We propose optimisation based algorithm using wavelet decomposition for segmentation of echocardiography images. The objective of the proposed method is to find an optimal threshold value for segmentation. This threshold value is used for separating the left ventricle by segmenting wavelet decomposed coefficients. The proposed method evaluates the optimal value of threshold using a nonlinear derivative free optimising algorithm. Its objective function is the contrast property of GLCM. The contrast would be highest at the correct position of the boundary between image regions. Accordingly, optimization task is formulated as a maximization problem. The proposed method is an iterative process for finding optimum threshold. Segmentation is carried out using this optimised threshold value. Further, the exact left ventricle contour is extracted using morphological operations and user provided location input. The proposed method is observed to be performing better and found to have 94% accuracy compared to ground truth labels.

中文翻译：

---

基于小波分解阈值的半自动数值优化方法超声心动图图像分割

我们提出了基于优化的算法，使用小波分解来分割超声心动图图像。所提出方法的目标是找到一个最佳的分割阈值。该阈值用于通过分割小波分解系数来分离左心室。所提出的方法使用非线性无导数优化算法来评估阈值的最佳值。它的目标函数是GLCM的对比度属性。在图像区域之间边界的正确位置处对比度最高。因此，优化任务被表述为一个最大化问题。所提出的方法是寻找最佳阈值的迭代过程。使用这个优化的阈值进行分割。更远，使用形态学操作和用户提供的位置输入来提取准确的左心室轮廓。所提出的方法被观察到表现更好，并且与地面实况标签相比具有 94% 的准确率。