Initial commit

Sources/ECGSynKit/ECGSyn.swift

@@ -0,0 +1,105 @@
+import Foundation
+import OdeInt
+import RealModule
+
+public struct ECGSyn {
+    public struct Attractor {
+        /// Angle of attractor in radians
+        public let θ: Double
+        /// Position of extremum above or below the z=0 plane.
+        public let a: Double
+        /// Width of the attractor.
+        public let b: Double
+        /// Angle rate factor adjustment `θ * pow(hrMean / 60.0, θrf)`
+        public let θrf: Double
+
+        public init(θ: Double, a: Double, b: Double, θrf: Double = 0.0) {
+            self.θ = θ
+            self.a = a
+            self.b = b
+            self.θrf = θrf
+        }
+
+        public init(deg: Double, a: Double, b: Double, θrf: Double = 0.0) {
+            self.init(θ: deg * .pi / 180, a: a, b: b, θrf: θrf)
+        }
+
+        static func make(deg: Double, _ a: Double, _ b: Double, _ θrf: Double = 0.0) -> Attractor {
+            Attractor(deg: deg, a: a, b: b, θrf: θrf)
+        }
+    }
+
+    public struct Parameters {
+        /// The ECG amplitude in mV.
+        public let range: (Double, Double) = (-0.4, 1.2)
+
+        /// Amplitude of the noise.
+        public let noiseAmplitude: Double = 0.0
+
+        /// Descriptors of the extrema/attractors for the dynamical model.
+        public let attractors: [Attractor] = [
+            .make(deg: -70, 1.2, 0.25, 0.25),
+            .make(deg: -15, -5.0, 0.1, 0.5),
+            .make(deg: 0, 30, 0.1),
+            .make(deg: 15, -7.5, 0.1, 0.5),
+            .make(deg: 100, 0.75, 0.4, 0.25),
+        ]
+    }
+
+    public static func generate(params: Parameters, rrSeries: ECGSynRRSeries<Double>) -> [Double] {
+        var rng = rrSeries.rng
+        let srInternal = rrSeries.timeParameters.srInternal
+
+        let hrSec = rrSeries.timeParameters.hrMean / 60.0
+        let hrFact = sqrt(hrSec)
+
+        // adjust extrema parameters for mean heart rate
+        let ti = params.attractors.map { $0.θ * pow(hrSec, $0.θrf) }
+        let ai = params.attractors.map { $0.a }
+        let bi = params.attractors.map { $0.b * hrFact }
+
+        let fhi = rrSeries.rrParamaters.fhi
+
+        let nt = rrSeries.count
+
+        let dt = 1.0 / Double(srInternal)
+        let ts = (0 ..< nt).map { Double($0) * dt }
+        let x0 = SIMD3<Double>(1.0, 0.0, 0.04)
+
+        let result = SIMD3<Double>.integrate(over: ts, y0: x0, tol: 1e-6) { x, t in
+            let ta = atan2(x[1], x[0])
+
+            let r0 = 1.0
+            let a0 = 1.0 - sqrt(x[0] * x[0] + x[1] * x[1]) / r0
+
+            let w0 = 2 * .pi / rrSeries.valueAt(t)
+
+            let zbase = 0.005 * sin(2 * .pi * fhi * t)
+
+            var dxdt = SIMD3<Double>(a0 * x[0] - w0 * x[1], a0 * x[1] + w0 * x[0], 0.0)
+
+            for i in 0 ..< ti.count {
+                let dt = remainder(ta - ti[i], 2 * .pi)
+
+                dxdt[2] += -ai[i] * dt * exp(-0.5 * (dt * dt) / (bi[i] * bi[i]))
+            }
+            dxdt[2] += -1.0 * (x[2] - zbase)
+
+            return dxdt
+        }
+
+        // extract z and downsample to output sampling frequency
+        var zresult = stride(from: 0, to: nt, by: rrSeries.timeParameters.decimateFactor).map { result[$0][2] }
+
+        let (zmin, zmax) = zresult.minAndMax()!
+        let zrange = zmax - zmin
+
+        // Scale signal between -0.4 and 1.2 mV
+        // add uniformly distributed measurement noise
+        for i in 0 ..< zresult.count {
+            zresult[i] = (params.range.1 - params.range.0) * (zresult[i] - zmin) / zrange + params.range.0
+            zresult[i] += params.noiseAmplitude * (2.0 * rng.nextDouble() - 1.0)
+        }
+        return zresult
+    }
+}