Source code for ultrasound_metrics.utils.phantom_utils

"""
Phantom utilities for ultrasound image analysis.

This module provides functions for loading and positioning digital phantoms
for use in ultrasound image registration and analysis.
"""

from pathlib import Path
from typing import Optional

import cv2
import numpy as np
from numpy.typing import NDArray
from skimage.transform import resize

# Pixel sizes below were determined empiricallly by aligning with US images


PHANTOMJPG_X_PX_SZE_MM = 0.102



PHANTOMJPG_Y_PX_SZE_MM = 0.105



def get_phantom_jpg(px_sze_mm: Optional[float] = None) -> tuple[NDArray[np.floating], dict[str, int]]:
    """
    Get digital phantom stored as jpg image.

    Parameters
    ----------
    px_sze_mm
        Desired pixel size in mm.

    Returns
    -------
    tuple
        Tuple of (phantom_image, dimensions_dict).
    """
    phantom_path = Path(__file__).parent / ".." / "assets"
    img = cv2.imread(str(phantom_path / "ATS359.jpg"), cv2.IMREAD_GRAYSCALE)
    phantom = np.array(img)

    nX = len(phantom[0])
    nY = len(phantom)

    if px_sze_mm is not None:
        resize_ratio_x = PHANTOMJPG_X_PX_SZE_MM / px_sze_mm
        resize_ratio_y = PHANTOMJPG_Y_PX_SZE_MM / px_sze_mm
        print(f"Resize ratio: {resize_ratio_x},{resize_ratio_y}")
        new_nX = int(nX * resize_ratio_x)
        new_nY = int(nY * resize_ratio_y)

        resized_phantom = resize(phantom, output_shape=(new_nY, new_nX))
    else:
        resized_phantom = phantom
        new_nX = nX
        new_nY = nY

    return resized_phantom, {"x": new_nX, "y": new_nY}


def find_position(
    us_image: NDArray[np.floating], phantom: NDArray[np.floating]
) -> tuple[NDArray[np.floating], dict[str, np.integer]]:
    """
    Find position of the targets in the US image on the digital phantom.

    Parameters
    ----------
    us_image
        Ultrasound image array.
    phantom
        Phantom image array.

    Returns
    -------
    tuple
        Tuple of (correlation_map, position_dict).
    """
    us_image = np.array(us_image)
    phantom = np.array(phantom)
    us_im_sze = (len(us_image[0]), len(us_image))
    corr = np.array(cv2.filter2D(phantom, ddepth=-1, kernel=us_image), dtype=float)
    max_lin_idx = np.argmax(corr)
    y, x = np.unravel_index(max_lin_idx, corr.shape)
    xpos = x - round(us_im_sze[0] / 2)
    ypos = y - round(us_im_sze[1] / 2)
    return corr, {"x": xpos, "y": ypos}