"Nadi Tarangini A Pulse Based Diagnostic System - CSE_ IIT Bombay"
Nadi Tarangini: A Pulse Based Diagnostic System Introduction. •Ayurveda: Indian Traditional Medicine for diagnosis. •Looking, listening, smelling, asking, touching. •The arterial pulse contains large amount of information about disorders. •Probably information in pulse under-used. •Pulse waveform: Physiological signal e.g. ECG. •Ayurveda Traditional Chinese Medicine (TCM). •Researchers already getting good results in TCM. •Previous methods: Pressure sensors , transducer, photoplethysmograph, infrared LED & three photodiodes, condenser microphone, etc. Observations and Discussion. Varying Pressure. At each pressure, the obtained pulse gives different insights about the body. Chaos Analysis. •Observation: The dynamics is changing in pulse. •Recurrence Plot: •Visually reveals information about correlations. •Structural changes in pulse. As the pressure applied on sensor increases Ri, j ( i || ( xi ) x j ||) where, i, j 1,..., N Variations with age. The patterns are different for three age-groups. •„below 25‟ pulses are more dominant in secondary peaks. •„25-50‟ group is relatively stable. •Older pulses are irregular in nature. •The pulse duration increases (rate decreases) as the age increases xi s : Phase space vectors (x): Heaviside function i : Cut-off distance Our goal: Pulse-based CAD system. •Imitate the skill of “feeling” the pulse. •Ayurvedic basis: types and sub-types of nadi. •Quantitative basis: machine learning algorithms. •Remove subjectiveness. Self-similar Nature. •Fractal forms: consists of subunits that resemble the structure of the overall object. •Physiological signals: •ECG (explored extensively). •Pulse (First time -- to our knowledge). •Step or cusp-like singularities. •Different portions have different scaling properties. •Multifractal formalism: The power-law scaling relationship. •Recurrence Quantification Analysis Descriptors: •Texture of a RP. •Large & small scale structures in a RP. Longest Recurrence Entropy Diagonal Normal Abnormal 0.062 0.116 96 179 3.428 2.330 Trend -0.059 -.0.034 Set-up of Nadi Tarangini. Fourier Analysis. Important Properties. •Sampling frequency 500Hz. •Rich in harmonics, Complete, Reproducible. •Time-domain features •percussion wave (P) •tidal wave (T) •valley (V) •dicrotic wave (D) WTMM tree: Hierarchical Organization of the singularities Multifractal Spectra have different mean and range values for various patients •The strength of the spectral harmonics: •Accounts for the morphology of the pulse. •Thus, can be used in detection of pulse. •First 100 Fourier coefficients of only vata pulse. Beat-to-Beat Alterations. •HRV has been analyzed for long to detect arrhythmias. •Similar analysis on long pulse data •Peak Detection: A pre-requisite for capturing such beat-tobeat variations. •accurately find the start and end of each beat (pulse cycle). Distinctly observable patterns. Final Remarks. •We have designed a high quality pulse DAQ system. • Incorporated recent developments in instrumentation technologies. •Our pulse waveforms, in the form of time series, have high details. •Pulse signal: typical properties of a physiological signal. •The information in arterial pulse is probably underused. •Rigorous machine learning algorithms could be applied on these waveforms to •Identify of types and sub-types of Nadis defined in Ayurvedic literature. •Diagnostic purposes to classify into possible disorders. Normal pulse behavior Blood-Pressure problem Fever: Very vibrating pulse Muscular (back-) pain Angle at main peak •Complex frequency B-spline wavelet. •Whenever there is a peak in the pulse waveform. •A positive and a negative spike in both the real and imaginary parts. •On our database of 79 waveforms: 100% accuracy. •Steady data for longer duration is required. •Time-domain, Frequency-domain, or using morphologybased features. •Amplitudes, energies, slopes, angles, entropies, velocities, etc. •It is the variations between consecutive beats that is critical, rather than the heart/pulse rate or the average values. •Variations in amplitudes, slopes, systolic & diastolic energies, and so on. •Machine learning algorithms can be applied to distinguish major and sub-types of nadis defined in Ayurveda. References. •Patent:- Dr. Ashok Bhat, Aniruddha Joshi and Anand Kulkarni, "A system for complete spectrum of the nadi pulses as a time-series", Indian Patent Application 197819, granted on 26th Dec 2005. •Paper:- A.J. Joshi, A. V. Kulkarni, S. Chandran, V. K. Jayaraman, B. D. Kulkarni, “Nadi Tarangini: A Pulse Based Diagnostic System”, In Proc. of the 29th IEEE-EMBC Conf., pp. 2207-2210, 2007. •Paper:- A. J. Joshi, S. Chandran, V. K. Jayaraman, B. D. Kulkarni, “Arterial Pulse System: Modern Methods For Traditional Indian Medicine”, In Proc. of the 29th IEEE-EMBC Conf., pp. 608-611, 2007. Aniruddha J. Joshi Research Scholar Computer Science and Engineering Department, IIT Bombay Dr. Ashok Bhat Ayurvedic Practitioner Pune Dr. Anand Kulkarni Ph.D. in Chemical U.I.C.T. Mumbai Prof. Sharat Chandean Professor CSE, IIT Bombay Dr. V. K. Jayaraman Scientist NCL, Pune Dr. B. D. Kulkarni Scientist NCL, Pune