"Identifying unproven cancer treatments on the health web: addressing accuracy, generalizability and scalability"Aphinyanaphongs, Yin; Fu, Lawrence D; Aliferis, Constantin F
GRANTS:1UL1RR029893/RR/NCRR NIH HHS/United States
Building machine learning models that identify unproven cancer treatments on the Health Web is a promising approach for dealing with the dissemination of false and dangerous information to vulnerable health consumers. Aside from the obvious requirement of accuracy, two issues are of practical importance in deploying these models in real world applications. (a) Generalizability: The models must generalize to all treatments (not just the ones used in the training of the models). (b) Scalability: The models can be applied efficiently to billions of documents on the Health Web. First, we provide methods and related empirical data demonstrating strong accuracy and generalizability. Second, by combining the MapReduce distributed architecture and high dimensionality compression via Markov Boundary feature selection, we show how to scale the application of the models to WWW-scale corpora. The present work provides evidence that (a) a very small subset of unproven cancer treatments is sufficient to build a model to identify unproven treatments on the web; (b) unproven treatments use distinct language to market their claims and this language is learnable; (c) through distributed parallelization and state of the art feature selection, it is possible to prepare the corpora and build and apply models with large scalability..
"Step-based cognitive virtual surgery simulation: an innovative approach to surgical education"Oliker, Aaron; Napier, Zachary; Deluccia, Nicolette; Qualter, John; Sculli, Frank; Smith, Brandon; Stern, Carrie; Flores, Roberto; Hazen, Alexes; McCarthy, Joseph
BioDigital Systems, LLC in collaboration with New York University Langone Medical Center Department of Reconstructive Plastic Surgery has created a complex, real-time, step-based simulation platform for plastic surgery education. These simulators combine live surgical footage, interactive 3D visualization, text labels, and voiceover as well as a high-yield, expert-approved testing mode to create a comprehensive virtual educational environment for the plastic surgery resident or physician..
"The BioDigital Human: A Web-based 3D Platform for Medical Visualization and Education"Qualter, John; Sculli, Frank; Oliker, Aaron; Napier, Zachary; Lee, Sabrina; Garcia, Julio; Frenkel, Sally; Harnik, Victoria; Triola, Marc
NYU School of Medicine's Division of Educational Informatics in collaboration with BioDigital Systems LLC (New York, NY) has created a virtual human body dataset that is being used for visualization, education and training and is accessible over modern web browsers..
"Integration of surgical simulation in plastic surgery residency training"Stern, Carrie; Oliker, Aaron; Napier, Zachary; Qualter, John; Deluccia, Nicolette; Sculli, Frank; Long, Sarah; Rosen, Joe; Hazen, Alexes
BioDigital Systems, LLC in collaboration with New York University Langone Medical Center Department of Reconstructive Plastic Surgery has created an interactive, step-based latissimus musculocutaneous flap simulator. Preliminary testing of fourteen residents (PGY1-6) demonstrates that simulator training results in significant improvement in an objective assessment of surgical knowledge (p < 0.0006, pre-training score: 81.0%, post-training score 92.7%). This study is the first in the field of plastic and reconstructive surgery to demonstrate objective improvement in surgical knowledge as a result of simulator training, suggesting the potential effectiveness of simulators for a panopoly of breast reconstruction options..
"Sociotechnical evaluation of a clinical transformation project in a specialized cancer care centre"Bishop, Margaret; Barnett, Jeff; Vlachaki, Maria T; Pai, Howard
The radiation therapy (RT) department at the British Columbia Cancer Agency - Vancouver Island Centre (VIC) is responsible for delivering radiation treatments to cancer patients from Vancouver Island, which has a population base of approximately 750,000. The purpose of this analysis is to examine a process transformation project undertaken by a VIC clinical champion using a sociotechnical approach and identify factors that influenced the project outcome. Beginning in January 2009, a radiation oncologist at VIC initiated a project to transform the clinical process of generating prescriptions for radiation therapy. The project objective was to replace the paper-based process for radiation therapy (RT) prescriptions with an electronic process to achieve benefits such as increased legibility, accuracy, and accessibility of prescriptions. The electronic prescription (e-Rx) process was designed and developed by health informatics students from the University of Victoria, and the new process was trialed and implemented for approximately half of the new patients seen by the VIC RT department. This pilot implementation was brought to a halt two weeks later, due to concerns raised by the RT department. Using a sociotechnical approach, the authors identify several factors that negatively impacted the project's successful implementation: lack of leadership endorsement and organizational strategy, insufficient formal and informal organizational power of the clinical champion, underestimation of complexity, and inadequate management of the implementation process. Although these factors have been well documented in the literature for large-scale system implementation projects, understanding the way by which they influence smaller-scale process transformation projects in highly specialized clinical settings may help future project managers and coordinators to set such projects up for success..
"The ONCO-I2b2 project: integrating biobank information and clinical data to support translational research in oncology"Segagni, Daniele; Tibollo, Valentina; Dagliati, Arianna; Perinati, Leonardo; Zambelli, Alberto; Priori, Silvia; Bellazzi, Riccardo
The University of Pavia and the IRCCS Fondazione Salvatore Maugeri of Pavia (FSM), has recently started an IT initiative to support clinical research in oncology, called ONCO-i2b2. ONCO-i2b2, funded by the Lombardia region, grounds on the software developed by the Informatics for Integrating Biology and the Bedside (i2b2) NIH project. Using i2b2 and new software modules purposely designed, data coming from multiple sources are integrated and jointly queried. The core of the integration process stands in retrieving and merging data from the biobank management software and from the FSM hospital information system. The integration process is based on a ontology of the problem domain and on open-source software integration modules. A Natural Language Processing module has been implemented, too. This module automatically extracts clinical information of oncology patients from unstructured medical records. The system currently manages more than two thousands patients and will be further implemented and improved in the next two years.
"Economic analysis of centralized vs. decentralized electronic data capture in multi-center clinical studies"Walden, Anita; Nahm, Meredith; Barnett, M Edwina; Conde, Jose G; Dent, Andrew; Fadiel, Ahmed; Perry, Theresa; Tolk, Chris; Tcheng, James E; Eisenstein, Eric L
GRANTS:UL1 TR000436/TR/NCATS NIH HHS/United States
BACKGROUND: New data management models are emerging in multi-center clinical studies. We evaluated the incremental costs associated with decentralized vs. centralized models. METHODS: We developed clinical research network economic models to evaluate three data management models: centralized, decentralized with local software, and decentralized with shared database. Descriptive information from three clinical research studies served as inputs for these models. MAIN OUTCOME MEASURES: The primary outcome was total data management costs. Secondary outcomes included: data management costs for sites, local data centers, and central coordinating centers. RESULTS: Both decentralized models were more costly than the centralized model for each clinical research study: the decentralized with local software model was the most expensive. Decreasing the number of local data centers and case book pages reduced cost differentials between models. CONCLUSION: Decentralized vs. centralized data management in multi-center clinical research studies is associated with increases in data management costs..
"Factors affecting distal end & global decompensation in coronal/sagittal planes 2 years after fusion"Miller, Daniel J; Jameel, Omar; Matsumoto, Hiroko; Hyman, Joshua E; Schwab, Frank J; Roye, David P Jr; Vitale, Michael G
INTRODUCTION: Decompensation of un-fused vertebrae is a potential complication of spinal instrumentation performed for adolescent idiopathic scoliosis (AIS). This can result in problems requiring revision surgery. The purpose of this study was to compare patients who decompensated in the sagittal/coronal plane and those who do not and to identify risk factors. METHODS: The Spinal Deformity Study Group data-base for AIS identified 908 patients at 2 years post-op. Coronal measures analyzed included coronal balance (CB), coronal position of the lower instrumented vertebra (CPL) and LIV tilt angle (LTA). Sagittal measures included sagittal balance (SB) and distal-junctional kyphosis (DJK). The incidence of decompensation at 2 years was: CB-16.83%, LTA-37.53%, CPL-21.17%, negative SB-51.88%, positive SB-7.62%, DJK-6.8%. Decompensated patients were compared to those who were not using preoperative, and 4-16 weeks post-op values. RESULTS: Numerous significant differences were found between patients who decompensated at 2 years and those who did not. CB was significantly influenced by larger height/weight, increased Cobb, preexisting CB and a thoracic LIV. In addition to other factors LTA decompensation was more likely to occur in JIS. CPL was associated with pelvic-obliquity and thoracic LIV. Post-operative sagittal balance could be predicted by pre-operative sagittal balance. DJK was also associated with larger weight and preoperative sagittal measures. DISCUSSION AND CONCLUSION: Less correction in sagittal/coronal planes is a risk factor for decompensation. Curve correction was significant in predicting coronal decompensation. Failure to control sagittal alignment was a risk factor in sagittal decompensation.
"Real-time complex cognitive surgical simulator with testing"Oliker, Aaron; Cutting, Court B
One of the greatest challenges facing surgical education is the inability to effectively test a surgeon's cognitive knowledge of a complex open surgery procedure. Cognitive knowledge is tested by paper, and more recently, computer-based and oral exams. Although these tools are used for testing in surgical education, they have been limited by providing a two-dimensional static representation of complex and dynamic, three-dimensional procedures.A three-dimensional interactive surgical simulator that will engage the surgeon, ask questions, test competency and provide feedback has the potential to revolutionize surgical education. Internet connectivity allows for rapid deployment of surgical modules, networked testing formats, data aggregation, comparative analysis and guided tutorials. Combined with the approval of a surgical society, this platform has the potential to set measurable quantitative surgical standards.
"Visualizing treatment options for breast reconstructive surgery"Qualter, John; Fana, Melissa; Deluccia, Nicolette; Colen, Kari; Scharf, Carrie; Hazen, Alexes
We propose that high-fidelity animations enhanced with real-time 3d interactivity, that demonstrate various breast reconstruction procedures will assist in a patient's decision-making process. These computer based modules will in no way replace a consultation with the physician; instead they will be added to the armamentarium of patient education.
"Local flaps: a real-time finite element based solution to the plastic surgery defect puzzle"Sifakis, Eftychios; Hellrung, Jeffrey; Teran, Joseph; Oliker, Aaron; Cutting, Court
One of the most fundamental challenges in plastic surgery is the alteration of the geometry and topology of the skin. The specific decisions made by the surgeon concerning the size and shape of the tissue to be removed and the subsequent closure of the resulting wound may have a dramatic affect on the quality of life for the patient after the procedure is completed. The plastic surgeon must look at the defect created as an organic puzzle, designing the optimal pattern to close the hole aesthetically and efficiently. In the past, such skills were the distillation of years of hands-on practice on live patients, while relevant reference material was limited to two-dimensional illustrations. Practicing this procedure on a personal computer  has been largely impractical to date, but recent technological advances may come to challenge this limitation. We present a comprehensive real-time virtual surgical environment, based on finite element modeling and simulation of tissue cutting and manipulation. Our system demonstrates the fundamental building blocks of plastic surgery procedures on a localized tissue flap, and provides a proof of concept for larger simulation systems usable in the authoring of complex procedures on elaborate subject geometry.
"The behaviour of fatigue-induced microdamage in compact bone samples from control and ovariectomised sheep"Kennedy, Oran D; Brennan, Orlaith; Mauer, Peter; O'Brien, Fergal J; Rackard, Susan M; Taylor, David; Lee, T Clive
This study investigates the effect of microdamage on bone quality in osteoporosis using an ovariectomised (OVX) sheep model of osteoporosis. Thirty-four sheep were divided into an OVX group (n=16) and a control group (n=18). Fluorochromes were administered intravenously at 3 monthly intervals after surgery to label bone turnover. After sacrifice, beams were removed from the metatarsal and tested in three-point bending. Following failure, microcracks were identified and quantified in terms of region, location and interaction with osteons. Number of cycles to failure (Nf) was lower in the OVX group relative to controls by approximately 7%. Crack density (CrDn) was higher in the OVX group compared to controls. CrDn was 2.5 and 3.5 times greater in the compressive region compared to tensile in control and OVX bone respectively. Combined results from both groups showed that 91% of cracks remained in interstitial bone, approximately 8% of cracks penetrated unlabelled osteons and less than 1% penetrated into labelled osteons. All cases of labelled osteon penetration occurred in controls. Crack surface density (CrSDn), was 25% higher in the control group compared to OVX. It is known that crack behaviour on meeting microstructural features such as osteons will depend on crack length. We have shown that osteon age also affects crack propagation. Long cracks penetrated unlabelled osteons but not labelled ones. Some cracks in the control group did penetrate labelled osteons. This may be due the fact that control bone is more highly mineralized. CrSDn was increased by 25% in the control group compared to OVX. Further study of these fracture mechanisms will help determine the effect of microdamage on bone quality and how this contributes to bone fragility..
"Text categorization models for identifying unproven cancer treatments on the web"Aphinyanaphongs, Yin; Aliferis, Constantin
GRANTS:LM007948-01/LM/NLM NIH HHS/United States;LM007948-02/LM/NLM NIH HHS/United States
The nature of the internet as a non-peer-reviewed (and largely unregulated) publication medium has allowed wide-spread promotion of inaccurate and unproven medical claims in unprecedented scale. Patients with conditions that are not currently fully treatable are particularly susceptible to unproven and dangerous promises about miracle treatments. In extreme cases, fatal adverse outcomes have been documented. Most commonly, the cost is financial, psychological, and delayed application of imperfect but proven scientific modalities. To help protect patients, who may be desperately ill and thus prone to exploitation, we explored the use of machine learning techniques to identify web pages that make unproven claims. This feasibility study shows that the resulting models can identify web pages that make unproven claims in a fully automatic manner, and substantially better than previous web tools and state-of-the-art search engine technology.
"A comparison of impact factor, clinical query filters, and pattern recognition query filters in terms of sensitivity to topic"Fu, Lawrence D; Wang, Lily; Aphinyanagphongs, Yindalon; Aliferis, Constantin F
GRANTS:LM007948-02/LM/NLM NIH HHS/United States;T15 LM 007450-03/LM/NLM NIH HHS/United States
Evaluating journal quality and finding high-quality articles in the biomedical literature are challenging information retrieval tasks. The most widely used method for journal evaluation is impact factor, while novel approaches for finding articles are PubMed's clinical query filters and machine learning-based filter models. The related literature has focused on the average behavior of these methods over all topics. The present study evaluates the variability of these approaches for different topics. We find that impact factor and clinical query filters are unstable for different topics while a topic-specific impact factor and machine learning-based filter models appear more robust. Thus when using the less stable methods for a specific topic, researchers should realize that their performance may diverge from expected average performance. Better yet, the more stable methods should be preferred whenever applicable.
"Learning causal and predictive clinical practice guidelines from data"Mani, Subramani; Aliferis, Constantin; Krishnaswami, Shanthi; Kotchen, Theodore
Clinical practice guidelines (CPG) propose preventive, diagnostic and treatment strategies based on the best available evidence. CPG enable practice of evidencebased medicine and bring about standardization of healthcare delivery in a given hospital, region, country or the whole world. This study explores generation of guidelines from data using machine learning, causal discovery methods and the domain of high blood pressure as an example..
"Modelling collaborative care information--the nursing perspective"Chu, Stephen
Collaborative care has been adopted by many countries as a preferred care delivery model to provide well coordinated holistic care to patients with chronic illnesses as they move between different healthcare sectors during the course of their illnesses. Effective communications and information exchange between collaborative care team members across the health care sectors is considered one of the critical success factors of collaborative care. This paper describes the adoption of a modified Health Level 7 development framework to analyze information exchange requirements, develop the standard clinical data contents and structures for populating the collaborative care messages. It also describes how example nursing data structures are mapped into sample Health Level 7 message segments.
"Learning Boolean queries for article quality filtering"Aphinyanaphongs, Yin; Aliferis, Constantin F
GRANTS:T15 LM 07450-01/LM/NLM NIH HHS/United States
Prior research has shown that Support Vector Machine models have the ability to identify high quality content-specific articles in the domain of internal medicine. These models, though powerful, cannot be used in Boolean search engines nor can the content of the models be verified via human inspection. In this paper, we use decision trees combined with several feature selection methods to generate Boolean query filters for the same domain and task. The resulting trees are generated automatically and exhibit high performance. The trees are understandable, manageable, and able to be validated by humans. The subsequent Boolean queries are sensible and can be readily used as filters by Boolean search engines.
"A novel algorithm for scalable and accurate Bayesian network learning"Brown, Laura E; Tsamardinos, Ioannis; Aliferis, Constantin F
GRANTS:P20 LM 007613-01/LM/NLM NIH HHS/United States;R01 LM 007948-01/LM/NLM NIH HHS/United States;T15 LM 07450-01/LM/NLM NIH HHS/United States
Bayesian Networks (BN) is a knowledge representation formalism that has been proven to be valuable in biomedicine for constructing decision support systems and for generating causal hypotheses from data. Given the emergence of datasets in medicine and biology with thousands of variables and that current algorithms do not scale more than a few hundred variables in practical domains, new efficient and accurate algorithms are needed to learn high quality BNs from data. We present a new algorithm called Max-Min Hill-Climbing (MMHC) that builds upon and improves the Sparse Candidate (SC) algorithm; a state-of-the-art algorithm that scales up to datasets involving hundreds of variables provided the generating networks are sparse. Compared to the SC, on a number of datasets from medicine and biology, (a) MMHC discovers BNs that are structurally closer to the data-generating BN, (b) the discovered networks are more probable given the data, (c) MMHC is computationally more efficient and scalable than SC, and (d) the generating networks are not required to be uniformly sparse nor is the user of MMHC required to guess correctly the network connectivity.
"Cognitive evaluation of an innovative psychiatric clinical knowledge enhancement system"Cohen, Trevor; Kaufman, David; White, Thomas; Segal, Gerald; Staub, Amy Bennett; Patel, Vimla; Finnerty, Molly
Psychiatric Clinical Knowledge Enhancement System (PSYCKES) is an innovative information system that presents patient medication history in tabular and graphical form. The system is designed to support therapeutic decision making. In this paper, we present a multifaceted cognitive evaluation of this system. The evaluation includes a cognitive walkthrough which is a task-analytic method for usability evaluation. We also conducted cognitive studies of two trainee and two attending psychiatrists using the system. One of the attending subjects is presented as a case study. An objective of this research is to characterize the way PSYCKES mediates reasoning. The study found that clinicians were able to use the system effectively to extract and coordinate information and draw appropriate inferences. The expert clinicians were better able to construct a coherent patient representation. The study also documented a few usability problems pertaining to the temporal integration of patient data. PSYCKES is a multifaceted tool that can significantly enhance therapeutic decision making..
"Methods for multi-category cancer diagnosis from gene expression data: a comprehensive evaluation to inform decision support system development"Statnikov, Alexander; Aliferis, Constantin F; Tsamardinos, Ioannis
GRANTS:P20 LM 007613-01/LM/NLM NIH HHS/United States;R01 LM007948-01/LM/NLM NIH HHS/United States
Cancer diagnosis is a major clinical applications area of gene expression microarray technology. We are seeking to develop a system for cancer diagnostic model creation based on microarray data. In order to equip the system with the optimal combination of data modeling methods, we performed a comprehensive evaluation of several major classification algorithms, gene selection methods, and cross-validation designs using 11 datasets spanning 74 diagnostic categories (41 cancer types and 12 normal tissue types). The Multi-Category Support Vector Machine techniques by Crammer and Singer, Weston and Watkins, and one-versus-rest were found to be the best methods and they outperform other learning algorithms such as K-Nearest Neighbors and Neural Networks often to a remarkable degree. Gene selection techniques are shown to significantly improve classification performance. These results guided the development of a software system that fully automates cancer diagnostic model construction with quality on par with or better than previously published results derived by expert human analysts.