Join the Competition

# ct_baseline_algorithm.py — key reconstruction logic

def fbp_reconstruct(sinogram, angles, img_size=None):
    """Filtered Back-Projection reconstruction."""
    n_angles, n_det = sinogram.shape
    # 1. Apply ramp filter in frequency domain
    freqs = np.fft.fftfreq(n_det)
    ramp = np.abs(freqs)
    for i in range(n_angles):
        proj_fft = np.fft.fft(sinogram[i])
        filtered[i] = np.real(np.fft.ifft(proj_fft * ramp))
    # 2. Back-project at each angle
    for i, angle in enumerate(angles):
        rotated = ndrotate(proj_img, angle, ...)
        recon += rotated
    return recon * np.pi / (2 * n_angles)

def estimate_center_offset(sinogram):
    """Compare 0° and 180° projections to find shift."""
    ...

# mri_baseline_algorithm.py — key reconstruction logic

def iterative_soft_threshold(y_kspace, mask, n_iter=30):
    """ISTA: iterative soft-thresholding with data consistency."""
    kspace_mag = np.expm1(y_kspace) * mask
    x = zero_filled_ifft(y_kspace, mask)
    for _ in range(n_iter):
        kx = np.fft.fftshift(np.fft.fft2(x))
        # Data consistency: keep measured k-space lines
        kx_dc = kx * (1 - mask) + kspace_mag * mask
        x = np.abs(np.fft.ifft2(np.fft.ifftshift(kx_dc)))
        x = soft_threshold(x, lam=0.005)
    return x

def estimate_b0_inhomog(x_hat):
    """Estimate B0 from spatial intensity variation."""
    ...