IEM-CNN Mammographic Calcification Dataset and Experimental Results

This repository contains the complete dataset, source code, trained models, preprocessing outputs, and experimental results generated during the development and validation of the Clinical Stability Index for Cross-Platform Deployment of Convolutional Neural Networks in Mammography (IEM-CNN) framework. The repository includes the original mammographic image patches, preprocessed and normalized datasets, data partitioning files, MATLAB (FP64) reference models, TensorFlow (FP32) exported models, TensorFlow Lite optimized models (FP32, FP16, and INT8), inference outputs, confusion matrices, performance metrics, Bland–Altman analyses, and the proposed IEM-CNN quantitative evaluation results. The primary objective of this repository is to ensure the complete reproducibility of the experimental workflow, from image preprocessing and model training to model conversion, quantization, deployment, and cross-platform validation. All files have been organized to facilitate independent verification of the reported results and to support future benchmarking, comparative studies, and methodological extensions in artificial intelligence for breast cancer diagnosis. This dataset accompanies the scientific manuscript entitled "Clinical Stability Index for Cross-Platform Deployment of Convolutional Neural Networks in Mammography" and has been made publicly available to promote transparency, reproducibility, and open scientific research in medical imaging and AI-assisted breast cancer diagnosis.