DeepPhe-CR: Natural Language Processing Software Services for Cancer Registrar Case Abstraction.

PURPOSE: Manual extraction of case details from patient records for cancer surveillance is a resource-intensive task. Natural Language Processing (NLP) techniques have been proposed for automating the identification of key details in clinical notes. Our goal was to develop NLP application programming interfaces (APIs) for integration into cancer registry data abstraction tools in a computer-assisted abstraction setting. METHODS: We used cancer registry manual abstraction processes to guide the design of DeepPhe-CR, a web-based NLP service API. The coding of key variables was performed through NLP methods validated using established workflows. A container-based implementation of the NLP methods and the supporting infrastructure was developed. Existing registry data abstraction software was modified to include results from DeepPhe-CR. An initial usability study with data registrars provided early validation of the feasibility of the DeepPhe-CR tools. RESULTS: API calls support submission of single documents and summarization of cases across one or more documents. The container-based implementation uses a REST router to handle requests and support a graph database for storing results. NLP modules extract topography, histology, behavior, laterality, and grade at 0.79-1.00 F1 across multiple cancer types (breast, prostate, lung, colorectal, ovary, and pediatric brain) from data of two population-based cancer registries. Usability study participants were able to use the tool effectively and expressed interest in the tool. CONCLUSION: The DeepPhe-CR system provides an architecture for building cancer-specific NLP tools directly into registrar workflows in a computer-assisted abstraction setting. Improved user interactions in client tools may be needed to realize the potential of these approaches.

DeepPhe-CR: Natural Language Processing Software Services for Cancer Registrar Case Abstraction.

Objective: The manual extraction of case details from patient records for cancer surveillance efforts is a resource-intensive task. Natural Language Processing (NLP) techniques have been proposed for automating the identification of key details in clinical notes. Our goal was to develop NLP application programming interfaces (APIs) for integration into cancer registry data abstraction tools in a computer-assisted abstraction setting. Methods: We used cancer registry manual abstraction processes to guide the design of DeepPhe-CR, a web-based NLP service API. The coding of key variables was done through NLP methods validated using established workflows. A container-based implementation including the NLP wasdeveloped. Existing registry data abstraction software was modified to include results from DeepPhe-CR. An initial usability study with data registrars provided early validation of the feasibility of the DeepPhe-CR tools. Results: API calls support submission of single documents and summarization of cases across multiple documents. The container-based implementation uses a REST router to handle requests and support a graph database for storing results. NLP modules extract topography, histology, behavior, laterality, and grade at 0.79-1.00 F1 across common and rare cancer types (breast, prostate, lung, colorectal, ovary and pediatric brain) on data from two cancer registries. Usability study participants were able to use the tool effectively and expressed interest in adopting the tool. Discussion: Our DeepPhe-CR system provides a flexible architecture for building cancer-specific NLP tools directly into registrar workflows in a computer-assisted abstraction setting. Improving user interactions in client tools, may be needed to realize the potential of these approaches. DeepPhe-CR: https://deepphe.github.io/.

Privacy-Preserving Deep Learning NLP Models for Cancer Registries.

Population cancer registries can benefit from Deep Learning (DL) to automatically extract cancer characteristics from the high volume of unstructured pathology text reports they process annually. The success of DL to tackle this and other real-world problems is proportional to the availability of large labeled datasets for model training. Although collaboration among cancer registries is essential to fully exploit the promise of DL, privacy and confidentiality concerns are main obstacles for data sharing across cancer registries. Moreover, DL for natural language processing (NLP) requires sharing a vocabulary dictionary for the embedding layer which may contain patient identifiers. Thus, even distributing the trained models across cancer registries causes a privacy violation issue. In this paper, we propose DL NLP model distribution via privacy-preserving transfer learning approaches without sharing sensitive data. These approaches are used to distribute a multitask convolutional neural network (MT-CNN) NLP model among cancer registries. The model is trained to extract six key cancer characteristics - tumor site, subsite, laterality, behavior, histology, and grade - from cancer pathology reports. Using 410,064 pathology documents from two cancer registries, we compare our proposed approach to conventional transfer learning without privacy-preserving, single-registry models, and a model trained on centrally hosted data. The results show that transfer learning approaches including data sharing and model distribution outperform significantly the single-registry model. In addition, the best performing privacy-preserving model distribution approach achieves statistically indistinguishable average micro- and macro-F1 scores across all extraction tasks (0.823,0.580) as compared to the centralized model (0.827,0.585).

Deep Transfer Learning Across Cancer Registries for Information Extraction from Pathology Reports.

Automated text information extraction from cancer pathology reports is an active area of research to support national cancer surveillance. A well-known challenge is how to develop information extraction tools with robust performance across cancer registries. In this study we investigated whether transfer learning (TL) with a convolutional neural network (CNN) can facilitate cross-registry knowledge sharing. Specifically, we performed a series of experiments to determine whether a CNN trained with single-registry data is capable of transferring knowledge to another registry or whether developing a cross-registry knowledge database produces a more effective and generalizable model. Using data from two cancer registries and primary tumor site and topography as the information extraction task of interest, our study showed that TL results in 6.90% and 17.22% improvement of classification macro F-score over the baseline single-registry models. Detailed analysis illustrated that the observed improvement is evident in the low prevalence classes.

Cervical Spine Alignment in Helmeted Skiers and Snowboarders With Suspected Head and Neck Injuries: Comparison of Lateral C-spine Radiographs Before and After Helmet Removal and Implications for Ski Patrol Transport.

OBJECTIVE: Current protocols for spine immobilization of the injured skier/snowboarder have not been scientifically validated. Observing changes in spine alignment during common rescue scenarios will help strengthen recommendations for rescue guidelines. METHODS: Twenty-eight healthy volunteers (18 men, 10 women) age 47±17 (range 20-73) (mean ±SD with range) underwent a mock rescue in which candidate patrollers completing an Outdoor Emergency Care course performed spine immobilization and back boarding in 3 scenarios: 1) Ski helmet on, no c-collar; 2) helmet on, with c-collar; and 3) helmet removed, with c-collar. After each scenario, a lateral radiograph was taken of the cervical spine to observe for changes in alignment. RESULTS: Compared with the control group (helmet on, no collar), we observed 9 degrees of increased overall (occiput-C7) cervical extension in the helmet on, with collar group (P < .001), and 17 degrees in the helmet off, with collar group (P < .001). There was increased extension at the occiput-C2 intersegment in the helmet on, with collar group (9 degrees, P < .001) and at both the occiput-C2 (9 degrees, P < .001) and C2-C7 (8 degrees, P < .001) intersegments in the helmet off, with collar group. CONCLUSION: Ski helmet removal and c-collar application each leads to increased extension of the cervical spine. In the absence of other clinical factors, our recommendation is that helmets should be left in place and c-collars not routinely applied during ski patrol rescue.

Functional Outcomes and Return to Sports After Acute Repair, Chronic Repair, and Allograft Reconstruction for Proximal Hamstring Ruptures.

BACKGROUND: There are limited data regarding outcomes and return to sports after surgery for acute versus chronic proximal hamstring ruptures. HYPOTHESIS: Surgery for chronic proximal hamstring ruptures leads to improved outcomes and return to sports but at a lower level than with acute repair. Proximal hamstring reconstruction with an Achilles allograft for chronic ruptures is successful when direct repair is not possible. STUDY DESIGN: Cohort study; Level of evidence, 3. METHODS: Between 2002 and 2012, a total of 72 patients with a traumatic proximal hamstring rupture (51 acute, 21 chronic) underwent either direct tendon repair with suture anchors (n = 58) or Achilles allograft tendon reconstruction (n = 14). Results from the Single Assessment Numeric Evaluation (SANE) for activities of daily living (ADL) and sports-related activities, Short Form-12 (SF-12), visual analog scale (VAS), and a patient satisfaction questionnaire were obtained. RESULTS: The mean time to surgery in the chronic group was 441.4 days versus 17.8 days in the acute group. At a mean follow-up of 45 months, patients with chronic tears had inferior sports activity scores (70.2% vs 80.3%, respectively; P = .026) and a trend for decreased ADL scores (86.5% vs 93.3%, respectively; P = .085) compared with those with acute tears. Patients with chronic tears, however, reported significant improvements postoperatively for both sports activity scores (30.3% to 70.2%; P < .01) and ADL scores (56.1% to 86.5%; P < .01). Greater than 5 to 6 cm of retraction in the chronic group was predictive of the need for allograft reconstruction (P = .015) and resulted in ADL and sports activity scores equal to those of chronic repair (P = .507 and P = .904, respectively). There were no significant differences between groups in SF-12, VAS, or patient satisfaction outcomes (mean, 85.2% satisfaction overall). CONCLUSION: Acute repair was superior to chronic surgery with regard to return to sports. Acute and chronic proximal hamstring repair and allograft reconstruction had favorable results for ADL. For low-demand patients or those with medical comorbidities, delayed repair or reconstruction might be considered with an expected 87% return to normal ADL. For patients who desire to return to sports, acute repair is recommended.

Injury patterns at a large Western United States ski resort with and without snowboarders: the Taos experience.

BACKGROUND: Differences in injury patterns among alpine skiers and snowboarders have previously been recognized, and controversy remains about the safety implications that snowboarding may pose to a ski resort. A change of policy at Taos Ski Valley provides a unique and modern perspective on the effect that snowboarders have on ski resort injuries. HYPOTHESIS: The addition of snowboarders to a large ski resort may result in a significant change in both the rate and pattern of injuries treated. STUDY DESIGN: Descriptive epidemiology study. METHODS: Patient records from the Mogul Medical Clinic at Taos Ski Valley were reviewed from the 2006-2007 ski season through the 2009-2010 season (approximately 2 years before and after snowboarding was allowed) and recorded for age, sex, diagnosis, body region, anatomic location, injury type, and sport (skiing, snowboarding). The total numbers of mountain visits for the time periods with and without snowboarding were used to determine injury rates. RESULTS: The overall rate of persons injured increased from 206.7 per 100,000 mountain visits without snowboarders to 233.8 with snowboarders. The relative risk ratio was 1.131, also represented as a 13.1% increased risk of injury (IRI) (statistically significant; 95% CI, 3.5%-23.6%). Increases were seen in the rate of upper extremity injuries (IRI, 39.1%; 95% CI, 14.3-69.4) and head/neck injuries (IRI 30.8%; not significant), while lower extremity and trunk/pelvis injuries remained relatively constant. Distal radius fractures, closed head injuries, and acromioclavicular separations showed statistically significant increases with the addition of snowboarding. The most frequent injuries among snowboarders were distal radius fractures, wrist sprains, closed head injuries, and acromioclavicular separations. The most frequent injuries among skiers were anterior cruciate ligament tears, knee sprains, closed head injuries, and gastrocnemius tears. The median age of injured persons decreased from 39 years (range, 4-100 years) without snowboarders to 31 years (range, 4-99 years) with snowboarders, and this was significant. Approximately 45% of injured persons were female, and this did not change with the addition of snowboarders. CONCLUSION: In this study, there was a small but statistically significant increase in the likelihood of injury with the addition of snowboarding to a large ski resort. It is likely that factors such as younger demographic, elevated risk-taking behavior, or increased mountain crowding are involved. The difference in injuries is largely because of a significant increase in distal radius fractures, closed head injuries, and acromioclavicular separations. On mountain safety precautions such as widening of runs and streamlining of high traffic areas, training medical providers to recognize and treat sport-specific injuries, and promoting the use of wrist guards and helmets may be useful in reducing the effect that snowboarders have on ski resort injuries.

Observing the solar corona with a tunable Fabry-Perot filter.

A solid Fabry-Perot etalon with a 0.16 A passband was used during the 180 s solar eclipse of 2006 for rapid scans of an emission line of the solar corona. The etalon was a Y-cut lithium niobate wafer coated with reflective and conductive (ITO) layers. Voltage applied perpendicular to the etalon face produced a passband shift of 0.0011 A V(-1). During the eclipse, 18 filtergrams were obtained at six 0.22 A steps across the profile of the forbidden [Fe X] spectral emission line at 6374.4 A, which results from the 10(6) K coronal plasma. The 9.3 x 9.3 arcmin field of view showed the structure of the corona above a newly emerged sunspot region. We discuss tests performed on the etalon before and after the eclipse. We also discuss the coronal observations, which show some features with 10 km s(-1) velocities in the line of sight.