# Head CT dose varied 3.4-fold across protocols

> A registry study found large dose differences among commonly used head-CT techniques, pointing to protocol optimization rather than a universal lowest-dose setting.

_Source: European Radiology study, checked against PubMed PMID 42684484, Crossref, Springer full text, and the originating Telegram post · 2026-09-14 · 5 min read · Verified against primary sources_

Canonical: https://iyu.app/e/head-ct-dose-cluster-analysis

## The 60-second version

A 904,209-scan registry found more than two-fold variation in size-adjusted head-CT dose across commonly used adult and pediatric protocols.

**Key points**

- The analysis separated adults from children and helical from axial scans, then clustered acquisition parameters with unsupervised machine learning.
- Mean adjusted dose varied by more than two-fold in all four groups and up to 3.4-fold; adult helical scans ranged from 372 to 998 mGy·cm.
- Lower-dose clusters often used lower tube current or voltage, but different parameter combinations could work.
- This is an observational dose study, not proof that one lowest-dose setting fits every diagnostic task.

**Verdict.** The strongest implication is for protocol review and standardization, with image quality and clinical judgment preserved.

## Full explainer

> **i** 这项研究分析的是扫描协议和辐射剂量指标，不是患者发生癌症或获得诊断的临床结局。低剂量方案必须仍能满足具体诊断任务。


### What was measured — A registry of 904209 head CT scans

Researchers analyzed **904,209** routine head CT scans recorded between 2015 and 2021: 864,182 from adults and 40,027 from children. The scans came from 131 facilities in seven countries. Adults and children were separated, as were helical and axial acquisitions, creating four comparison groups.


### What the model found — Dose varied within familiar techniques

K-means clustering grouped scans by tube current, voltage, collimation, scan length and, for helical scans, pitch. After adjustment for patient head size, the mean dose differed by more than two-fold in every group and by up to 3.4-fold. In adult helical scans, the dose-length product ranged from 372 to 998 mGy·cm across clusters.

- **904,209** — head CT scans analyzed
- **131** — facilities across 7 countries
- **2.7×** — adult-helical adjusted DLP range


### How to read it — Optimization is a protocol problem

Lower-dose clusters generally used lower tube current or voltage, but no single setting won everywhere. The study suggests that departments can compare their routine protocols, find high-dose outliers and standardize well-tested lower-dose approaches while checking that image quality remains adequate.

- **Study design:** Retrospective scan-level analysis of an international CT dose registry.
- **Population:** 864,182 adult scans and 40,027 pediatric scans; not a treatment trial.
- **Main metric:** Patient-size-adjusted dose-length product and CTDIvol.
- **Clinical boundary:** Radiation metrics do not establish diagnostic equivalence or future cancer risk.

> **⚑ Caveat:** The registry does not prove that the lowest-dose cluster is appropriate for every patient, scanner or clinical question. About 11.9% of examinations were excluded for missing or problematic data, and image quality was inferred from routine use rather than tested as a randomized outcome.

> The actionable finding is not one magic dose. It is the large, reviewable spread among protocols that are already in routine use.


### What to do with it — Keep the scan justified and optimized

Patients should not treat the study as a reason to refuse a medically indicated CT or to demand one universal setting. The useful question is whether the radiology service reviews protocol variation and optimizes exposure for the diagnostic task, patient size and scanner.


## Primary sources

- [Originating Telegram post 1479](https://t.me/CNSmydream/1479)
- [European Radiology paper, DOI 10.1007/s00330-026-12842-8](https://doi.org/10.1007/s00330-026-12842-8)
- [PubMed PMID 42684484](https://pubmed.ncbi.nlm.nih.gov/42684484/)
- [Springer full text](https://link.springer.com/article/10.1007/s00330-026-12842-8)

---
_Published by iyu (https://iyu.app) — the day's AI news, checked against primary sources and rewritten in plain language. Free to quote with attribution and a link to the canonical URL._