NeuroTrace AI: Intelligent Fetal Brain Plane Identification with Explainable Decision Support and Secure Medical Record Traceability

Abstract

Accurate identification of standard fetal brain ultrasound planes is essential for prenatal diagnosis and early detection of neurological abnormalities. However, manual interpretation of ultrasound images is often challenging due to image noise, low contrast, and dependence on operator expertise. This paper presents NeuroTrace AI, an intelligent framework for automated fetal brain plane identification using deep transfer learning, Explainable AI decision support, and secure medical record traceability. The proposed system employs a VGG19-based transfer learning model to classify fetal brain ultrasound images into four categories: Trans-cerebellum, Trans-thalamic, Trans-ventricular, and Other. Image preprocessing and data augmentation techniques are applied to improve model robustness and generalization across diverse ultrasound imaging conditions. Following classification, an Explainable AI module generates clinically meaningful explanations to support healthcare professionals in understanding the predicted anatomical plane and its diagnostic relevance. To enhance security and transparency, blockchain technology is integrated to store prediction outcomes, confidence scores, timestamps, and associated metadata in a tamper-resistant manner, ensuring data integrity and traceability. A web-based interface enables real-time image upload, automated analysis, and report generation for clinical use. Experimental evaluation demonstrates that the proposed framework achieves reliable classification performance while improving interpretability, security, and trustworthiness. NeuroTrace AI offers an intelligent and scalable solution for AI-assisted prenatal fetal brain assessment and supports more consistent clinical decision-making.