Frank Meng

Query formulation from high-level concepts for relational databases

A new query formulation system based on a semantic graph model is presented. The graph provides a semantic model for the data in the database with user-defined relationships. The query formulator allows users to specify their requests and constraints in high-level concepts. The query candidates are formulated based on the user input by a graph search algorithm and ranked according to a probabilistic information measure. English-like query descriptions can also be provided for users to resolve ambiguity when multiple queries are formulated from a user input. For complex queries, we introduce an incremental approach, which assists users to achieve a complex query goal by formulating a series of simple queries. A prototype system with a multimodal interface using the high-level query formulation techniques has been implemented on top of a cooperative database system (CoBase) at UCLA.

Automatic generation of repeated patient information for tailoring clinical notes

Dictating clear, readable, and accurate clinical notes can be a time-consuming task for physicians. Clinical notes often contain information concerning the patients medical history and current medical condition which is propagated from one clinical note to all follow-up clinical notes for the same patient. In this paper, we present a system which, given a clinical note, automatically determines what information should be repeated, and then generates this information for the physician for a new clinical note. We use semantic patterns for capturing the rhetorical category of sentences, which we show to be useful for determining whether the sentence should be repeated. Our system is shown to perform better than a baseline metric based on precision/recall results. Such a system would allow clinical notes to be more complete, timely, and accurate.

A methodology to integrate clinical data for the efficient assessment of brain-tumor patients

Careful examination of the medical record of brain-tumor patients can be an overwhelming task for the neuroradiologist. The number of clinical documents alone may approach 100 for a patient that has a 3-year-old brain tumor. The neuroradiologists evaluation of a patients brain tumor involves examining the current imaging exam and checking for previous imaging exams that may occur pre- or post-treatment. The goal of this research is to develop an effective method to review all of the pertinent patient information from the medical record. We have designed and developed a medical system that incorporates Hospital Information Systems, Radiology Information Systems, and Picture Archiving and Communications Systems information. Our research improves clinical review of patients data by organizing image display, removing unnecessary documents, and mining for key clinical scenarios that are important in the assessment and care of brain-tumor patients.

Automating the generation of lexical patterns for processing free text in clinical documents

OBJECTIVE Many tasks in natural language processing utilize lexical pattern-matching techniques, including information extraction (IE), negation identification, and syntactic parsing. However, it is generally difficult to derive patterns that achieve acceptable levels of recall while also remaining highly precise. MATERIALS AND METHODS We present a multiple sequence alignment (MSA)-based technique that automatically generates patterns, thereby leveraging language usage to determine the context of words that influence a given target. MSAs capture the commonalities among word sequences and are able to reveal areas of linguistic stability and variation. In this way, MSAs provide a systemic approach to generating lexical patterns that are generalizable, which will both increase recall levels and maintain high levels of precision. RESULTS The MSA-generated patterns exhibited consistent F1-, F.5-, and F2- scores compared to two baseline techniques for IE across four different tasks. Both baseline techniques performed well for some tasks and less well for others, but MSA was found to consistently perform at a high level for all four tasks. DISCUSSION The performance of MSA on the four extraction tasks indicates the methods versatility. The results show that the MSA-based patterns are able to handle the extraction of individual data elements as well as relations between two concepts without the need for large amounts of manual intervention. CONCLUSION We presented an MSA-based framework for generating lexical patterns that showed consistently high levels of both performance and recall over four different extraction tasks when compared to baseline methods.

A Bayesian model for estimating multi-state disease progression

A growing number of individuals who are considered at high risk of cancer are now routinely undergoing population screening. However, noted harms such as radiation exposure, overdiagnosis, and overtreatment underscore the need for better temporal models that predict who should be screened and at what frequency. The mean sojourn time (MST), an average duration period when a tumor can be detected by imaging but with no observable clinical symptoms, is a critical variable for formulating screening policy. Estimation of MST has been long studied using continuous Markov model (CMM) with Maximum likelihood estimation (MLE). However, a lot of traditional methods assume no observation error of the imaging data, which is unlikely and can bias the estimation of the MST. In addition, the MLE may not be stably estimated when data is sparse. Addressing these shortcomings, we present a probabilistic modeling approach for periodic cancer screening data. We first model the cancer state transition using a three state CMM model, while simultaneously considering observation error. We then jointly estimate the MST and observation error within a Bayesian framework. We also consider the inclusion of covariates to estimate individualized rates of disease progression. Our approach is demonstrated on participants who underwent chest x-ray screening in the National Lung Screening Trial (NLST) and validated using posterior predictive p-values and Pearsons chi-square test. Our model demonstrates more accurate and sensible estimates of MST in comparison to MLE.

DICOM structured report to track patient's radiation dose to organs from abdominal CT exam

The dramatic increase of diagnostic imaging capabilities over the past decade has contributed to increased radiation exposure to patient populations. Several factors have contributed to the increase in imaging procedures: wider availability of imaging modalities, increase in technical capabilities, rise in demand by patients and clinicians, favorable reimbursement, and lack of guidelines to control utilization. The primary focus of this research is to provide in depth information about radiation doses that patients receive as a result of CT exams, with the initial investigation involving abdominal CT exams. Current dose measurement methods (i.e. CTDIvol Computed Tomography Dose Index) do not provide direct information about a patients organ dose. We have developed a method to determine CTDIvol normalized organ doses using a set of organ specific exponential regression equations. These exponential equations along with measured CTDIvol are used to calculate organ dose estimates from abdominal CT scans for eight different patient models. For each patient, organ dose and CTDIvol were estimated for an abdominal CT scan. We then modified the DICOM Radiation Dose Structured Report (RDSR) to store the pertinent patient information on radiation dose to their abdominal organs.

Automated generation of individually customized visualizations of diagnosis-specific medical information using novel techniques of information extraction

Managing pediatric patients with neurogenic bladder (NGB) involves regular laboratory, imaging, and physiologic testing. Using input from domain experts and current literature, we identified specific data points from these tests to develop the concept of an electronic disease vector for NGB. An information extraction engine was used to extract the desired data elements from free-text and semi-structured documents retrieved from the patient’s medical record. Finally, a Java-based presentation engine created graphical visualizations of the extracted data. After precision, recall, and timing evaluation, we conclude that these tools may enable clinically useful, automatically generated, and diagnosis-specific visualizations of patient data, potentially improving compliance and ultimately, outcomes.

Feasibility of Automating Patient Acuity Measurement Using a Machine Learning Algorithm

Background and Purpose: One method of determining nurse staffing is to match patient demand for nursing care (patient acuity) with available nursing staff. This pilot study explored the feasibility of automating acuity measurement using a machine learning algorithm. Methods: Natural language processing combined with a machine learning algorithm was used to predict acuity levels based on electronic health record data. Results: The algorithm was able to predict acuity relatively well. A main challenge was discordance among nurse raters of acuity in generating a gold standard of acuity before applying the machine learning algorithm. Conclusions: This pilot study tested applying machine learning techniques to acuity measurement and yielded a moderate level of performance. Higher agreement among the gold standard may yield higher performance in future studies.

Automatic Classification of Ultrasound Screening Examinations of the Abdominal Aorta

Our work facilitates the identification of veterans who may be at risk for abdominal aortic aneurysms (AAA) based on the 2007 mandate to screen all veteran patients that meet the screening criteria. The main research objective is to automatically index three clinical conditions: pertinent negative AAA, pertinent positive AAA, and visually unacceptable image exams. We developed and evaluated a ConText-based algorithm with the GATE (General Architecture for Text Engineering) development system to automatically classify 1402 ultrasound radiology reports for AAA screening. Using the results from JAPE (Java Annotation Pattern Engine) transducer rules, we developed a feature vector to classify the radiology reports with a decision table classifier. We found that ConText performed optimally on precision and recall for pertinent negative (0.99 (0.98–0.99), 0.99 (0.99–1.00)) and pertinent positive AAA detection (0.98 (0.95–1.00), 0.97 (0.92–1.00)), and respectably for determination of non-diagnostic image studies (0.85 (0.77–0.91), 0.96 (0.91–0.99)). In addition, our algorithm can determine the AAA size measurements for further characterization of abnormality. We developed and evaluated a regular expression based algorithm using GATE for determining the three contextual conditions: pertinent negative, pertinent positive, and non-diagnostic from radiology reports obtained for evaluating the presence or absence of abdominal aortic aneurysm. ConText performed very well at identifying the contextual features. Our study also discovered contextual trigger terms to detect sub-standard ultrasound image quality. Limitations of performance included unknown dictionary terms, complex sentences, and vague findings that were difficult to classify and properly code.