workup was found in 47% (71,071). Across institutions, workup ranged between 24% and 52%; 40% (60,749) had evidence of a PN in the upper lung location. We found 52,789 patients (35%) had a maximum PN ≥6mm and of those, 55% (28,958) had a workup (Table). For patients with a PN ≥6mm, workup for non-English speaking patients was 9% lower than for English speaking patients (47% vs 56%), 17% lower for patients 18-44 years than patients 45-90 years (39% vs 56%), 47% lower for patients presenting with a chest injury rather than cancer screening (28% vs 75%), and 9% higher for patients with initial imaging via chest computed tomography (CT) rather than other CT (58% vs 49%). After the start of the COVID-19 pandemic (January 2020), overall workup decreased 3.8% and prevalence of symptoms such as dyspnea and chest pain increased. Conclusions: The descriptive findings of this analysis demonstrate gaps and variations in workup for patients with PNs across different demographics and clinical scenarios. A subsequent project is underway to explore strategies for improvement in adherence to guideline-recommended care to benefit patients by diagnosing lung cancer at an earlier, potentially curative stage. [Table: see text]

Integration of Augmented Reality and Voice Recognition in Learning English for Children

Application development by combining two technologies, namely Augmented Reality and Voice Recognition, can make learning media regarding object recognition at home interact directly with 3D virtual objects. The technology can also help the pronunciation or pronunciation of sentences in English. Natural Language Processing (NLP) is used to understand human language so that machines can understand and process it. This ability supports Voice Recognition to have intelligence and interact like humans. wit.ai is an open-source NLP platform that can support speech-to-text application development. The merging of the two technologies in this development using the wit.ai platform. With the wit.ai platform that is used to understand voice commands and perform tasks as needed for applications regarding object recognition at home, users will be able to interact with objects at home through the given voice commands. In the Black Box testing, each functionality got the results that all the features had functioned properly. User Acceptance Test was also carried out and the average test results were 95.77% and 93.26% on a Likert scale with test results on 13 respondents aged 6-9 years old who have tried the application and 14 respondents as observers when 13 respondents aged 6-9 years tried the application. These results show that the application can be accepted and used as a tool in learning media.

Using Natural Language Processing to Explore Mental Health Insights From UK Tweets During the COVID-19 Pandemic: Infodemiology Study.

There is need to consider the value of soft intelligence, leveraged using accessible natural language processing (NLP) tools, as a source of analyzed evidence to support public health research outputs and decision-making. The aim of this study was to explore the value of soft intelligence analyzed using NLP. As a case study, we selected and used a commercially available NLP platform to identify, collect, and interrogate a large collection of UK tweets relating to mental health during the COVID-19 pandemic. A search strategy comprised of a list of terms related to mental health, COVID-19, and lockdown restrictions was developed to prospectively collate relevant tweets via Twitter's advanced search application programming interface over a 24-week period. We deployed a readily and commercially available NLP platform to explore tweet frequency and sentiment across the United Kingdom and identify key topics of discussion. A series of keyword filters were used to clean the initial data retrieved and also set up to track specific mental health problems. All collated tweets were anonymized. We identified and analyzed 286,902 tweets posted from UK user accounts from July 23, 2020 to January 6, 2021. The average sentiment score was 50%, suggesting overall neutral sentiment across all tweets over the study period. Major fluctuations in volume (between 12,622 and 51,340) and sentiment (between 25% and 49%) appeared to coincide with key changes to any local and/or national social distancing measures. Tweets around mental health were polarizing, discussed with both positive and negative sentiment. Key topics of consistent discussion over the study period included the impact of the pandemic on people's mental health (both positively and negatively), fear and anxiety over lockdowns, and anger and mistrust toward the government. Using an NLP platform, we were able to rapidly mine and analyze emerging health-related insights from UK tweets into how the pandemic may be impacting people's mental health and well-being. This type of real-time analyzed evidence could act as a useful intelligence source that agencies, local leaders, and health care decision makers can potentially draw from, particularly during a health crisis.

170 An Integrated Voice Recognition and Natural Language Processing Platform to Automatically Extract Thoracolumbar Injury Classification Score (TLICS) Features from Radiology Reports

INTRODUCTION: Many predictive models have been reported in the literature for clinical outcomes after spine surgery. However, the real-time implementation of these predictive scores in clinical practice is limited by the time intenstive nature of manually abstracting relevant predictors from the electronic health record. Natural language processing (NLP) provides a framework to automate data abstraction for commonly used predictive scores in spine surgery. METHODS: We retrieved and annotated the radiology reports of all Mayo Clinic patients with an International Classification of Diseases(ICD)-9/10 code corresponding to a fracture of the thoracolumbar spine between January 2005 and October 2020. Annotated data was used to train an n-gram NLP model using machine learning (ML) methods including random forest(RF), stepwise linear discriminant analysis (sLDA), k-nearest neighbors (kNN), and penalized logistic regression (pLR) models. RESULTS: 1,085 spine radiology reports were included in our analysis consististing of 868 to train our ML-NLP model and 217 for validation. Our dataset included 483 compression, 401 burst, 103 translational/rotational, and 98 distraction fractures. 103 reports had documented an injury of the PLC. The overall accuracy of the RF model for fracture morphology feature detection was 76.96% versus 65.90% in the stepwise LDA, 50.69% in the kNN, and 62.67% in the pLR. The overall accuracy to detect PLC integrity was highest in the RF model at 83.41% versus 70.97% in sLDA, 70.05% in kNN, and 67.74% in pLR model. Our RF model was implemented in the backend of a web application where users can dictate reports and have TLICS score features be automatically extracted. CONCLUSION: We have developed a robust ML-NLP model for extracting TLICS score features from radiology reports, which we deployed in a convenient web application that can be integrated prospectively into clinical practice and scaled to other commonly used predictive scores.

Abstract 14253: Combining History Of Present Illness With Risk Stratification Scores To Predict Intensive Care Admission For Suspicious Acute Coronary Syndrome Patients In The Emergency Department

Introduction: Patient symptom data are recorded in free text notes in the electronic health records and are difficult to extract. Natural language processing (NLP) is a technique that can be used to mine key patient symptoms that can be combined with risk stratification tools to improve patient outcomes. Hypothesis: Adding symptom data to the National Early Warning Score (NEWS) and quick sequential organ failure assessment score (qSOFA) will improve the area under the receiver operating curve (AUC) performance. Methods: NEWS scores and qSOFA scores were calculated with assistance from an emergency medicine physician for each patient with suspicion of acute coronary syndrome at initial emergency department (ED) assessment. Independent reviewers annotated the outcome of admission to the intensive care unit as a binary outcome (yes/no). The Apache clinical Text Analysis Knowledge Extraction System (cTAKES), an open access NLP platform, was used to preprocess and run analysis on patient free text ED history of present illness notes. Patient symptom data were extracted and used as patient features/input variables and combined with variables from the NEWS and qSOFA scores. Logistic regression model was used to do prediction. We used DeLong’s test to determine the statistical difference between each scores’ AUC. Results: 70 patients out of 1195 had an intensive care admission. The AUC for the NEWS and qSOFA scores were 0.75 and 0.67, respectively. After adding symptoms features, the AUCs improved to 0.81 and 0.79, respectively. There were statistical differences between the AUCs between each score. Conclusion: Adding patient symptom data improves performance of both the NEWS and qSOFA risk stratification scores. Enhanced risk stratification scores with symptom data could identify those at high risk for clinical deterioration and could help augment accurate decision making at ED disposition.

Abstract 9389: Apache Clinical Text Analysis Knowledge Extraction System (ctakes) Identifies One Out of Five Suspicious Symptoms of Acute Coronary Syndrome from Emergency Department Nursing Triage Notes

Introduction: Acute coronary syndrome is a time sensitive condition requiring rapid identification, but nurses only identify 40%-50% of high-risk patients at emergency department (ED) triage. Symptoms are the driving force behind prompt ED care of patients with ACS. Symptom data are recorded as free text in nursing narratives. Standardized extraction of symptoms via natural language processing will create usable clinical data that could be used in predictive analytics, to improve early identification of ACS and ensure life-saving treatments. Hypothesis: The Apache Clinical Text Analysis and Knowledge Extraction System (cTAKES) will recognize symptom concepts in nursing narratives with an acceptable F1 score of greater than or equal to 0.9. Methods: A nurse scientist and a physician annotated 1200 ED nursing narratives of patients with suspicion of ACS for the following symptoms predictive of ACS: chest pain, chest discomfort, shortness of breath, fatigue and sweating. Inter-rater reliability (Cohen’s kappa statistic) was calculated with 10% overlap. cTAKES, which is an open-source natural language processing platform, was applied to extract evidence of the 5 annotated symptoms. Recall (sensitivity), precision (positive predictive value) and F1 score (harmonic mean of precision and recall) were calculated to assess extraction quality for each symptom. Results: Recall, precision and F1 score for each symptom are as follows: chest pain: 0.71, 0.97, 0.823; chest discomfort: 0.6, 0.9, 0.72; fatigue: 0.9, 0.26, 0.41; shortness of breath: 0.93, 0.96, 0.95; and sweating: 0.54, 1, 0.7, respectively. Cohen’s kappa was 0.9. Conclusions: cTAKES had fair results across all 5 ACS predictive symptoms, with shortness of breath meeting our hypothesized F1 score goal. Failure to recognize symptoms at a clinically acceptable value might be due to the linguistic and content variation present in ED nursing initial narratives. Customized natural language processing pipelines trained specifically for ED nursing narratives have significant potential for accurately identifying symptoms and improving early identification of ACS.

Evaluation of patient burden using online social media in mild cognitive impairment

AbstractBackgroundDetection and diagnosis of Mild Cognitive Impairment due to Alzheimer’s Disease (MCI‐AD) can be challenging due to the subtlety of changes in early stages of disease progression. As a result, opportunities to capture the patient burden in MCI‐AD can be limited.MethodWe used Natural Language Processing (NLP) to analyze challenges and issues that were the most frequently reported online by patients living with MCI‐AD and caregivers. Patients’ self‐reports and caregivers’ reports on patients were identified based on online narratives posted between January 1998 and December 2019 across 84 social media sources. The RLytics NLP platform was used in combination with manual curation to codify verbatim symptoms and impairments against standardized medical taxonomies such as WHO‐ICF and MedDRA, and further into the following categorizations: Social, Physical, Emotional, Cognitive, and Role Activity (SPEC‐R).Result63,933 narratives from 311 patients with MCI‐AD and 1,454 caregivers were qualified into the sample for analysis. The most frequently reported issues varied between patients and caregivers. Cognitive issues (e.g. memory impairments and comprehension‐related) were the most frequently reported by both groups, however, patients reported at a higher rate. This may suggest a deeper concern with losing cognitive ability felt by patients themselves at the disease stage of MCI. This is further supported by a higher reporting rate of anxiety and depressive disorders among patients relative to caregivers. In contrast, certain role activity, physical and social issues (e.g. driving, fatigue, walking and mobility, relationships with spouse or partner) are reported by caregivers at a higher rate than patients, suggesting that these issues may have a greater impact on the daily lives of caregivers or may be more readily recognizable than cognitive issues.ConclusionThe SPEC‐R framework helps identify, understand, and prioritize the most pressing concerns at the symptom and functional level by patients and caregivers. It can uncover areas of disease burden in MCI‐AD that are less frequently associated with MCI, particularly in the non‐cognitive domains, that may not be sufficiently captured by early diagnostic processes in the clinical setting.

Patient treatment and outcome after breast cancer orbital and periorbital metastases: a comprehensive case series including analysis of lobular versus ductal tumor histology

BackgroundBreast cancer is the most common malignancy to spread to the orbit and periorbit, and the invasive lobular carcinoma (ILC) histologic subtype of breast cancer has been reported to form these ophthalmic metastases (OM) more frequently than invasive ductal carcinomas (IDC). We herein report our single academic institution experience with breast cancer OM with respect to anatomical presentation, histology (lobular vs. ductal), treatment, and survival.MethodsWe employed the natural language processing platform, TIES (Text Information Extraction System), to search 2.3 million de-identified patient pathology and radiology records at our institution in order to identify patients with OM secondary to breast cancer. We then compared the resultant cohort, the “OM cohort,” to two other representative metastatic breast cancer patient (MBC) databases from our institution. Histological analysis of selected patients was performed.ResultsOur TIES search and manual refinement ultimately identified 28 patients who were diagnosed with breast cancer between 1995 and 2016 that subsequently developed OM. Median age at diagnosis was 54 (range 28–77) years of age. ER, PR, and HER2 status from the 28 patients with OM did not differ from other patients with MBC from our institution. The relative proportion of patients with ILC was significantly higher in the OM cohort (32.1%) than in other MBC patients in our institution (11.3%, p = 0.007). Median time to first OM in the OM cohort was 46.7 months, and OM were the second most frequent first metastases after bony metastases. After diagnosis of the first distant metastasis of any kind, median survival of patients with ILC (21.4 months) was significantly shorter than that of patients with IDC (55.3 months, p = 0.03). Nine patients developed bilateral OM. We observed a significant co-occurrence of OM and central nervous system metastases (p = 0.0053). The histological analysis revealed an interesting case in which the primary tumor was of a mixed ILC/IDC subtype, while only ILC was present in the OM.ConclusionsOM from breast cancer are illustrative of the difference in metastatic behavior of ILC versus IDC and should be considered when treating patients with ILC, especially in those with complaints of visual acuity changes.

PND83 IDENTIFYING NEUROMYELITIS OPTICA PATIENTS FROM INSURANCE CLAIMS DATA USING NFERX, A NATURAL LANGUAGE PROCESSING-BASED PLATFORM

Neuromyelitis optica spectrum disorder (NMO) is a rare auto-immune disease affecting the central nervous system and has clinical characteristics similar to multiple sclerosis (MS), making clinical diagnosis challenging. This study aims to improve the identification of NMO patients by leveraging nferX, a natural language processing platform, to enable encoding of patient claims that are specific to NMO versus MS. Disease phenotypes associated with NMO or MS were identified from the biomedical literature using the nferX platform. Patients with at least 1 NMO or MS diagnosis code were identified in the IQVIA Pharmetrics Plus database. Accompanying diagnosis, procedure, and drug codes present in greater than 0.1% of patients and at least 2-fold more prevalent in either cohort were mapped to nferX-identified phenotypes. These phenotypes’ relative association with NMO versus MS were quantified using nferX’s adaptation of the pointwise mutual information metric. Associations were aggregated into a summary score for each patient from the number of corresponding claims in the year preceding the initial NMO or MS diagnosis claim. Separation of scores between cohorts of NMO and MS patients identified with existing standard algorithms was assessed. A total of 307,166 patients with at least 1 NMO or MS claim between 2006-2018 were identified. 3,132 NMO patients and 102,611 MS patients were selected using previously published algorithms. The Cohen’s d separation of summary scores between these two cohorts was 1.50, suggesting significant separation. By applying a cutoff of one standard deviation above the mean summary score, 14,822 NMO patients were identified, including 13,150 patients without any claims for NMO. Use of the nferX platform has the potential to improve identification of NMO patients from claims data. Further validation of the approach is needed to support its future use.

MP31-02 IS DIGITAL ETHNOGRAPHY THE FOCUS GROUP OF THE FUTURE? FOCUS GROUPS VS. SOCIAL MEDIA ANALYSIS OF WOMENʼS EXPERIENCE WITH OVERACTIVE BLADDER (OAB)

INTRODUCTION AND OBJECTIVE: Qualitative methods assessing women’s perspectives on living with OAB have traditionally been obtained via focus groups and interviews. Now a plethora of cyber forums have emerged, allowing women to discuss their OAB experiences anonymously. We compared patient perceptions of OAB generated through focus groups to those gleaned from online social media. METHODS: For the digital ethnography analysis, 2,618 posts pertaining to OAB from 203 social media sites were identified by using keywords as search terms in a Java-based natural language processing platform and a social media data mining service. Of these, 200 posts were randomly selected for inclusion. The focus groups included were conducted previously (Anger et al, 2011). Women evaluated in urology clinic for OAB symptoms were recruited to participate in one of five focus groups, totaling 33 patients. Data from both methods were analyzed using grounded theory methodology. The data sets were coded independently and then compared. RESULTS: Each major theme that emerged from focus group data was echoed in the themes uncovered in analysis of social media posts. While major themes were generally similar, several unique subthemes were identified in each group (Table 1). The subthemes unique to the focus groups were centered on symptoms and management. The subthemes unique to the social media group were more personal, reflecting a sense of comfortability with sharing private insights anonymously. A new major theme was identified in the digital ethnography analysis: Online Community Engagement. The subthemes under this theme highlighted that women are pursuing self-education via online resources instead of relying on information from providers. CONCLUSIONS: This study highlights the reliability of qualitative data collected through social media compared to focus groups, as well as social media’s ability to procure personal, unbiased information. Digital ethnography provides immediate access to data on a large, diverse population, obviating the logistics required for focus groups. It presents an opportunity for providers to gain awareness of their patients’ perceptions, as well as what information their patients seek, facilitating a more effective patient-provider relationship.Source of Funding: n/a

Abstract P3-01-12: Breast cancer orbital and periorbital metastases can be bilateral, are associated with invasive lobular histology, and can co-occur with brain metastases

Abstract Breast cancer is responsible for most ophthalmic metastases (OM) to the orbit and periorbit. OM are associated with poor prognosis, as 95% of affected patients subsequently die of breast cancer. Previous studies have demonstrated an increased propensity for invasive lobular carcinoma (ILC) to spread to the eye compared to invasive ductal carcinoma (IDC). ILC is less common than IDC, and it can exhibit an unusual pattern of metastatic spread, including to sites such as the ovaries and GI tract. We herein report our single academic institution experience with OM from breast cancer with respect to anatomical presentation, histology (lobular vs. ductal), treatment and survival. We used the natural language processing platform TIES (Text Information Extraction System) to identify OM caused by primary breast cancer in a database of over 2.3 million patients with electronically stored pathology and radiology reports at the University of Pittsburgh Medical Center. Search terms in TIES were translated into ontologies consisting of the NCI Metathesaurus’ synonyms and abbreviations. The search was thus not reliant on specific words. We then reviewed clinical notes, pathology reports, radiology reports, therapeutic regimens, and outcome data for the cases identified through TIES. Data from the patients identified through TIES was also compared to a large institutional database featuring 1,366 patients with metastatic breast cancer (MBC). Histological slides from 3 patients were analyzed. We identified 28 patients diagnosed with primary breast cancer between 1995 and 2016 and subsequent OM. Median age at diagnosis was 54, with a range of 28 to 77. ER, PR, and HER2neu status from the 28 patients with OM did not differ from other patients with MBC in our institutional database. The relative proportion of patients with ILC was significantly higher in the patient cohort with OM (32.1%) than in the metastatic institutional database (11.3%, p=0.007). Median OS in the OM cohort was 78.4 months; distant metastasis free survival (DMFS) was 34.2 months. These survival times did not differ significantly from those patients in the large institutional metastatic database. DMFS tended to be longer (35.05 months) for patients with ILC compared to IDC (23.34 months), supporting a tendency for late relapse. Additionally, after a diagnosis of first metastasis, median survival of patients with ILC (21.4 months) was significantly shorter than that of patients with IDC (55.2 months) (p=0.03). OM were the second most frequent site of first metastasis in the OM cohort after bony metastases. Median time to first OM was 46.7 months. Of the 9 patients who developed bilateral OM, 4 had ILC, 1 had IDC, 2 had a mixed ILC/IDC, and 2 had an unknown histology. We observed a significant co-occurrence of OM and central nervous system (CNS) metastases (p=0.018). Of 14 patients that developed OM and CNS metastases, only 3 were diagnosed with ILC compared to 9 patients diagnosed with IDC. 57.1% of the patients with OM received radiation therapy to the eye, and 25 patients received at least one line of systemic therapy. The histological analysis revealed an interesting case in which the primary tumor was of a mixed ILC/IDC subtype, while only ILC was present in the OM. To our knowledge, our report of 28 patients is the largest analysis of the histological subtype of breast cancer OM. Through our focus on anatomical presentation, histological subtype, treatment, and survival we provide a broad overview of this rare complication of breast cancer. Our data suggests that OM from breast cancer can often impact both eyes, can be associated with CNS metastases, and are more frequent in patients with ILC than IDC. Citation Format: Martin Blohmer, Li Zhu, Jennifer M Atkinson, Sushil Beriwal, Joshua L. Rodriguez-Lopez, Margaret Rosenzweig, George C Tseng, Peter C Lucas, Adrian V Lee, Steffi Oesterreich, Rachel C Jankowitz. Breast cancer orbital and periorbital metastases can be bilateral, are associated with invasive lobular histology, and can co-occur with brain metastases [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr P3-01-12.

Understanding EFL Linguistic Models through Relationship between Natural Language Processing and Artificial Intelligence Applications

Natural Language Processing (NLP) platforms have recently reported a higher adoption rate of Artificial Intelligence (AI) applications. The purpose of this research is to examine the relationship between NLP and AI in the application of linguistic tasks related to morphology, parsing, and semantics. To achieve this objective, a theoretical framework was designed to investigate the direct and indirect impact of the relationship between NLP platforms and AI applications, such as machine learning and deep learning. Theoretically, this study contributes to examining the relationship between NLP platforms and AI applications through selected linguistic models from the English as a Foreign Language (EFL) perspective. Practical implications are derived from syntactic and semantic variables when AI applications are used. The results of this study suggest that AI applications can use to support NLP tasks, particularly the adaptation of deep learning applications that can prove useful in extracting analytical inferences and enhancing NLP approaches applied to EFL texts. The conclusion drawn of this study is that if NLP caters to knowledge-rich AI techniques, it can make significant advances in the linguistics disciplines of morphology, parsing, and semantics.

Understanding EFL Linguistic Models through Relationship between Natural Language Processing and Artificial Intelligence Applications

Natural Language Processing (NLP) platforms have recently reported a higher adoption rate of Artificial Intelligence (AI) applications. The purpose of this research is to examine the relationship between NLP and AI in the application of linguistic tasks related to morphology, parsing, and semantics. To achieve this objective, a theoretical framework was designed to investigate the direct and indirect impact of the relationship between NLP platforms and AI applications, such as machine learning and deep learning. Theoretically, this study contributes to examining the relationship between NLP platforms and AI applications through selected linguistic models from the English as a Foreign Language (EFL) perspective. Practical implications are derived from syntactic and semantic variables when AI applications are used. The results of this study suggest that AI applications can use to support NLP tasks, particularly the adaptation of deep learning applications that can prove useful in extracting analytical inferences and enhancing NLP approaches applied to EFL texts. The conclusion drawn of this study is that if NLP caters to knowledge-rich AI techniques, it can make significant advances in the linguistics disciplines of morphology, parsing, and semantics.