Characterizing the genetic alterations leading to the more aggressive forms of estrogen receptor positive (ER+) breast cancers are of critical significance in breast cancer management. Here we identify recurrent rearrangements between estrogen receptor gene ESR1 and its neighbor CCDC170, which are enriched in the more aggressive and endocrine-resistant luminal-B tumors, through large-scale analyses of breast cancer transcriptome and copy number alterations. Further screening of 200 ER+ breast cancers identifies eight ESR1-CCDC170 positive tumors. These fusions encode N-terminally truncated CCDC170 proteins (ΔCCDC170). When introduced into ER+ breast cancer cells, ΔCCDC170 leads to markedly increased cell motility and anchorage-independent growth, reduced endocrine sensitivity, and enhanced xenograft tumor formation. Mechanistic studies suggest that ΔCCDC170 engages Gab1 signalosome to potentiate growth factor signaling and enhance cell motility. Together, this study identifies neoplastic ESR1-CCDC170 fusions in a more aggressive subset of ER+ breast cancer, which suggests a new concept of ER pathobiology in breast cancer.
To describe and apply a semi-quantitative MRI scoring system for multi-feature analysis of cartilage defect repair in the knee by osteochondral allografts, and to correlate this scoring system with histopathologic, micro-computed tomography (μCT), and biomechanical reference standards using a goat repair model.
Fourteen adult goats had two osteochondral allografts implanted into each knee: one in the medial femoral condyle (MFC) and one in the lateral trochlea (LT). At 12 months, goats were euthanized and MRI was performed. Two blinded radiologists independently rated nine primary features for each graft, including cartilage signal, fill, edge integration, surface congruity, calcified cartilage integrity, subchondral bone plate congruity, subchondral bone marrow signal, osseous integration, and presence of cystic changes. Four ancillary features of the joint were also evaluated, including opposing cartilage, meniscal tears, synovitis, and fat-pad scarring. Comparison was made with histological and μCT reference standards as well as biomechanical measures. Interobserver agreement and agreement with reference standards was assessed. Cohen’s kappa, Spearman’s correlation, and Kruskal-Wallis tests were used as appropriate.
There was substantial agreement (κ>0.6, p<0.001) for each MRI feature and with comparison against reference standards, except for cartilage edge integration (κ=0.6). There was a strong positive correlation between MRI and reference standard scores (ρ=0.86, p<0.01). OCAMRISS was sensitive to differences in outcomes between the types of allografts.
We have described a comprehensive MRI scoring system for osteochondral allografts and have validated this scoring system with histopathologic and μCT reference standards as well as biomechanical indentation testing.
cartilage repair; osteochondral allografts; MRI scoring system
Stroke remains a major cause of disability. The cost of stroke rehabilitation is substantial. Understanding the factors that predict response to inpatient stroke rehabilitation may be useful, for example, to best individualize the content of therapy, or to maximize the efficiency with which resources are directed. This review reviewed the literature and found that numerous variables were associated with outcome after inpatient stroke rehabilitation. The strongest evidence exists for factors such as age, stroke subtype, nutritional status, psychosocial factors such as living with family prior to stroke or presence of a caregiver. Functional status on admission, urinary incontinence, post-stroke infection, and aphasia each can also impact prognosis. Strengths and weaknesses of cited studies are considered in an attempt to inform design of future studies examining the factors that predict response to inpatient rehabilitation after stroke.
stroke rehabilitation predict
MR microneurography is a noninvasive technique that provides visualization of the microanatomy of peripheral nerves, otherwise available only with histopathology. The objective of this study was to present a protocol to visualize the microstructure of peripheral nerves in vivo, using a 3T MRI scanner with a clinical set of coils and sequences. The tibial and the common peroneal nerves of healthy volunteers were imaged above the medial malleolus and at the level of the fibular head, respectively. The acquired images provided details about the internal structure of peripheral nerves, with visualization of the fascicles, the interfascicular fat, the epineurium, and the perineurium. MR microneurography can be performed in a clinical setting with acceptable imaging times and can be a potentially powerful tool that complements standard MR neurography.
References to anatomical entities in medical records consist not only of explicit references to anatomical locations, but also other diverse types of expressions, such as specific diseases, clinical tests, clinical treatments, which constitute implicit references to anatomical entities. In order to identify these implicit anatomical entities, we propose a hierarchical framework, in which two layers of named entity recognizers (NERs) work in a cooperative manner. Each of the NERs is implemented using the Conditional Random Fields (CRF) model, which use a range of external resources to generate features. We constructed a dictionary of anatomical entity expressions by exploiting four existing resources, i.e., UMLS, MeSH, RadLex and BodyPart3D, and supplemented information from two external knowledge bases, i.e., Wikipedia and WordNet, to improve inference of anatomical entities from implicit expressions. Experiments conducted on 300 discharge summaries showed a micro-averaged performance of 0.8509 Precision, 0.7796 Recall and 0.8137 F1 for explicit anatomical entity recognition, and 0.8695 Precision, 0.6893 Recall and 0.7690 F1 for implicit anatomical entity recognition. The use of the hierarchical framework, which combines the recognition of named entities of various types (diseases, clinical tests, treatments) with information embedded in external knowledge bases, resulted in a 5.08% increment in F1. The resources constructed for this research will be made publicly available.
Intensity modulated proton therapy (IMPT) plans are normally generated utilizing multiple field optimization (MFO) techniques. Similar to photon based IMRT, MFO allows for the utilization of a simultaneous integrated boost in which multiple target volumes are treated to discrete doses simultaneously, potentially improving plan quality and streamlining quality assurance and treatment delivery. However, MFO may render plans more sensitive to the physical uncertainties inherent to particle therapy. Here we present clinical examples of a single-field integrated boost (SFIB) technique for spot scanning proton therapy based on single field optimization (SFO) treatment-planning techniques.
Methods and materials
We designed plans of each type for illustrative patients with central nervous system (brain and spine), prostate and head and neck malignancies. SFIB and IMPT plans were constructed to deliver multiple prescription dose levels to multiple targets using SFO or MFO, respectively. Dose and fractionation schemes were based on the current clinical practice using X-ray IMRT in our clinic. For inverse planning, dose constraints were employed to achieve the desired target coverage and normal tissue sparing. Conformality and inhomogeneity indices were calculated to quantify plan quality. We also compared the worst-case robustness of the SFIB, sequential boost SFUD, and IMPT plans.
The SFIB technique produced more conformal dose distributions than plans generated by sequential boost using a SFUD technique (conformality index for prescription isodose levels; 0.585 ± 0.30 vs. 0.435 ± 0.24, SFIB vs. SFUD respectively, Wilcoxon matched-pair signed rank test, p < 0.01). There was no difference in the conformality index between SFIB and IMPT plans (0.638 ± 0.27 vs. 0.633 ± 0.26, SFIB vs. IMPT, respectively). Heterogeneity between techniques was not significantly different. With respect to clinical metrics, SFIB plans proved more robust than the corresponding IMPT plans.
SFIB technique for scanning beam proton therapy (SSPT) is now routinely employed in our clinic. The SFIB technique is a natural application of SFO and offers several advantages over SFUD, including more conformal plans, seamless treatment delivery and more efficient planning and QA. SFIB may be more robust than IMPT and has been the treatment planning technique of choice for some patients.
Proton therapy; Spot scanning; Single-field optimization; Single field integrated boost; SFIB
Cortical bone is typically regarded as “MR invisible” with conventional clinical magnetic resonance imaging (MRI) pulse sequences. However, recent studies have demonstrated that free water in the microscopic pores of cortical bone has a short T2* but a relatively long T2, and may be detectable with conventional clinical spin echo (SE) or fast spin echo (FSE) sequences. In this study we describe the use of a conventional two-dimensional (2D) FSE sequence to assess cortical bone microstructure and measure cortical porosity using a clinical 3T scanner. Twelve cadaveric human cortical bone samples were studied with MRI and micro computed tomography (μCT) (downsampled to the same spatial resolution). Preliminary results show that FSE-determined porosity is highly correlated (R2 = 0.83; P < 0.0001) with μCT porosity. Bland Altman analysis suggested a good agreement between FSE and μCT with tight limit of agreement at around 3%. There is also a small bias of -2% for the FSE data, which suggested that the FSE approach slightly underestimated μCT porosity. The results demonstrate that cortical porosity can be directly assessed using conventional clinical FSE sequences. The clinical feasibility of this approach was also demonstrated on six healthy volunteers using 2D FSE sequences as well as 2D ultrashort echo time (UTE) sequences with a minimal echo time (TE) of 8 μs, which provide high contrast imaging of cortical bone in vivo.
Fast spin echo; UTE; porosity; cortical bone; μCT
To create an end-to-end system to identify temporal relation in discharge summaries for the 2012 i2b2 challenge. The challenge includes event extraction, timex extraction, and temporal relation identification.
An end-to-end temporal relation system was developed. It includes three subsystems: an event extraction system (conditional random fields (CRF) name entity extraction and their corresponding attribute classifiers), a temporal extraction system (CRF name entity extraction, their corresponding attribute classifiers, and context-free grammar based normalization system), and a temporal relation system (10 multi-support vector machine (SVM) classifiers and a Markov logic networks inference system) using labeled sequential pattern mining, syntactic structures based on parse trees, and results from a coordination classifier. Micro-averaged precision (P), recall (R), averaged P&R (P&R), and F measure (F) were used to evaluate results.
For event extraction, the system achieved 0.9415 (P), 0.8930 (R), 0.9166 (P&R), and 0.9166 (F). The accuracies of their type, polarity, and modality were 0.8574, 0.8585, and 0.8560, respectively. For timex extraction, the system achieved 0.8818, 0.9489, 0.9141, and 0.9141, respectively. The accuracies of their type, value, and modifier were 0.8929, 0.7170, and 0.8907, respectively. For temporal relation, the system achieved 0.6589, 0.7129, 0.6767, and 0.6849, respectively. For end-to-end temporal relation, it achieved 0.5904, 0.5944, 0.5921, and 0.5924, respectively. With the F measure used for evaluation, we were ranked first out of 14 competing teams (event extraction), first out of 14 teams (timex extraction), third out of 12 teams (temporal relation), and second out of seven teams (end-to-end temporal relation).
The system achieved encouraging results, demonstrating the feasibility of the tasks defined by the i2b2 organizers. The experiment result demonstrates that both global and local information is useful in the 2012 challenge.
Context-Free Grammar; Syntax; Labeled Sequential Pattern; Dependency Tree; Coordination
There are close to 70,000 new cases of primary central nervous system tumors diagnosed annually in the United States. Meningiomas, gliomas, nerve sheath tumors and pituitary tumors account for 85% of them. There is abundant literature on these commonly occurring tumors but data from the literature on infrequently encountered tumors such as atypical teratoid/rhabdoid tumor, choroid plexus carcinoma, ganglioglioma, hemangiopericytoma, and pleomorphic xanthoastrocytoma are limited. This review provides an overview of the clinicopathologic and therapeutic aspects of these rare primary central nervous system tumors.
atypical teratoid/rhabdoid tumor; choroid plexus carcinoma; ganglioglioma; hemangiopericytoma; pleomorphic xanthoastrocytoma
Ras can act on the plasma membrane (PM) to mediate extracellular signaling and tumorigenesis. To identify key components controlling Ras PM localization, we performed an unbiased screen to seek Schizosaccharomyces pombe mutants with reduced PM Ras. Five mutants were found with mutations affecting the same gene, S. pombe
erf2 (sp-erf2), encoding sp-Erf2, a palmitoyltransferase, with various activities. sp-Erf2 localizes to the trans-Golgi compartment, a process which is mediated by its third transmembrane domain and the Erf4 cofactor. In fission yeast, the human ortholog zDHHC9 rescues the phenotypes of sp-erf2 null cells. In contrast, expressing zDHHC14, another sp-Erf2-like human protein, did not rescue Ras1 mislocalization in these cells. Importantly, ZDHHC9 is widely overexpressed in cancers. Overexpressing ZDHHC9 promotes, while repressing it diminishes, Ras PM localization and transformation of mammalian cells. These data strongly demonstrate that sp-Erf2/zDHHC9 palmitoylates Ras proteins in a highly selective manner in the trans-Golgi compartment to facilitate PM targeting via the trans-Golgi network, a role that is most certainly critical for Ras-driven tumorigenesis.
To develop ultrashort echo time (UTE) magnetic resonance imaging (MRI) techniques to image the zone of calcified cartilage (ZCC), and quantify its T2*, T1 and T1ρ.
In this feasibility study a dual inversion recovery ultrashort echo time (DIR-UTE) sequence was developed for high contrast imaging of the ZCC. T2* of the ZCC was measured with DIR-UTE acquisitions at progressively increasing TEs. T1 of the ZCC was measured with saturation recovery UTE acquisitions at progressively increasing saturation recovery times. T1ρ of the ZCC was measured with spin-locking prepared DIR-UTE acquisitions at progressively increasing spin-locking times.
The feasibility of the qualitative and quantitative DIR-UTE techniques was demonstrated on phantoms and in six cadaveric patellae using a clinical 3T scanner. On average the ZCC has a short T2* ranging from 1.0 to 3.3 ms (mean ± standard deviation = 2.0 ± 1.2 ms), a short T1 ranging from 256 to 389 ms (mean ± standard deviation = 305 ± 45 ms), and a short T1ρ ranging from 2.2 to 4.6 ms (mean ± standard deviation = 3.6 ± 1.2 ms).
UTE MR based techniques have been developed for high resolution imaging of the ZCC and quantitative evaluation of its T2*, T1 and T1ρ relaxation times, providing noninvasive assessment of collagen orientation and proteoglycan content at the zone of calcified cartilage and the bone cartilage interface. These measurements may be useful for non-invasive assessment of the ZCC, including understanding the involvement of this tissue component in osteoarthritis.
Ultrashort echo time; adiabatic inversion recovery; dual inversion recovery; T1ρ; ZCC
Desmoplastic small round cell tumor (DSCRT) is an uncommon pediatric tumor with a poor prognosis. Aggressive multimodality therapy is the current treatment approach, however treatment toxicity is of concern. We report our results with whole abdominopelvic intensity-modulated radiation therapy (WAP-IMRT) as a component of multimodality therapy for DSCRT at a single institution.
Medical records of all patients with DSCRT who received WAP-IMRT as part of definitive treatment at MD Anderson (2006-2010) were identified and reviewed.
Eight patients with DSRCT received WAP-IMRT with a median follow-up of 15.2 months. All patients received multiple courses of chemotherapy followed by surgical debulking of intra-abdominal disease; seven also had intraoperative hyperthermic cisplatin. WAP-IMRT was delivered to a total dose of 30 Gy post-operatively; four patients received a simultaneous boost (6-10 Gy) to sites of gross residual disease. Seven patients received concurrent chemotherapy during WAP-IMRT. No RTOG grade 4 nausea, vomiting, or diarrhea occurred during RT. Red cell transfusions were given to two patients to maintain hemoglobin levels of greater than 10 g/dL. Grade 4 cytopenia requiring growth factor support occurred in only one patient; no other significant cytopenias were noted. WAP-IMRT resulted in 25% lower radiation doses to the lumbosacral vertebral bodies and pelvic bones than conventional RT plans. The median time to local or distant failure after WAP-IMRT was 8.73 months in seven patients. One patient who had completed RT 20 months before the last follow-up remains alive without evidence of disease. Five patients (63%) experienced treatment failure in the abdomen. Distant failure occurred in three patients (37.5%).
WAP-IMRT with concurrent radiosensitizing chemotherapy was well tolerated after aggressive surgery for DSCRT. Enhanced bone sparing with IMRT probably accounts for the low hematologic toxicity (vs. conventional WAP RT). This modality should be considered as an additional local-regional control option for DSRCT.
Desmoplastic small round-cell tumor (DSRCT); whole abdominopelvic radiotherapy; pediatric cancer; sarcoma; peritoneal sarcomatosis; IMRT
Peripheral nerve injury induces upregulation of the calcium channel alpha-2-delta-1 proteins in the dorsal root ganglia and dorsal spinal cord that correlates with neuropathic pain development. Similar behavioral hypersensitivity was also observed in injury-free transgenic mice (TG) over-expressing the alpha-2-delta-1 proteins in neuronal tissues. To investigate pathways regulating alpha-2-delta-1 protein-mediated behavioral hypersensitivity, we examined whether spinal serotonergic 5-HT3 receptors are involved similarly in the modulation of behavioral hypersensitivity induced by either peripheral nerve injury in a nerve injury model or neuronal alpha-2-delta-1 over-expression in the TG model.
The effects of blocking behavioral hypersensitivity in these two models by intrathecal or systemic injections of 5-HT3 receptor antagonist, ondansetron, were compared.
Our data indicated that the TG mice displayed similar behavioral hypersensitivities to non-painful mechanical stimulation (tactile allodynia) and painful thermal stimulation (thermal hyperalgesia) as that observed in the nerve injury model. Interestingly, tactile allodynia and thermal hyperalgesia in both models can be blocked similarly by intrathecal, but not systemic, injection of ondansetron.
Our data suggest that spinal 5-HT3 receptors are likely play a role in alpha-2-delta-1-mediated behavioral hypersensitivities through a descending serotonergic facilitation.
Brain metastases present a therapeutic challenge because patients with metastatic cancers live longer now than in the recent past due to systemic therapies that, while effective, may not penetrate the blood-brain barrier. In the present study the authors sought to validate the Diagnosis-Specific Graded Prognostic Assessment (DS-GPA), a new prognostic index that takes into account the histological characteristics of the primary tumor, and the Radiation Therapy Ontology Group Recursive Partitioning Analysis (RPA) system by using a single-institution database of patients who were treated initially with stereotactic radiosurgery (SRS) alone for brain metastases.
Investigators retrospectively identified adult patients who had undergone SRS at a single institution, MD Anderson Cancer Center, for initial treatment of brain metastases between 2003 and 2010 but excluded those who had undergone craniotomy and/or whole-brain radiation therapy at an earlier time; the final number was 251. The Leksell Gamma Knife was used to treat 223 patients, and a linear accelerator was used to treat 28 patients. The patient population was grouped according to DS-GPA scores as follows: 0–0.5 (7 patients), 1 (33 patients), 1.5 (25 patients), 2 (63 patients), 2.5 (14 patients), 3 (68 patients), and 3.5–4 (41 patients). The same patients were also grouped according to RPA classes: 1 (24 patients), 2 (216 patients), and 3 (11 patients). The most common histological diagnoses were non–small cell lung cancer (34%), melanoma (29%), and breast carcinoma (16%). The median number of lesions was 2 (range 1–9) and the median total tumor volume was 0.9 cm3 (range 0.3–22.9 cm3). The median radiation dose was 20 Gy (range 14–24 Gy). Stereotactic radiosurgery was performed as the sole treatment (62% of patients) or combined with a salvage treatment consisting of SRS (22%), whole-brain radiation therapy (12%), or resection (4%). The median duration of follow-up was 9.4 months.
In this patient group the median overall survival was 11.1 months. The DS-GPA prognostic index divided patients into prognostically significant groups. Median survival times were 2.8 months for DS-GPA Scores 0–0.5, 3.9 months for Score 1, 6.6 months for Score 1.5, 12.9 months for Score 2, 11.9 months for Score 2.5, 12.2 months for Score 3, and 31.4 months for Scores 3.5–4 (p < 0.0001). In the RPA groups, the median overall survival times were 38.8 months for Class 1, 9.4 months for Class 2, and 2.8 months for Class 3 (p < 0.0001). Neither the RPA class nor the DS-GPA score was prognostic for local tumor control or new lesion–free survival. A multivariate analysis revealed that patient age > 60 years, Karnofsky Performance Scale score ≤ 80%, and total lesion volume > 2 cm3 were significant adverse prognostic factors for overall survival.
Application of the DS-GPA to a database of patients with brain metastases who were treated with SRS appears to be valid and offers additional prognostic refinement over that provided by the RPA. The DS-GPA may also allow for improved selection of patients to undergo initial SRS alone and should be studied further.
stereotactic radiosurgery; Gamma Knife surgery; brain metastasis; prognostic factor; Diagnosis-Specific Graded Prognostic Assessment; Recursive Partitioning Analysis
Objective. To present a unique case report of a Lisfranc fracture in a patient with a bipartite medial cuneiform and to evaluate the prevalence of the bipartite medial cuneiform in a retrospective review of 1000 magnetic resonance (MR) imaging studies of the foot. Materials and Methods. Case report followed by a retrospective review of 1000 MR imaging studies of the foot for the presence or absence of a bipartite medial cuneiform. Results. The incidence of the bipartite medial cuneiform is 0.1%. Conclusion. A bipartite medial cuneiform is a rare finding but one with both clinical and surgical implications.
This Phase I/II study tests the hypothesis that single-fraction SBRT for previously un-irradiated spinal metastases is a safe, feasible, and efficacious treatment approach.
All patients were evaluated by a multidisciplinary team. Spinal MRI was performed before treatment and at regular intervals to both define target volume and response to treatment. SBRT was delivered to a peripheral dose of 16–24 Gy in 1 fraction while limiting dose to the spinal cord. Higher doses were used for renal cell histology. The NCI Common Toxicity Criteria 2.0 and McCormick neurological function score were used as toxicity assessment tools.
A total of 61 patients harboring 63 tumors of the non-cervical spine were enrolled and treated with SBRT between 2005 and 2010 on a prospective Phase I/II trial at the University of Texas M. D. Anderson Cancer Center. Mean follow-up was 20 months. Actuarial 18-month imaging local control for all patients was 88%. Actuarial 18-month overall survival for all patients was 64%. Median survival for all patients was 30 months. No significant differences in outcomes were noted with respect to tumor histology and SBRT dose. Two patients experienced radiation adverse events (Grade 3 or higher). Actuarial 18-month freedom from neurologic deterioration from any cause as was 82%.
This Phase I/II data support an expanded indication for SBRT as first-line treatment of selected spinal metastases patients. Additional studies that can prospectively identify predictive factors for spinal cord toxicity after SBRT are warranted to minimize the incidence of this serious yet rare complication.
radiation; stereotactic body radiotherapy; radiosurgery; spinal metastases
Cortical dysplasias (CDs) are highly epileptogenic lesions with a good prognosis of seizure freedom, if totally resected. However, their accurate delineation and resection can be difficult, and depend on the extent of pathology and lesion location. Intraoperative neurophysiologic assessments are valuable in these situations. We present an illustrative case of intractable epilepsy where judicious use of intraoperative neurophysiologic–techniques guided resection of precentral CD, under general anesthesia and in the absence of preoperative electrophysiologic mapping data. Ictal onset was accurately delineated using electrocorticography (ECoG). Phase reversal of the median somatosensory-evoked potentials (MSSEPs) localized the central sulcus (CS). Motor evoked potentials (MEPs) triggered by high-frequency monopolar anodal electrical cortical stimulation at the primary motor cortex (PMC) threshold delineated the PMC. Using this technique, PMC and the corticospinal tract (CST) were continuously monitored during resection. No changes in MEPs from the pre-resection baseline were seen; no residual abnormal activity was present in the postresection ECoG. The patient emerged from surgery without deficits and has been seizure free during a 10-month follow-up. Staged multimodal intraoperative neurophysiology can be used successfully under general anesthesia to guide resection of epileptogenic lesions within the precentral gyrus, as an add-on or, in certain situations, as a viable alternative to preoperative electrophysiologic mapping.
Electrocorticography; Motor mapping; Cortical dysplasia; Central sulcus
To create a highly accurate coreference system in discharge summaries for the 2011 i2b2 challenge. The coreference categories include Person, Problem, Treatment, and Test.
An integrated coreference resolution system was developed by exploiting Person attributes, contextual semantic clues, and world knowledge. It includes three subsystems: Person coreference system based on three Person attributes, Problem/Treatment/Test system based on numerous contextual semantic extractors and world knowledge, and Pronoun system based on a multi-class support vector machine classifier. The three Person attributes are patient, relative and hospital personnel. Contextual semantic extractors include anatomy, position, medication, indicator, temporal, spatial, section, modifier, equipment, operation, and assertion. The world knowledge is extracted from external resources such as Wikipedia.
Micro-averaged precision, recall and F-measure in MUC, BCubed and CEAF were used to evaluate results.
The system achieved an overall micro-averaged precision, recall and F-measure of 0.906, 0.925, and 0.915, respectively, on test data (from four hospitals) released by the challenge organizers. It achieved a precision, recall and F-measure of 0.905, 0.920 and 0.913, respectively, on test data without Pittsburgh data. We ranked the first out of 20 competing teams. Among the four sub-tasks on Person, Problem, Treatment, and Test, the highest F-measure was seen for Person coreference.
This system achieved encouraging results. The Person system can determine whether personal pronouns and proper names are coreferent or not. The Problem/Treatment/Test system benefits from both world knowledge in evaluating the similarity of two mentions and contextual semantic extractors in identifying semantic clues. The Pronoun system can automatically detect whether a Pronoun mention is coreferent to that of the other four types. This study demonstrates that it is feasible to accomplish the coreference task in discharge summaries.
Natural language processing; information retrieval; clinical decision support; biomedical informatics; text processing; medical records
A system that translates narrative text in the medical domain into structured representation is in great demand. The system performs three sub-tasks: concept extraction, assertion classification, and relation identification.
The overall system consists of five steps: (1) pre-processing sentences, (2) marking noun phrases (NPs) and adjective phrases (APs), (3) extracting concepts that use a dosage-unit dictionary to dynamically switch two models based on Conditional Random Fields (CRF), (4) classifying assertions based on voting of five classifiers, and (5) identifying relations using normalized sentences with a set of effective discriminating features.
Macro-averaged and micro-averaged precision, recall and F-measure were used to evaluate results.
The performance is competitive with the state-of-the-art systems with micro-averaged F-measure of 0.8489 for concept extraction, 0.9392 for assertion classification and 0.7326 for relation identification.
The system exploits an array of common features and achieves state-of-the-art performance. Prudent feature engineering sets the foundation of our systems. In concept extraction, we demonstrated that switching models, one of which is especially designed for telegraphic sentences, improved extraction of the treatment concept significantly. In assertion classification, a set of features derived from a rule-based classifier were proven to be effective for the classes such as conditional and possible. These classes would suffer from data scarcity in conventional machine-learning methods. In relation identification, we use two-staged architecture, the second of which applies pairwise classifiers to possible candidate classes. This architecture significantly improves performance.
xy804280; thu; text processing; natural language processing; medical records
To develop a system to extract follow-up information from radiology reports. The method may be used as a component in a system which automatically generates follow-up information in a timely fashion.
A novel method of combining an LSP (labeled sequential pattern) classifier with a CRF (conditional random field) recognizer was devised. The LSP classifier filters out irrelevant sentences, while the CRF recognizer extracts follow-up and time phrases from candidate sentences presented by the LSP classifier.
The standard performance metrics of precision (P), recall (R), and F measure (F) in the exact and inexact matching settings were used for evaluation.
Four experiments conducted using 20 000 radiology reports showed that the CRF recognizer achieved high performance without time-consuming feature engineering and that the LSP classifier further improved the performance of the CRF recognizer. The performance of the current system is P=0.90, R=0.86, F=0.88 in the exact matching setting and P=0.98, R=0.93, F=0.95 in the inexact matching setting.
The experiments demonstrate that the system performs far better than a baseline rule-based system and is worth considering for deployment trials in an alert generation system. The LSP classifier successfully compensated for the inherent weakness of CRF, that is, its inability to use global information.
Text processing; natural language processing; medical records
To build large collections of medical terms from semi-structured information sources (e.g. tables, lists, etc.) and encyclopedia sites on the web. The terms are classified into the three semantic categories, Medical Problems, Medications, and Medical Tests, which were used in i2b2 challenge tasks. We developed two systems, one for Chinese and another for English terms. The two systems share the same methodology and use the same software with minimum language dependent parts. We produced large collections of terms by exploiting billions of semi-structured information sources and encyclopedia sites on the Web. The standard performance metric of recall (R) is extended to three different types of Recall to take the surface variability of terms into consideration. They are Surface Recall (), Object Recall (), and Surface Head recall (). We use two test sets for Chinese. For English, we use a collection of terms in the 2010 i2b2 text. Two collections of terms, one for English and the other for Chinese, have been created. The terms in these collections are classified as either of Medical Problems, Medications, or Medical Tests in the i2b2 challenge tasks. The English collection contains 49,249 (Problems), 89,591 (Medications) and 25,107 (Tests) terms, while the Chinese one contains 66,780 (Problems), 101,025 (Medications), and 15,032 (Tests) terms. The proposed method of constructing a large collection of medical terms is both efficient and effective, and, most of all, independent of language. The collections will be made publicly available.
The bony shell of the turtle is an evolutionary novelty not found in any other group of animals, however, research into its formation has suggested that it has evolved through modification of conserved developmental mechanisms. Although these mechanisms have been extensively characterized in model organisms, the tools for characterizing them in non-model organisms such as turtles have been limited by a lack of genomic resources. We have used a next generation sequencing approach to generate and assemble a transcriptome from stage 14 and 17 Trachemys scripta embryos, stages during which important events in shell development are known to take place. The transcriptome consists of 231,876 sequences with an N50 of 1,166 bp. GO terms and EC codes were assigned to the 61,643 unique predicted proteins identified in the transcriptome sequences. All major GO categories and metabolic pathways are represented in the transcriptome. Transcriptome sequences were used to amplify several cDNA fragments designed for use as RNA in situ probes. One of these, BMP5, was hybridized to a T. scripta embryo and exhibits both conserved and novel expression patterns. The transcriptome sequences should be of broad use for understanding the evolution and development of the turtle shell and for annotating any future T. scripta genome sequences.
Shortening of telomeres, which are essential for maintenance of genomic integrity, is a mechanism commonly associated with the aging process. Here we ascertained whether changes in telomere lengths or telomerase activity correlated with age in normal human mammary epithelial cells (HMEC), or with phenotypes of aging in breast. Accordingly, flow cytometry fluorescence in situ hybridization (flowFISH) was used to determine relative telomere lengths (RTL), and telomerase activity was measured by the telomeric repeat amplification protocol (TRAP), in a collection of 41 primary HMEC strains established from women aged 16 to 91 years.
RTL measurements of HMEC strains that were heterogeneous with respect to lineage composition revealed no significant associations between telomere length with age, maximum observed population doublings, or with lineage composition of the strains. However, within strains, luminal epithelial and cKit-expressing epithelial progenitor cells that were flow cytometry-enriched from individual HMEC strains exhibited significantly shorter telomeres relative to isogenic myoepithelial cells (P < 0.01). In unsorted strains, detectable telomerase activity did not correlate with RTL. Telomerase activity declined with age; the average age of strains that exhibited TRAP activity was 29.7 ± 3.9y, whereas the average age of strains with no detectable TRAP activity was 49.0 ± 4.9y (P < 0.01). Non-detectable TRAP activity also was correlated with phenotypes of aging previously described in HMEC strains; increased proportions of CD227-expressing luminal epithelial cells (P < 0.05) and cKit-expressing progenitor cells (P < 0.05).
Telomere shortening did not correlate with the chronological ages of HMEC strains, whereas decreased telomerase activity correlated with age and with lineage distribution phenotypes characteristic of aging.
Aging; Human mammary epithelial cell; HMEC; Telomere; Telomerase
While Ras is well-known to function on the plasma membrane (PM) to mediate growth factor signaling, increasing evidence suggests that Ras has complex roles in the cytoplasm. To uncover these roles, we screened a cDNA library and isolated H-Ras-binding proteins that also influence Ras functions. Many isolated proteins regulate trafficking involving endosomes; CHMP6/VPS20 and VPS4A, which interact with ESCRT-III, were chosen for further study. We showed that the binding is direct and occurs in endosomes. Furthermore, the binding is most efficient when H-Ras has a functional effector-binding-loop and is GTP-bound and ubiquitylated. CHMP6 and VPS4A also bound N-Ras, but not K-Ras. Repressing CHMP6 and VPS4A blocked Ras-induced transformation, which correlated with inefficient Ras localization to the PM as measured by cell fractionation and photobleaching. Moreover, silencing CHMP6 and VPS4A also blocked EGFR recycling. These data suggest that Ras interacts with key ESCRT-III components to promote recycling of itself and EGFR back to the PM to create a positive feedback loop to enhance growth factor signaling.
signal transduction; oncogene; tumorigenesis; trafficking; ESCRT-III