Initial evaluation of a continuous speech recognition program for radiology

The aims of this work were to measure the accuracy of one continuous speech recognition product and dependence on the speaker's gender and status as a native or nonnative English speaker, and evaluate the product's potential for routine use in transcribing radiology reports. IBM MedSpeak/Radiology software, version 1.1 was evaluated by 6 speakers. Two were nonnative English speakers, and 3 were men. Each speaker dictated a set of 12 reports. The reports included neurologic and body imaging examinations performed with 6 different modalities. The dictated and original report texts were compared, and error rates for overall, significant, and subtle significant errors were computed. Error rate dependence on modality, native English speaker status, and gender were evaluated by performing ttests. The overall error rate was 10.3 +/- 3.3%. No difference in accuracy between men and women was found; however, significant differences were seen for overall and significant errors when comparing native and nonnative English speakers (P = .009 and P = .008, respectively). The speech recognition software is approximately 90% accurate, and while practical implementation issues (rather than accuracy) currently limit routine use of this product throughout a radiology practice, application in niche areas such as the emergency room currently is being pursued. This methodology provides a convenient way to compare the initial accuracy of different speech recognition products, and changes in accuracy over time, in a detailed and sensitive manner.     

Improving language models for radiology speech recognition

Speech recognition systems have become increasingly popular as a means to produce radiology reports, for reasons both of efficiency and of cost. However, the suboptimal recognition accuracy of these systems can affect the productivity of the radiologists creating the text reports. We analyzed a database of over two million de-identified radiology reports to determine the strongest determinants of word frequency. Our results showed that body site and imaging modality had a similar influence on the frequency of words and of three-word phrases as did the identity of the speaker. These findings suggest that the accuracy of speech recognition systems could be significantly enhanced by further tailoring their language models to body site and imaging modality, which are readily available at the time of report creation.     

Experience with implementation of a radiology speech recognition system

Recent advances in speech recognition technology have allowed development of computer systems for real-time radiologist-driven generation of reports. The transition to a speech recognition system is a technically complex process with many potential pitfalls that can decrease efficiency and disrupt workflow. In our recent experience with installation of such a system in an academic radiology department, factors that have worked against optimal performance have included environmental logistics, hardware incompatibilities, radiology information system interface problems, lack of suitable training, and inadequate technical support. Communication of our experience is intended to allow radiologists to anticipate complications of these systems and make informed decisions regarding the feasibility of such a system in their practices. With this information, potential buyers should be able to carefully scrutinize specifications for prospective systems and, by avoiding many of the possible pitfalls, make an easier transition to a speech recognition environment.     

Comparison of feature evaluation criteria for speech recognition based on electromyography

In this paper, we present a performance comparison of 14 feature evaluation criteria and 4 classifiers for isolated Thai word classification based on electromyography signals (EMG) to find a near-optimal criterion and classifier. Ten subjects spoke 11 Thai number words in both audible and silent modes while the EMG signal from five positions of the facial and neck muscles were captured. After signal collection and preprocessing, 22 EMG features widely used in the EMG recognition field were computed and were then evaluated based on 14 evaluation criteria including both independent criteria (IC) and dependent criteria (DC) for feature evaluation and selection. Subsequently, the top nine features were selected for each criterion, and were used as inputs to classifiers. Four types of classifier were employed with 10-fold cross-validation to estimate classification performance. The results showed that features selected with a DC on a Fisher’s least square linear discriminant classifier (D_FLDA) used with a linear Bayes normal classifier (LBN) gave the best average accuracies, of 93.25 and 80.12% in the audible and the silent modes, respectively.     

Development of a (silent) speech recognition system for patients following laryngectomy

Surgical voice restoration post-laryngectomy has a number of limitations and drawbacks. The present gold standard involves the use of a tracheo-oesophageal fistula (TOF) valve to divert air from the lungs into the throat, which vibrates, and from this, speech can be formed. Not all patients can use these valves and those who do are susceptible to complications associated with valve failure. Thus there is still a place for other voice restoration options.

With advances in electronic miniaturization and portable computing power a computing-intensive solution has been investigated. Magnets were placed on the lips, teeth and tongue of a volunteer causing a change in the surrounding magnetic field when the individual mouthed words. These changes were detected by 6 dual axis magnetic sensors, which were incorporated into a pair of special glasses. The resulting signals were compared to training data recorded previously by means of a dynamic time warping algorithm using dynamic programming. When compared to a small vocabulary database, the patterns were found to be recognised with an accuracy of 97% for words and 94% for phonemes. On this basis we plan to develop a speech system for patients who have lost laryngeal function.     

An evaluation of a behavioral program for chronic pain

The current study evaluated a behavioral program for the management of chronic pain. One hundred twenty-one patients were treated. Primary pain complaints were predominantly in the low back or head/neck/face regions. Patients participated in a 4-week inpatient treatment package consisting of controlled medication reduction, physical therapy, behavioral group therapy, self-monitoring, contracting, and biofeedback/relaxation training. Several measures of verbal/nonverbal pain behavior and physical functioning were obtained over the baseline, treatment, and follow-up periods. Results showed significant reductions in analgesic medication use and verbal/nonverbal pain behavior and improvements in physical functioning, employment status, and pro-health behaviors which were maintained at 12-month follow-up.The authors wish to thank the following people for their assistance: D. Bugold, N. Flynn, A. Glavan, L. Graham, J. Henkel-Johnson, E. Martin, P. Panyan, B. Patric, K. Slack, J. Sommerfeld, K. Wahman, S. Wahman, B. Wilk, and M. Wipson. This research was supported in part by a grant from the Paul F. Dwan Foundation.     

ERPs during continuous recognition memory for words

Event-related brain potentials (ERPs) were recorded from 10 young adults during a version of the continuous recognition memory paradigm. Words were presented after lags of either 2, 8 or 32 intervening items (equiprobable) following their first presentation, and subjects were required on each trial to make a choice: new (never presented previously) or old (previously presented) response. To assess the effect of probability of new to old items, words were presented in separate blocks with ratios of new to old of 2:1 and 1:1. Reaction time increased and successful recognition decreased systematically as the lag between first and second presentations of an item increased, supporting the distinction between primary (immediate memory) and secondary memory for verbal material. However, there were no systematic effects of item lag on the ERP components. ERPs to new items were characterized by larger N300 and smaller P300 amplitudes (from about 250 to 700 ms) than those to old items. These amplitude differences between old and new ERPs were interpreted as primarily reflecting repetition as opposed to semantic priming effects. These old/new effects did not interact with probability, suggesting that frequency of occurrence is not a major determinant the ERP old/new difference. Old items elicited a late negativity following the behavioral response that was interpreted as due to the presence of a "positive slow wave," with a frontally oriented distribution to new words that was absent in the ERPs to old words. Similarly, subtraction of ERPs elicited by new items that were subsequently unrecognized from those subsequently recognized, showed that underlying the ERP subsequent "memory effect" was a "frontal positive slow wave," dissociable from P300 on the basis of differences in scalp distribution. Since positive slow wave has been interpreted as reflecting "further processing," the data suggest that such processing, possibly similar to elaboration (Graf & Mandler, 1984), enhanced the probability of subsequent recognition.     

Computer program for intestinal spike bursts recognition

A FORTRAN program has been developed for locating intestinal spike bursts and for estimating their strength. Tested against human scanning, the reliability rate was 92% and the misrecognition rate was 2.5%.This program was applied to the automatisation of the Migrating Myoelectric Complex analysis. A first method computed the percentage of Basic Electrical Rhythm (BER) cycles with superimposed spike bursts. A second one was based on the evaluation of spike bursts strength.Supported by grant no. ATP 177540 and no. CRL 74.518507 from the INSERM     

Initial experience with a radiology imaging network to newborn and intensive care units

A digital image network has been installed in the James Whitcomb Riley Hospital for Children on the Indiana University Medical Center to create a limited all digital imaging system. The system is composed of commercial components, Philips/AT&T CommView system, (Philips Medical Systems, Shelton, CT; AT&T Bell Laboratories, West Long Beach, NJ) and connects an existing Philips Computed Radiology (PCR) system to two remote workstations that reside in the intensive care unit and the newborn nursery. The purpose of the system is to display images obtained from the PCR system on the remote workstations for direct viewing by referring clinicians, and to reduce many of their visits to the radiology reading room three floors away. The design criteria includes the ability to centrally control all image management functions on the remote workstations to relieve the clinicians from any image management tasks except for recalling patient images. The principal components of the system are the Philips PCR system, the acquisition module (AM), and the PCR interface to the Data Management Module (DMM). Connected to the DMM are an Enhanced Graphics Display Workstation (EGDW), an optical disk drive, and a network gateway to an ethernet link. The ethernet network is the connection to the two Results Viewing Stations (RVS) and both RVSs are approximately 100 m from the gateway. The DMM acts as an image file server and an image archive device. The DMM manages the image data base and can load images to the EGDW and the two RVSs. The system has met the initial design specifications and can successfully capture images from the PCR and direct them to the RVSs.(ABSTRACT TRUNCATED AT 250 WORDS)     

Managing predefined templates and macros for a departmental speech recognition system using common software

The authors have developed a networked database system to create, store, and manage predefined radiology report definitions. This was prompted by complete departmental conversion to a computer speech recognition system (SRS) for clinical reporting. The software complements and extends the capabilities of the SRS, and 2 systems are integrated by means of a simple text file format and import/export functions within each program. This report describes the functional requirements, design considerations, and implementation details of the structured report management software. The database and its interface are designed to allow all radiologists and division managers to define and update template structures relevant to their practice areas. Two key conceptual extensions supported by the template management system are the addition of a template type construct and allowing individual radiologists to dynamically share common organ system or modality-specific templates. In addition, the template manager software enables specifying predefined report structures that can be triggered at the time of dictation from printed lists of barcodes. Initial experience using the program in a regional, multisite, academic radiology practice has been positive.     

语音识别技术在咳嗽声音自动识别中的应用

目的 利用双门限法、Mel频率倒普系数(MFCC)法及矢量量化(VQ)法的语音识别技术对咳嗽声音进行自动识别.方法 在安静环境下,对5例健康成年人和15例咳嗽患者的非咳嗽和咳嗽声音进行录音,分别随机分为训练样本和测试样本.训练样本用于生成咳嗽识别软件的码本,并用该码本对测试样本进行自动识别分析.同时与人工识别的结果进行对比,计算敏感性、特异性,记录两种方法的识别时间.结果 用于码本生成的咳嗽声音和非咳嗽声音均为200次,测试样本的咳嗽和非咳嗽声音分别为375次和125次.人工识别和通过码本自动识别测试样本的时间分别为33 min 18 s和1min 35 s;码本自动识别咳嗽声音的敏感性和特异性分别为98.93%和100%.结论 基于VQ的双门限法及MFCC法可用于咳嗽声音的自动识别.     

Development and evaluation of a multi-family psychoeducational program for cystic fibrosis

The development, implementation and evaluation of a psychoeducational program for families with a child affected by cystic fibrosis (CF) is described. Aim of the program was to strengthen the families' coping with CF-related problems and to improve adherence with chest physiotherapy.Sixteen families from an outpatient CF-clinic participated. Parents and children were educated both individually as well as together in multi-family groups. Teaching, practicing and group discussions were balanced in each session. In a pre-post-design the following variables were evaluated: parental coping, parental health beliefs, children's coping, adherence, and knowledge about CF.The parental coping patterns and health beliefs remained unchanged in the study group. Children developed more search for social support, whereas their competence and optimism decreased and withdrawal increased slightly as perceived by the parents. There is a subgroup with poor family functioning and adherence at pretest which improves after the intervention.Family-centered psychoeducational intervention may be a promising supportive strategy for children with CF, especially if it is dedicated to families with poor adaptation to the disease.     

Qualitative and quantitative evaluation of charge capture by a radiology information system

Installation of a radiology information management system (RIS) is usually justified on the basis of improved departmental efficiency and improved charge capture. However, evaluation of the success of these expected improvements is often difficult. The installation and operation of such a system in a medium-sized tertiary care hospital has permitted the effects of the RIS on the operation of the department to be studied and the improvements in charge capture provided by the system to be quantitatively assessed. As a result of a side-by-side comparison with a conventional check-sheet manual billing system, it is apparent that the RIS reduces the errors inherent in manual systems. Subjectively, it is also apparent that personnel prefer the computerized system to the manual charge sheets.     

Development and evaluation of a presurgical preparation program

Three presurgical preparation programs were developed and evaluated in an Australian hospital utilizing an additive component design. The component basic to all three preparation programs was modeling. This technique was compared with the additional components of teaching child coping skills and parent coping skills via videotape. Subjects were 28 children between the ages of 4 and 13 years who were scheduled for elective surgery. Anxiety of both the children and parents was assessed by self-report and behavior observation. Results indicated that there was no further anxiety reduction by the addition of child and parent coping skills. Results are discussed in terms of the viability of teaching coping skills via videotape particularly to parents. Methodological difficulties associated with research in this area are examined.     

An evaluation of a juvenile court volunteer program

A longitudinal study of the impact of three separate volunteer programs working with juvenile delinquents was undertaken at one juvenile court. The youth were followed for a period of one year with measures of official and unofficial delinquent behavior obtained four times during that year. Other data were also collected. The study results indicated two of the programs (i.e., volunteer probation officer and group counseling) had a negligible effect on reducing probationers' delinquent behavior compared to a control group. Youth who participated in the third group, a volunteer tutor program, actually increased their delinquent behavior when compared to a control group within the first 6 months of their probation. However, one year after probation, the differences were negligible. Recommendations are made which include divesting the juvenile court of client responsibility for providing volunteer services.     

An initial evaluation of a mindful parenting program

A growing number of children are experiencing marital transition. The effects of divorce on children have typically been considered deleterious, although factors can buffer the difficulty of postdivorce adjustment. One of these factors is a positive relationship with a parental figure. Unfortunately, divorce often overwhelms parents with a series of changes that compromise their parenting skills. One new approach to improving parenting after divorce is mindful parenting, which aims to enhance interpersonal and emotional connection in the parent-child relationship. This program is intended to facilitate parents' self-awareness, their mindfulness, and their intentionality in responding to their child's needs. The present study reports on the implementation of the Mindful Parenting Program, delivered in two groups to 12 recently divorced parents with preschool-aged children. Program effectiveness was conducted on two levels. First, mindfulness measured by the Toronto Mindfulness Scale revealed significant increases over the intervention and posttest period. Second, in-home behavioral observations conducted pre- and postintervention revealed no changes in parent-child relationships. These findings are discussed within the larger context of facilitating effective parenting postdivorce.     

Initial evaluation of a commercial EPID modified to a novel direct-detection configuration for radiotherapy dosimetry

Electronic portal imaging devices (EPIDs) integrated with medical linear accelerators utilize an indirect-detection EPID configuration (ID-EPID). Amorphous silicon ID-EPIDs provide high quality low dose images for verification of radiotherapy treatments but they have limitations as dosimeters. The standard ID-EPID configuration includes a high atomic number phosphor scintillator screen, a 1 mm copper layer, and other nonwater equivalent materials covering the detector. This configuration leads to marked differences in the response of an ID-EPID compared to standard radiotherapy dosimeters such as ion chambers in water. In this study the phosphor and copper were removed from a standard commercial EPID to modify the configuration to a direct-detection EPID (DD-EPID). Using solid water as the buildup and backscatter for the detector, dosimetric measurements were performed on the DD-EPID and compared to standard dose-in-water data for 6 and 18 MV photons. The sensitivity of the DD-EPID was approximately eight times less than the ID-EPID but the signal was sufficient to produce accurate and reproducible beam profile measurements for open beams and an intensity-modulated beam. Due to the lower signal levels it was found necessary to ensure that the dark field correction (no radiation) DD-EPID signal was stable or updated frequently. The linearity of dose response was comparable to the ID-EPID but with a greater under-response at low doses. DD-EPID measurements of field size output factors and beam profiles at the depth of maximum dose (dmax), and tissue-maximum ratios between the depths of 0.5 and 10 cm, were in close agreement with dose in water measurements. At depths beyond dmax the DD-EPID showed a greater change in response to field size than ionisation chamber measurements and the beam penumbrae were broader compared to diode scans. The modified DD-EPID configuration studied here has the potential to improve the performance of EPIDs for dose verification of radiotherapy treatments.