Scaling-up NLP Pipelines to Process Large Corpora of Clinical Notes

Introduction: This article is part of the Focus Theme of Methods of Information in Medicine on “Big Data and Analytics in Healthcare”.

Objectives: This paper describes the scale-up efforts at the VA Salt Lake City Health Care System to address processing large corpora of clinical notes through a natural language processing (NLP) pipeline. The use case described is a current project focused on detecting the presence of an indwelling uri-nary catheter in hospitalized patients and subsequent catheter-associated urinary tract infections.

Methods: An NLP algorithm using v3NLP was developed to detect the presence of an indwelling urinary catheter in hospitalized patients. The algorithm was tested on a small corpus of notes on patients for whom the presence or absence of a catheter was already known (reference standard). In planning for a scale-up, we estimated that the original algorithm would have taken 2.4 days to run on a larger corpus of notes for this project (550,000 notes), and 27 days for a corpus of 6 million records representative of a national sample of notes. We approached scaling-up NLP pipelines through three techniques: pipeline replication via multi-threading, intra-annotator threading for tasks that can be further decomposed, and remote annotator services which enable annotator scale-out.

Results: The scale-up resulted in reducing the average time to process a record from 206 milliseconds to 17 milliseconds or a 12-fold increase in performance when applied to a corpus of 550,000 notes.

Conclusions: Purposely simplistic in nature, these scale-up efforts are the straight forward evolution from small scale NLP processing to larger scale extraction without incurring associated complexities that are inherited by the use of the underlying UIMA framework. These efforts represent generalizable and widely applicable techniques that will aid other computationally complex NLP pipelines that are of need to be scaled out for processing and analyzing big data.

Residential care homes for the mentally ill

Aims and Method

This study describes residents in seven care homes, reviews their usage of mental health services and evaluates cost implications of psychiatric health care provision.

Results

The patients are predominantly male with multiple diagnoses who are receiving psychiatric health care, but in general lack structured rehabilitation services. Forty-seven per cent of the residents moved into the trust catchment area in order to occupy the placement.The cost associated with the provision of differing models of out-patients care varies considerably.

Clinical Implications

These vulnerable residents are costing the mental health service relatively little, although the total cost to society is higher.This study points to the necessity of multiagency planning for 'new long-stay' patients.

Diagnosis of Mycobacterial infections (Tuberculosis and Paratuberculosis) in tissue samples using molecular (inhouse multiplex PCR, PCR and TaqMan real-time PCR), histopathology and immunohistochemical techniques

An in-house multiplex PCR designed for differential detection of M. bovis, M. avium paratuberculosis (MAP) and M. smegmatis in tissue samples, besides PCR and TaqMan real-time PCR targeting esxA (ESAT-6) and esxB (CFP-10) gene of M. tuberculosis complex (MTC) and IS900 gene targeting MAP were carried out in the present study. Tissue samples comprising of lung, mediastinal lymph node, intestine and mesenteric lymph node from cattle and buffaloes (n=16) suspected to have died of tuberculosis (TB) or paratuberculosis (JD) were collected at post-mortem. Out of 16 animals, only 4 (25%) cases were positive for M. bovis by in-house designed multiplex PCR, where as 9 (56.25%) cases were positive for MTC by both conventional and real-time PCR targeting esxA (ESAT-6) and esxB (CFP-10) genes. Out of 9 TB positive animals, 2 (22.22%) had co-infection with MAP as confirmed by IS900 PCR, real-time PCR and also by in-house multiplex PCR. No samples were found positive for M. smegmatis by the in-house multiplex PCR. Analytical sensitivity of in-house multiplex primers targeting M. bovis, MAP and M. smegmatis were 170 fg/µl, 300 fg/µl and 51 fg/µl of genomic DNA respectively. Analytical sensitivity of primers for both conventional and real time PCR tareting esxA were 8 pg/µl and 800 fg/µl, esxB were 800 fg/µl and 80 fg/µl and IS900 were 30 fg/µl and 3 fg/µl of M. tuberculosis and MAP genomic DNA respectively. The tissue samples were also subjected to histopathology and Immunohistochemical (IHC) staining using ESAT-6 and CFP-10 antibodies for TB and anti-JD anti-serum (raised in rabbit) for JD. All the animals positive for TB (9 of 9) and JD (2 of 2) in PCR and real-time PCR were also positive in histopathology and IHC. Thus, combination of molecular, histopathology and IHC techniques increased the accuracy of TB and JD diagnosis, all of which had same sensitivity and specificity except the in-house multiplex PCR which was comparatively less sensitive in M. bovis detection. However, the molecular methods were found to be more convenient and rapid, capable of making the diagnosis within 6 hours. The study indicated that Real-time PCR was ten times more sensitive than the conventional PCR.

A Rare Association of Congenital Asplenia with Jejunal Arteriovenous Malformation

Patient: Male, 21

Final Diagnosis: AVM intestinal bleeding

Symptoms: Melena

Medication: —

Clinical Procedure: EGD/colonoscopy/enteroscopy

Specialty: Gastroenterology and Hepatology

Synthesis, characterization of nano-sized anatase TiO2 and its adsorption behaviour for environmental contaminant

TiO2 is one of the most studied material due to its unique properties like chemical stability, non toxicity and cost effectiveness.  Nano-sized TiO2   has been prepared by sol -gel method using titanium butoxide as precursor. X- ray diffraction analysis was used to characterize the phase and crystal size of the synthesized TiO2 particles. It was found that sizes of the particles ranged from 6 to 12 nanometre. FT-IR spectrum was also recorded for the TiO2  particles to detect the remaining organic residue .Heavy metals like Cr, Cu, Fe, Ni etc are considered major environmental contaminant in water due to their non biodegradable nature and adverse health effect in living beings. In this work the adsorption behavior of heavy metals towards nano-sized TiO2 was investigated.

Nuclear morphometry and texture analysis on cytological smears of thyroid neoplasms: a study of 50 cases

BACKGROUND

Fine needle aspiration cytology (FNAC) is a reliable and reproducible diagnostic technique for thyroid lesions with certain limitations. Computed morphometric methods have been introduced with a view to improve the diagnostic yield of thyroid aspirates. However, a review of the existing literature revealed conflicting reports regarding morphometric parameters in thyroid neoplasms.

MATERIALS AND METHODS

This study included 50 cases of thyroid lesions (20 cases of colloid goitre, 15 of follicular adenoma, 5 of follicular carcinoma and 10 papillary carcinomas). Digital images of cytologic smears of these cases were captured using a dedicated photomicrography system and nuclear profiles traced manually. With self-designed image analysis software, nuclear morphometric measurements, including texture analysis, were performed. Discriminant analysis was performed including the morphometric parameters and percentage of correctly classified nuclei noted.

RESULTS

Nuclear morphometry parameters showed that papillary thyroid carcinoma had the highest perimeter, area, radius and elongation factor compared to other thyroid lesions. Discriminant analysis revealed that altogether 77.9% of cells could be correctly classified to their lesion category based on the nuclear morphometric and textural parameters. Of the neoplastic cases, 84.5% of cells of follicular neoplasms and 72.5% of papillary carcinoma were classified to the respective category.

CONCLUSION

Nuclear morphometry, including texture analysis, can assist in the cytologic diagnosis of thyroid lesions, considering the high degree of accuracy of classification. Further studies and methodological refinements can achieve higher accuracy.

Abstract P4-03-05: Role of stromal discoidin domain receptor-1 in breast cancer

This abstract was not presented at the symposium.

Comparison of diagnostic tests for the detection of bovine brucellosis in the natural cases of abortion

Rapid and precise diagnosis in natural field cases of bovine abortion caused by Brucella abortus warrants the use of the most sensitive and reliable diagnostic methods. In the present study, bacterial isolation, serology, gross, histopathology, immunohistochemistry and polymerase chain reaction technique(s) were applied. Sero-prevalence studies showed the rate of 28.86% positive cases using the competitive ELISA. Histopathological changes were mainly seen in the placenta, fetal lungs, kidney, liver and spleen. Immunohistochemical (IHC) staining of Brucella spp. was evident as brown, finely granular intracytoplasmic staining in trophoblasts of placental sections and in section(s) of liver, lung, kidney and spleen. Twenty-eight out of the 103 samples (17 from stomach contents, 3 from placental cotyledons, 2 from vaginal discharges and 6 from pooled fetal tissues) produced 193 bp amplicon specific for Brucella genus. Moreover, the species-specific primers amplified a 498 bp amplicon which corresponded to B. abortus. Comparison of diagnostic tests revealed PCR and IHC provide a reliable test for the diagnosis of bovine brucellosis in aborted fetal tissue and placental cotyledons whereas serology is most important for detection of Brucella positive animals in a herd.

Shape Description of Curved Surfaces from Contact Sensing Using Surface Normals

We describe a spline-based framework for incorporating both position and surface-normal information into estimates of curvaturebased shape parameters. Such position and normal data can be obtained with different types of contact sensors, for example, tactile sensors, force-moment sensors, etc. The heart of our framework is an extended B-spline formulation that incorporates the surfacenormal information into a B-spline surface fit of the position data. The surface-normal information provides additional constraints for the surface fit, and can also significantly improve the approximation of the surface. Curvature-based shape parameters applied to this B-spline surface are then used to characterize the local shape of the object surface. Preliminary experiments with simple primitives—spherical, cylindrical, and planar shapes—and an off theshelf force/moment sensor to obtain the position and surfacenormal data show that despite coarse resolution of the sensor, this approach succeeds in qualitative shape recognition.

Validation of Romanowsky staining as a novel screening test for the detection of faecal cryptosporidial oocysts

Cryptosporidiosis is an emerging waterborne protozoan disease and one of the major causes of neonatal diarrhea in humans and animals. But the disease remains under diagnosed due to lack of availability of special stains in majority of laboratories at primary health centers. Therefore, it requires a rapid screening test for routine diagnosis in conventional laboratory set up. In this pursuit, the present study was planned. During this study, fecal samples from 100 representative animals randomly selected from 17 out breaks of bovine calf diarrhea, were stained with modified Ziehl Neelsen staining (mZN) and Leishman's stain to demonstrate cryptosporidial oocysts and for routine fecal examination, respectively. By mZN staining, 25 cases confirmed the presence of cryptosporidial oocysts. However, examination of Leishman's stained fecal smears revealed round hollow unstained bodies resembling cryptosporidia in 20 cases. Therefore, a comparative morphometric analysis was made between the two techniques to determine their relative efficacy in demonstrating cryptosporidia in the feces of affected animals. The analyses showed that the Leishman's stain can be effective in making a presumptive diagnosis of cryptosporidiosis with a little experience. Confirmation of cryptosporidiosis was done by histopathological examination of intestinal sections of calves died during these out breaks. The findings appear to have great clinical value for routine laboratory screening of fecal samples for cryptosporidiosis as conventional Romanowsky stains are readily available and used for multipurpose examination in most of the laboratories at grass root level. Perusal of literature proved this to be the first attempt at easy diagnostics for cryptosporidiosis.