• Share
  • Email
  • Embed
  • Like
  • Save
  • Private Content
Fn2410141018
 

Fn2410141018

on

  • 247 views

 

Statistics

Views

Total Views
247
Views on SlideShare
247
Embed Views
0

Actions

Likes
0
Downloads
0
Comments
0

0 Embeds 0

No embeds

Accessibility

Categories

Upload Details

Uploaded via as Adobe PDF

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

    Fn2410141018 Fn2410141018 Document Transcript

    • A.Lavanya, Sireesh Babu/ International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com Vol. 2, Issue 4, July-August 2012, pp.1014-1018 Development of Heart Prescreening Device using low Power MSP430 device A.Lavanya*, Sireesh Babu** *M.Tech student, ECE.Dept, PKSK college **ASSOC.prof, ECE.Dept , PKSK collegeAbstract- Inventions are done keeping oppurtunities ratio is 3:10 000 and is highly biased toward urbanand expectations of developing economies in terms areas, thus creating a rural urban divide in cardiacof health care. The designed system is a point-of- healthcare. The designed device also enables acare (POC) device that can send heart-care paramedic to contribute significantly in ruralservices to the rural areas and can install urban healthcare in case of unavailability of a trainedmethodologies in healthcare delivery.The product physician, and in prescreening programs.design uses several innovations including theeffective use of adaptive and multiresolution The presently available electronicssignal-processing techniques for acquisition, stethoscopes like HD Medical-HDfono and 3Mdenoising, segmentation, and characterization of Stethos are aimed for educational and trainingthe heart sounds (HS) and murmurs using an purposes. In literature, reference of devices forultralow-power embedded Mixed Signal cardiac prescreening and diagnosis using HS areProcessor. The device is able to provide indicative rare.diagnosis of cardiac conditions and classify a In the present study, we use a Texassubject into either normal, abnormal, ischemic, or Instruments’ Mixed Signal Processor (TI-MSPvalvular abnormalities category. Preliminary results 430) based handheld device to innovate andgenerated by the prototype states the applicability develop a Point-of-Care (POC) heart care andof the device as a rescreening tool that can be used patient monitoring system that can be used forby paramedics in rural outreach programs. It is also noninvasive cardiac screening. This system isfound from the feedback of medical proffessional utilized in determining human subjects susceptiblethat they can helpful in determining congenital to cardiac ailments or showing early symptoms,heart diseases. This system aims to determine a thus mitigating early diagnosis and enabling aframework for utilization of automated HS analysis demographic study of cardiac cases. In future wesystem for community healthcare and healthcare plan to integrate the system with Microsoft CE-inclusion. SQL-based patient database management system and haptic interfaces using Microsoft Windows CEIndex Terms—Adaptive signal processing, for easy operation.biomedical signal processing, cardiovascularsystem, data acquisition. II. SYSTEM ARCHITECTURE Cardiology is a grand challenge in healthcareI. INTRODUCTION and has inspired in envisioning the design concept MEDICAL practitioners use cardiac with several intended benefits including low costauscultation for an objective clinical screening of semi real-time analysis of data, door to doorcardiac conditions and prediagnosis. It is, however, recording services enabled by mobile ornoted that a real-time system for the same is ultramobile devices, demographic dataconspicuously lacking in development. The prediagnostic input, and expert intervention in casenonstationary nature of cardiac signals also makes of emergency through GSM/GPRS networks.the generation of a robust system for heart sound(HS) analysis difficult. Diagnosing heart diseases A. Target Specificationsusing stethoscope and ECG are two fundamental Several technological innovations are built inapproaches due their efficiency, simplicity, and into the system prototype: 1) development based onnoninvasiveness [1]–[3]. HS auscultation highly ultralow-power embedded processors (TI-MSPdepends on the hearing ability, skill, and 430); 2) expert intervention of medicalexperience of a cardiologist [1]. In real time, lung practitioners by remote monitoring at tertiarysounds interfere with the HS making it noisy, healthcare centers; 3) the system incorporates statewhich leads to increased difficulty in diagnosis of of the art-denoising algorithms suited for field useheart diseases using signal-analysis techniques. An in the rural nonclinical setup; 4) the data analysis isintegrated HS analysis tool thus becomes vital to based on physiological considerations and medicalassist the cardiologist [2]. Study shows that in domain knowledge has been employed fordeveloping countries, cardiologist-to-population classifying the signal in four preliminary classes 1014 | P a g e
    • A.Lavanya, Sireesh Babu/ International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com Vol. 2, Issue 4, July-August 2012, pp.1014-1018that includes normal and three abnormal conditions(see Fig. 1). Finally, when the signal is fed into TI- MSP430, it samples the analog input, applies the signal-processing algorithms, and separates the HS signal from the lung sound accordingly. The signal is segmented for each beat and clustered based on its frequency and temporal features. B. Hardware Platform: TI-MSP 430 and Its We aim to optimize the device capabilities Advantages in Embedded Applicationsboth in vivo and in vitro by making the design and The internal architecture of TI-MSP 430 comprisesdevelopment cycles flexible and adaptive. As of multiple analog and digitalmodules throughdepicted in Fig. 2, TI-MSP430 is the heart of the which CPU can communicate with the externalsystem; it handles the role of data acquisition as interfaces and devices. Several features make thewell as signal processing for the input composite TI-MSP430 suitable for low-power and portablesignal. The preamplification unit consists of a low- applications. The CPU is small and efficient with apass filter (LPF) and high-pass filter (HPF) circuit. large number of registers. Several low-powerThis LPF–HPF circuit in combination with the operating modes reduce power consumption andamplification unit forms the signal conditioning extend battery life. There is a wide choice ofblock that performs the primary filtering operation clocks: from the low-frequency watch crystal toto delimit the unnecessary components of the input high-frequency digitally controlled oscillatorsignal. It also amplifies the signal of interest to an (DCO). The hardware module provides RS232 andextent so that it can be properly gated to the input RS485 for communicating with computer, audioof TI-MSP430 [3]. It is, however, not able to inputs and outputs, switching applications, etc., todeinterlace the overlapped heart and lungs sounds make the system versatile.for which wavelet-based denoising is introduced inthe processing unit. Postamplification unit C. Hardware Implementationcomprises the signal conditioning block that The schematic of the system including theperforms the filtering operation to remove the steps for implementation of the algorithms has beenunnecessary region of the input signal and depicted in Fig. 3. The HPF-LPF circuit designedamplifies the signal of interest. The in the present instrumentation consists of a 4thsignalprocessing algorithms are implemented on order Butterworth Low Pass Filter (cut-off ∼ 1000the TI-MSP430 to minimize lung sounds, ambient Hz) and a 4th order Butterworth High Pass Filternoise, and acoustical distortions. (cutoff ∼ 20 Hz). The Sallen-Key Topology is used for filter design because of its inherent gain accuracy. Moreover, Butterworth filter provides a satisfactorily flat pass-band magnitude response and attenuation of∼ −3 db at the cut-off frequency (see Fig. 4). The implementations of the algorithms are carried out in the IAR workbench (V 5.4), using 1015 | P a g e
    • A.Lavanya, Sireesh Babu/ International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com Vol. 2, Issue 4, July-August 2012, pp.1014-1018“C” language given its simplicity and versatility in instead of being derived separately, consists of aembedded programming. The programs are delayed version of the primary (input) signal. Thecompiled in the TI-MSP430 IAR C/C++ compiler. delay is provided to decorrelate the noise signal soBoth A/D and D/A operations are performed in low that the adaptive filter (used as a linear predictionpower mode (LPM0) that draws a power∼ 95μA. filter) cannot predict the noise signal while easilyThis feature of the program reduces the power predicting the signal of interest. Thus, the outputconsumption at the time of A/D and D/A will contain only the signal of interests, which willconversions. The CPU only commands over the be again subtracted from the desired signal and thefiltering program during which the TI-MSP430 error signal will then be used to adapt the filterswitches from low power mode to active mode (I ∼ weights to minimize the error.We implement the600 μA). ALE model using the least mean square (LMS) algorithm and the recursive least square (RLS) algorithm. B. Wavelet-Based Denoising It is observed that orthogonal wavelet transform compresses the energy in a signal into few large components, whereas the noise is disorderly and characterized by small coefficients scattered throughout the transform. We can neglect these smaller coefficients from the wavelet- decomposed details, and thus, reduce the noise [10]. In DWT, the energy content is concentrated in larger wavelet coefficients from that we can easily reconstruct back the original signal. In the first stage of algorithm, all the 1) SureShrink: Sureshrink is a smoothnessproper settings for ADC, DAC and port 6 are adaptive algorithm and works at multipleinitialized. The ADC process is established by levels of wavelet decomposition and uses thesetting the ADC12SC bit. The sample values are principle of Stain’s unbiased risk estimatorfed to the input of the filter algorithm. For faster (SURE) for risk estimates [11]. For ouroperation, the filtering process is executed with application, we modified the SureShrinkhigh clock frequency. The filtered signal data is algorithm and parameterized the equations toaccumulated in the DAC12_1 data register for D/A reduce computational cost.operation. The process continues iteratively, andwhenever a sample value comes to the input buffer 2) BayesShrink: BayesShrink is a waveletof the filter it modifies its array, performs filtering shrinkage method that has emerged muchoperation and channelizes the result to the earlier and is outperformed by newer emergentDAC12_1 data register. DAC output can be techniques. The same is, however, included inobtained at the port 6.6 pin and is been gated to the the study because of its computational costLCD for visual inspection. efficiency that make it better suited for fast real-time and embedded operations.III. SIGNAL DENOISING AND SEGMENTATION C. Segmentation of HSs There are severalsignal-processing The signals are decomposed to level 5schemes requiring a separate noise reference for using Daubachis 6 (db6) wavelets. The detaildenoising of HS adaptively. This study is based on coefficients of levels 4, 5, and 6 are used for theapproaches where single channel is required for HS calculation of Shannon’s entropy. The subbands aredenoising. In designing the embedded device, we scaled and added up to give a partialhave used two classes of algorithms for denoising: reconstruction.1) adaptive signalprocessing algorithms, and 2)multiresolution signal processing using wavelets The normalized Shannon’s entropy is[9]. calculated from the signal and is used to generate the signal envelope. The same is then passed to a peak detector. The subbands are scaled up andA. Adaptive Line Enhancer added to give a partial reconstruction that amplifies The adaptive line enhancer is a system that is the S1–S2 factors. In case we have a detected peak,used frequently in biological signal processing to we pass on the data to a boundary-calculationextract the desired biosignal from the background algorithm else recalculate the peak points bywideband noise [4]–[8]. Its reference signal, readjusting the threshold adaptively. Other entropy 1016 | P a g e
    • A.Lavanya, Sireesh Babu/ International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com Vol. 2, Issue 4, July-August 2012, pp.1014-1018measures are being employed and evaluated of the HS and its temporal properties. The valvularincluding fuzzy entropy, Kapur’s entropy, and disorders can be easily picked up by identifying theapproximate entropy measures. The measures for murmurs in the segmented HS. When concomitantenvelop detection presents an interesting ECG–PCG signal recordings are studied, we cancomparison, however, the same lies beyond the find the evidence of atrial fibrillation. Ischemicscope of the present article (see Fig. 5). heart diseases can be diagnosed when there is an early or late heart failure, S3 Gallop, or a loud P2IV. PERFORMANCE EVALUATION (loud P2 often indicates a pulmonaryThe data were acquired in situ using the hypertension); an atrial septal defect is indicated byinstrumentation designed. Total 72 samples were a split of second HS. Coronary stenosis is generallyobtained from 17 volunteers. The evaluation is known to produce sounds due to turbulent flow ofbased on the comparison of the power spectral blood in the occluded arteries. Normally, thedensity (PSD) of the signals both at the input of the sounds are masked and are not audible clearlysystems and the output of the systems for LMS- during the systolic phase but the same can beALE and RLS-ALE, and wavelet methods. picked up by precision sensors during the relativelyHSrecover and NOISE reduction are computed in quiet diastolic phase [5]. The extraction of usefulterms of percentages (see Table I) information from the diastolic sounds associated with coronary occlusions using the adaptive signal- processing algorithms and the use of clinical examination variables can yield encouraging results. The signals are highly attenuated and complex so high precision microphones are required to detect the sound signals. The feature vectors extracted from the diastolic sounds analyzed, in addition, other physical examination variables like pulse rate, ambient/body temperature, and patient details are registered using the system that can be used by fuzzy and neural network classifiers located at a remote server to validate the results and minimize false detections. The clustering of the systolic and the diastolic segments are done with the help of K- means algorithm♠ [12], where K = 2, a fuzzy inference engine† is used to incorporate the data from the other peripherals (see Table II for results). The frequency features were determined withwhere xHS (n) is the original HS, xhs noi (n) is reference to the S1 and S2 peaks determined usingtheHS corrupted with noise signal, and y(n) is the the entropy plots (see Fig. 5).output recovered signal after application of thefiltering techniques. It is evident that LMS andBayesShrink algorithms are able to recover the HSsignal from wideband noisy background withsufficient accuracy. However, the performance ofthe SureShrink and RLS-ALE algorithm is muchbetter than the LMS-ALE and BayesShrink. The e-General Medical and the Texas Heart Instituteonline datasets were also employed for validationof the methods using synthetically added whiteGaussian noise (see Fig. 6).V. CLASSIFICATION AND ITS CLINICAL SIGNIFICANCE The system is designed to classify the signalsinto: 1) abnormal and 2) normal classes at theonset. The abnormal classes once screened arefurther classified into: 1) valvular heart diseases(VHD), 2) ischemic heart disease (IHD), and 3)abnormal undetermined (AbU). Thesubclassification is based on frequency signatures 1017 | P a g e
    • A.Lavanya, Sireesh Babu/ International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com Vol. 2, Issue 4, July-August 2012, pp.1014-1018 [2] R. M. Rangayyan, Biomedical Signal Analysis: A Case-Study Approach. New York: IEEE Press/Wiley, 2002. [3] L. Cromwell, F. J.Weibell, and E. A. Pfeiffer, Biomedical Instrumentation &Measurements, 2nd ed. Cromwell: Books, 1980. [4] M. Kompis and E. Russi, “Adaptive heart- noise reduction of lung sounds recorded by a single microphone,” in Proc. 14th Ann. Int. Conf. IEEE EMBS, 2003, pp. 2416–2419. [5] Y. M. Akay, M. Akay, W. Welkowitz, J. L. Semmlow, and J. B . Kostis, “Noninvasive acoustical detection of coronary artery disease: a comparative study of signal processing methods,” IEEE Trans. Biomed. Eng., vol. 40, no. 6, pp. 571–578, Jun. 1993. [6] V. K. Iyer, P. A. Ramamoorthy, H. Fan, and Y. Ploysongsang, “Reduction of heart sounds from lung sounds by adaptive filtering,” IEEE Trans. Biomed. Eng., vol. 33, no. 12, pp. 1141–1148, Dec. 1986. The applicability of the system can further [7] L. J. Hadjileontiadis and S.M. Panas,be extended to pediatric heart care and can be used “Adaptive reduction of heart sounds fromfor early detection of ventricular septal defect lung sounds using fourth-order statistics,”(VSD), bicuspid aortic valve disorders and patent IEEE Trans. Biomed. Eng., vol. 44, no. 7,ductus arteriosus (PDA). As per statistics, 2–5 pp. 642–348, Jul. 1997.children are detected with congenital heart defects [8] W. K. Ma, Y. T. Zhang, and F. S. Yang, “Aper 1000 live births. Hence, use of a noninvasive fast recursive-least-squares adaptive notchand safe methodology, i.e., auscultation, for filter and its applications to biomedicalidentification of these conditions can provide a signals,” in Medical and Biological Eng.paradigm shift in pediatric and neonatal healthcare. Comput., Springer, vol. 37, no. 1, Jan. 1999. [9] T. R. Reed,N. E. Reed, and P. Fritzson,VI. CONCLUSION “Heart sound analysis for symptom The presence of any abnormality in the detection and computer-aided diagnosis,”subject’s heart alters the temporal domain signature Simul. Modeling Pract. Theory, vol. 12, pp.of the HS, and accordingly, it can be sensed by the 129–146, 2004.device. These latent changes in the signal are [10] D. L. Donoho and I.M. Johnstone,monitored in a real-time basis after processing on “Adapting to unknown smoothness viaTI-MSP430. The system has a special color scheme wavelet shrinkage,” J. Amer. Statist. Assoc.,to indicate a patient risk; further, a networked vol. 90, p. 1200, Dec. 1995.cardiologist is alerted for any abnormalities in the [11] F. Luisier, T. Blu, and M. Unser, “A newrecorded HS. This system design facilitates the use SURE approach to image denoising:as a POC device owing to its properties of Interscale orthonormalwaveletportability and real-time operation. It is primarily thresholding,” IEEE Trans. Image Proc., vol.aimed to identify the population suffering from 16, no. 3, pp. 593–606, Mar. 2007.various heart ailments through sufficient screening [12] Z. Syed, D. Leeds, D. Curtis, F. Nesta, R. A.enabling early detection leading to timely Levine, and J. Guttag, “A framework for thediagnosis, and targeted remedial measures. The analysis of acoustical cardiac signals,” IEEEmethodology thus developed is the core technology Trans. Biomed. Eng., vol. 54, no. 4, pp.for a rural heart care delivery and diagnosis system 651–662, Apr. 2007.in form of an embedded handheld device, whichcan be carried to the village households ensuringsustainable healthcare delivery.REFERENCES[1] Morton E. Travel, “Cardiac auscultation: A glorious past- but does it have a future?,” Circulation, vol. 93, pp. 1250–1253, 1996. 1018 | P a g e