Technology and Devices
Oral Presentations
-
Unobtrusive Monitoring of Sleep
Zachary Beattie, Tamara HayesKeywords: Unobtrusive Monitoring, Sleep Abstract:Sleep is a vital part of living. It consumes, or at least it should, about one third of everyone’s life. While many are able to enjoy the restorative effects of a ‘good night sleep’ there are a significant amount of others who, plagued by various sleep disorders, cannot. The Institute of Medicine reports that 50 to 70 million Americans suffer from what they refer to as disorders of sleep and wakefulness. They also estimate that the medical cost for these disorders is in the range of hundreds of billions of dollars a year. Part of this exorbitant amount of money is used for diagnosing and monitoring sleep disorders using polysomnography (PSG). Besides being expensive PSG is also obtrusive and inconvenient. Patients who are already struggling with sleep are physically wired to several sensors and asked to sleep normally. PSGs are also not performed frequently enough for individual patients to detect the night-by-night variance that many sleep disorders exhibit. We propose using load cells fitted under the supports of the bed as an alternative, unobtrusive method to monitoring sleep.
-
The effect of phoneme durations on formant frequencies
Akiko Amano Kusumoto, John-Paul HosomKeywords: Speech Modification, Speech Intelligibility Abstract:Clear (CLR) speech, which is spoken deliberately clearly, as if talking to hard-of-hearing listeners, is known to be more intelligible than conversational (CNV) speech, which is spoken as if talking with a colleague [Picheny 1986]. Many studies have been conducted to improve speech intelligibility by mimicking CLR speech characteristics, such as lengthening phoneme durations [Uchanski 1996]. The reason for the limited success of lengthening phoneme durations might be because the characteristics of formant frequencies are disrupted.
In order to modify formant frequencies or phoneme durations and preserve the characteristics of natural speech, we need to first examine the effects of speaking style and phoneme durations on formant frequencies. The objective of this study is to characterize the formant target frequency and transition frequency based on the speaking style and phoneme duration.
The database contains four test words (wheel, will, well, and wail) in a carrier sentence with 2 speaking styles and 3 speaking rates (CNV, CNV /slow, CLR /slow, CLR , and CLR /fast). The results of acoustic analysis showed that the F2 target of CLR speech was higher than that of CNV speech. The results suggest that formant target values of F2 were determined based on the speaking style, independent of phoneme duration. The slope of F2 transition from the consonant to the vowel was also significantly different between CLR and CNV speech, regardless of the speaking rate. The success of this research may lead to an appropriate signal processing algorithm to improve vowel intelligibility.
-
Image Retrieval: Experiences at ImageCLEF
Jayashree Kalpathy Cramer, Steven Bedrick, William HershKeywords: Image Retrieval Abstract:Advances in digital imaging technologies and the increasing prevalence of Picture Archival and Communication Systems (PACS) have led to substantial growth in the number of digital images stored in hospitals and medical systems in recent years. Retrieval of clinical images can be an important in patient care, education, and research. However, image retrieval systems generally do not perform as well as their text counterparts. Clinical images are most often retrieved using the patient or study identifier. They can also be retrieved using manual annotations associated with the image. However, the process of annotation can be a time consuming and error prone. More recently, content-based image retrieval (CBIR) methods, where the image itself is used as the query, have been studied extensively in computer vision. However, the global, low level visual features automatically extracted by these algorithms do not always correspond to high level concepts that a user has in mind for searching. The role of image retrieval in diagnostic medicine can be quite complex, making it difficult for the user to express his information needs appropriately. Purely CBIR has not been very successful in medicine as context is very important for clinical images. Image retrieval in medicine needs to evolve from purely visual retrieval to a more holistic, case-based approach that incorporates various multimedia data sources. We present our hybrid approach to image retrieval, using the image and the annotations in this paper. We present the result of the evaluation performed as part of the international ImageCLEF challenge.
-
The Effects of Hands Free Communication Devices on Clinical Communication: Communication Changes Among Nurses, Nurse Managers, and Information Technology Staff
Joshua Richardson, Joan Ash, PhD, MBA, MLSKeywords: Communication, Computer-Supported Cooperative Work, Wireless Devices Abstract:Hands Free Communication Device (HFCD) systems are a relatively new information and communication technology. HFCD systems enable clinicians to directly contact and communicate with one another using wearable, voice-controlled badges that utilize VoIP (voice-over Internet Protocol) and are interlinked to one another over a wireless local area network (WLAN). This qualitative study employed a grounded theory, multiple perspectives approach to understand how the use of HFCDs affected communication in the hospitals that implemented them. The study generated five themes revolving around the impact HFCD systems had on communication: 1) communication access, 2) control, 3) training, 4) organizational change, and 5) environment and infrastructure. The results demonstrated that communication access provided HFCD users with the ability to quickly call others for help or information. However, the availability and immediacy of HFCD communications required users to learn how to manage, or control, the number and types of calls they and others received. Study participants noted that training was an important factor in proper use of HFCDs. In addition, participants described ways in which HFCD communications brought about change within and across work teams. The physical environment and infrastructure had considerable impact upon the effectiveness and reliability of HFCD communications. In conclusion, users came to rely upon HFCD systems but their implementation brought about organizational and communication changes that need to be addressed in future HFCD system installations.
-
Pipeline Iteration to Improve the Automated Identification of Syntax in Spoken Language
Kristy HollingsheadKeywords: Clinical Diagnosis, Speech, Language Abstract:We present pipeline iteration, an approach that uses output from later stages of a pipeline to constrain earlier stages of the same pipeline. We demonstrate that use of this technique can produce significant improvements in the automated identification of syntactic structures in spoken language. Pipeline systems typically consist of a sequence of processing stages such that the output from one stage provides the input to the next. Each stage in such a pipeline is used to filter the set of possible solutions, and output a smaller set for consideration in the next stage. Pipeline systems are ubiquitous, not only in automated systems such as speech analysis, but also, it could be argued, in human decision-making processes such as triage evaluation. Thus, a generic technique for improving the accuracy of a pipeline system could be widely beneficial. We present such a technique, which we term "pipeline iteration." In pipeline iteration, output from later stages of a pipeline is used as a filter for earlier stages of the same pipeline, in a second pass through the pipeline. We demonstrate that use of this technique can produce significant improvements in the automated identification of syntactic structures in spoken language. We briefly discuss syntactic structures in language, the process of automatically identifying such structures, and highlight a few interesting problems with identifying syntax in spoken language. Finally, we point to existing research on using syntax in the diagnosis of Mild Cognitive Impairment, and argue that improving automated syntax identification will bring us closer to improved diagnoses of cognitive disorders with a language component.
Poster Presentations
-
PRISM: Procedure for Representing Intonation in the Superpositional Model
Taniya Mishra, Jan van SantenKeywords: Speech Analysis And Synthesis; Prosody; Intonation Abstract:This paper describes and applies a new algorithm for decomposing pitch curves into component curves, in accordance with the General Superpositional Model of Intonation. According to this model, which is a generalization of the Fujisaki model, a pitch contour can be described as the sum of component curves that are each associated with different phonological levels, including the phrase, foot, and phoneme. The algorithm assumes that the phrase curve is locally linear during intervals spanned by a foot. The algorithm was evaluated using synthetically generated curves, and was found to accurately recover the synthetic component curves. The algorithm was also evaluated in a perceptual experiment, where speech generated by concatenation of accent curves was shown to produce better speech quality than speech based on direct concatenation of “raw” pitch curve fragments.
-
Polarized light camera (polCAM) to determine margins of skin cancer
Ravikant Samatham, Dr. Steven L. JacquesKeywords: Basal Cell Carcinoma, Mohs Surgery, Polarization Light Camera, Polarization Abstract:Surgery strives to remove skin cancer while maximally preserving normal tissue. It is often challenging to excise skin cancer, like basal cell cancer or squamous cell cancer, because the margins of the cancer extend beyond the clinically observable margins. In this study, a polarized light camera (polCAM) was used to determine the margins of skin cancer. When linearly polarized light propagates through light-scattering media like tissues the polarization state of the light is randomized, but superficially scattered light retains the polarization of incident light. The polCAM creates an image based only on the superficially scattered light, which reveals a fabric pattern of skin. Skin cancer disrupts this pattern, and the doctor can detect this disruption and identify cancer margins. Normally, the doctor relies on changes in skin color due to blood perfusion or melanin pigment and on surface texture changes. The camera detects the colorless subtle sub-surface changes in skin structure based on scattering of light by superficial tissue layers. In an NIH -sponsored pilot clinical study, a hand-held polCAM was designed and tested on several patients undergoing Mohs surgery for skin cancer. The study compared the margins detected by the polCAM with the clinical margins visually observed by the doctor and the margins specified by dermatopathologist after examining the excised cancer tissue. The camera could detect margins that were not recognized by the doctor but later confirmed by the pathologist. We are testing the hypothesis that polCAM can guide surgical excision of skin cancer by identifying the true cancer margins.
-
Test
Heather Hill, Miguel Humberto Torres-UrquidyKeywords: Dental School, Cis, Clinical Information Systems Abstract:Although Clinical Information Systems (CIS) are gaining widespread acceptance in dental schools, their impact on users is not well understood. We conducted separate qualitative studies of the impact of CIS on users in two dental schools and then compared our results. We found five themes in common. By understanding the factors that impact CIS implementation we believe that dental schools will be better prepared to manage them.