Optimization of Radiation Dose in Pediatric CT Imaging: A Dose Reduction Framework Using AI-Based Protocol Customization

Project Abstract

This study presents a comprehensive framework to reduce radiation exposure in pediatric computed tomography (CT) by leveraging artificial intelligence (AI) for protocol customization without compromising diagnostic accuracy. Recognizing the heightened radiosensitivity of children and the cumulative lifetime risk associated with ionizing radiation, the research aims to develop an end-to-end pipeline that personalizes scanning parameters based on patient-specific factors, clinical indication, and image quality requirements. The framework integrates a multi-source dataset comprising pediatric CT scans, corresponding dose metrics, and clinical outcomes from multiple institutions to ensure diversity across age, weight, height, and development stages. We implement a supervised learning model capable of predicting optimal tube current (mA), tube potential (kVp), pitch, and scan range that achieve diagnostic-quality reconstructions at the lowest feasible dose. The AI system is trained with a dual objective minimizing effective dose while preserving image metrics critical for pediatric radiology, including lesion conspicuity, organ delineation, and noise-resolution trade-offs. To address potential biases and variability in scanner hardware, domain adaptation techniques are employed to generalize across different CT platforms and reconstruction algorithms. A core component of the framework is an automated protocol recommender that translates AI predictions into vendor-specific protocol settings, with built-in safety checks and fail-safes to prevent radiologist workflow disruption. The methodology also encompasses advanced denoising and reconstruction strategies, such as model-based iterative reconstruction and deep learning-enhanced image reconstruction, to further enhance image quality at reduced doses. The study evaluates performance through a multi-criterion analysis quantitative image quality metrics (signal-to-noise ratio, contrast-to-noise ratio, modulation transfer function), diagnostic accuracy in simulated and real-world cases, and cumulative dose reduction across representative pediatric cohorts. A prospective validation phase is conducted in a controlled clinical environment to compare AI-driven protocols against conventional pediatric CT protocols across common indications, including chest and abdomen imaging for trauma, infectious processes, congenital anomalies, and oncologic follow-up. The results are analyzed for statistical significance, robustness to motion and pediatric cooperation, and potential impact on workflow efficiency. The health physics perspective includes dosimetric assessments aligned with the ALARA principle, dose-length product (DLP) tracking, effective dose estimation, and patient-specific risk modeling to quantify expected reductions. The study also contemplates ethical and regulatory considerations, data privacy, and interoperability with existing hospital information systems. Findings are expected to demonstrate substantial dose reductions in pediatric populations while maintaining, or even enhancing, diagnostic confidence and lesion detectability. The proposed framework aspires to establish a scalable, reproducible approach for pediatric CT imaging that can be integrated into routine clinical practice, promoting safer imaging without compromising the quality of care. This research contributes to the development of AI-assisted, dose-optimized imaging paradigms and provides a foundation for future translational studies and multi-center implementations aimed at standardizing pediatric CT dose reduction.

Project Overview

What This Project Is About

---

A plain-language overview of the topic and what the project investigates.

---

---

The Problem It Addresses

---

What problem or gap this project tackles and why it matters to the field or society.

---

---

Objectives of the Project

---

1. Reduce unnecessary radiation exposure in pediatric CT scans.
2. Develop AI-based guidelines to tailor scan protocols to a child’s size and condition.
3. Evaluate how protocol customization affects image quality and diagnostic accuracy.
4. Propose a practical framework for clinic adoption and safety compliance.

---

---

What You Will Do Step by Step

---

1. Review existing pediatric CT protocols and dose-reduction techniques.
2. Collect and anonymize pediatric scan data with different protocol settings.
3. Train a simple AI model to suggest protocol adjustments based on patient data.
4. Test the model on retrospective cases to compare dose and image quality outcomes.
5. Validate findings with radiology experts and assess practicality in clinics.

---

---

Expected Outcome

Anticipated results include a demonstrable reduction in radiation dose without compromising diagnostic usefulness, plus a clear set of steps for implementing AI-guided protocol customization in real-world pediatric imaging settings.