Ultrasound Image Normalisation Software

Development Background In the research of artificial intelligence in medical ultrasound imaging, the original image intensity is affected by factors such as device model, gain setting, and depth compensation, resulting in significant differences in data distribution from different sources, which seriously affects the model's generalization ability. Although open-source libraries such as nibabel and numpy support data normalization, they typically require users to have programming skills and manually implement operations such as percentile cropping and Min Max mapping. At present, there is still a lack of a lightweight tool for ordinary users that is programming free, environment free, and specifically designed for standardizing ultrasound image intensity. To this end, we have developed this software that encapsulates professional normalized logic into independent executable programs, providing a one click operation experience and significantly reducing the threshold for use.  Software Purpose This software is designed specifically for standardizing ultrasound medical imaging and is used to perform the following operations on ultrasound images in NIfTI (. nii. gz) format: 1) Independently calculate the 1% and 99% intensity percentiles for each image; 2) Crop pixel values that exceed this range; 3) Apply Min Max normalization to map the results to the [0,1] interval; 4) Batch processing of ultrasound data within the entire folder, suitable for deep learning dataset construction, multi center data integration, and clinical research preprocessing. Software functions and features 1) Zero environment dependency: No need to install Python or any third-party libraries, just unzip and use; 2) Dynamic adaptive normalization: Independently calculate percentiles for each image to avoid global parameter bias; 3) Retain original spatial information: do not change metadata such as image size, orientation, affine matrix, etc; 4) Output standardization: The result is of type float32, with values strictly limited to [0.0, 1.0]; 5) Batch efficient processing: Automatically traverse all. nii.gz files in the input directory and output normalized results with the same name; 6) Graphic interface: Dark color eye protection theme, easy and intuitive operation, and non-technical background users can also easily get started.