Software for Evaluating Medical NifTI Image Label Processing and Segmentation Performance

Development background: In the research of medical imaging artificial intelligence, researchers often face two major challenges: firstly, the original segmentation label data may have inconsistent label IDs, require batch modification, or require understanding of the volume distribution of various structures; Secondly, after model development, there is a lack of a programming free and integrated tool to comprehensively and quantitatively evaluate segmentation results. Although there are open source libraries such as nibabel and medpy that support related calculations, they usually require users to have strong programming skills and manually write scripts to implement the complete process from file reading to indicator calculation, and then to result summary. At present, there is a lack of a lightweight desktop tool that integrates tag management and performance evaluation for ordinary researchers and clinical doctors. To this end, we have developed this software, which encapsulates the above professional analysis logic into an independent graphical program, providing an intuitive and efficient one click operation experience, significantly reducing the technical threshold. Software Usage: This software is designed specifically for medical image segmentation tasks and is used to perform the following operations on segmentation labels in NIfTI (. nii,. nii. gz) format: 1) Analysis: Batch check the existence of specific labels and generate detailed reports containing physical volume (mm ³); 2) Preprocessing: Batch replace specified label values with new values to solve label inconsistency issues; 3) Evaluation: Compare the predicted segmentation results with the Ground Truth standard, calculate multiple authoritative performance indicators including Dice, Recall, and HD95, and automatically generate statistical summaries. Software functions and features: 1) Dual module integration: integrating the two core functions of "label analysis and remapping" and "segmentation performance evaluation" into a single interface, meeting the full process requirements from data preparation to result verification; 2) Zero programming dependency: Provides a graphical user interface (GUI) that allows users to complete complex data processing and analysis tasks without writing any code; 3) Accurate physical units: All volume calculations are based on voxel spacing in the NIfTI file header, with results measured in mm ³, in compliance with medical standards; 4) Comprehensive evaluation indicators: supporting the six core indicators recognized by the computing industry (Dice, IoU, Recall, Precision, RVE, HD95), providing multidimensional insights into model performance; 5) Intelligent statistical summary: Automatically generate "SUMMAY" rows for each evaluation label, and intelligently select "mean ± standard deviation" or "median [interquartile range]" for description based on data distribution; 6) Easy export of results: All analysis reports and evaluation results can be easily exported as CSV files with just one click, making it convenient for archiving, sharing, and subsequent statistical analysis.