⚛️ biophysics

Quantitative Susceptibility Mapping for Differentiating Hydroxyapatite and Calcium Oxalate Breast Calcifications at 3T: A Phantom Study

This phantom study demonstrates that at 3T, combining Quantitative Susceptibility Mapping (QSM) for detecting hydroxyapatite and R2* mapping for detecting calcium oxalate enables the differentiation of benign and malignancy-associated breast microcalcifications from a single multi-echo acquisition.

Original authors: Misak, K., De Vita, E., Clark, C. A., Cashmore, M. T., Walker-Samuel, S.

Published 2026-07-17

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Misak, K., De Vita, E., Clark, C. A., Cashmore, M. T., Walker-Samuel, S.

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Imagine you are a detective trying to solve a mystery inside a crowded room. In the world of medicine, specifically breast imaging, there is a very common puzzle: tiny, invisible specks of calcium, called microcalcifications, often show up on X-rays. These specks are like little breadcrumbs that can signal the start of breast cancer. However, here is the tricky part: about 70 to 80% of the time, these specks are actually harmless "false alarms" caused by benign (non-cancerous) processes. Currently, doctors cannot tell the difference between the dangerous "bad guys" and the harmless "good guys" just by looking at the picture. Because they can't tell them apart, they have to perform biopsies—taking a tiny sample of tissue with a needle—to be sure. This means thousands of women undergo painful and stressful procedures every year just to find out they are perfectly fine.

To solve this, scientists are looking at a special property of matter called "magnetic susceptibility." Think of this as how much a material gets "pulled" or "pushed" by a magnetic field, kind of like how some metals stick to a fridge magnet while others don't. In an MRI scanner, which uses powerful magnets, different types of calcium have slightly different "magnetic personalities." One type, called hydroxyapatite (HA), is the one linked to cancer and has a strong magnetic personality. The other type, calcium oxalate (CaOx), is the harmless kind and has a very weak magnetic personality. The big question is: Can we build a special MRI technique that is sensitive enough to spot the difference between these two tiny magnetic personalities without needing a needle?

This paper is a "phantom study," which is a fancy way of saying the researchers built a fake model to test their ideas before trying them on real people. They created a set of test tubes filled with a jelly-like substance that mimics human breast tissue. Inside these tubes, they hid tiny particles of both the dangerous hydroxyapatite and the harmless calcium oxalate. They then used a 3 Tesla MRI scanner (a very strong magnet, similar to what hospitals use) to take pictures of these fake tumors.

The researchers used a special imaging trick called Quantitative Susceptibility Mapping (QSM). You can think of QSM as a super-sensitive magnetic compass that tries to measure exactly how much each tiny speck is pushing or pulling on the magnetic field. They also looked at something called R2* relaxation, which is like measuring how quickly the signal from the speck "fades away" after the magnet is turned on.

Here is what they found: The magnetic compass (QSM) was excellent at spotting the dangerous hydroxyapatite particles. It saw them clearly in 18 out of 24 attempts. However, it completely missed the harmless calcium oxalate particles; it saw zero of them. This is actually a good thing! It means that if the scanner sees a strong magnetic signal, it's almost certainly the dangerous kind. If it doesn't see a signal, it might be the harmless kind (or nothing at all).

But since QSM missed the harmless kind, the researchers added the second tool: the R2* measurement. This tool was much better at seeing the harmless calcium oxalate, spotting it in 22 out of 24 attempts. By combining the two tools, they created a simple "two-step" classification system. If the magnetic compass sees a strong signal, it's likely the dangerous type. If the compass sees nothing but the fading signal (R2*) is still there, it's likely the harmless type.

The team also ran computer simulations, which they called a "digital twin," to understand why the measurements weren't perfect. They discovered that the main reason the numbers were lower than expected wasn't just because the particles were small, but because of the mathematical rules used to clean up the image. These rules, called "regularization," accidentally smoothed out the tiny signals too much, making the particles look weaker than they really were.

In short, this study shows that at the strength of magnets used in hospitals today, we might be able to tell the difference between dangerous and harmless breast calcifications without a needle. The dangerous kind lights up on the magnetic map, while the harmless kind stays hidden on the map but shows up on the fading signal. While this is just a test with fake jelly and plastic tubes, it suggests that a new, non-invasive way to avoid unnecessary biopsies could be possible in the future.

Technical Summary: Quantitative Susceptibility Mapping for Differentiating Hydroxyapatite and Calcium Oxalate Breast Calcifications at 3T

Problem Statement
Breast microcalcifications are critical early indicators of malignancy, yet current imaging modalities cannot distinguish between malignancy-associated hydroxyapatite (HA) and benign-associated calcium oxalate (CaOx). This inability leads to 70–80% of breast biopsies being unnecessary. While HA and CaOx are both diamagnetic, they possess distinct magnetic susceptibilities (HA: $\Delta\chi \approx -5$ to $-7$ ppm; CaOx: $\Delta\chi \approx 0$ to $-1$ ppm relative to water). Previous studies demonstrated susceptibility differences at ultra-high fields (7 T and 9.4 T) using macroscopic samples, but no study had quantified susceptibility recovery or demonstrated compositional differentiation at the clinical field strength of 3 T, particularly for sub-voxel particles within tissue-mimicking environments. Furthermore, standard Quantitative Susceptibility Mapping (QSM) pipelines, designed for distributed cortical sources in neuroimaging, may not be optimal for focal point-like calcifications in heterogeneous breast tissue due to partial volume effects (PVE) and regularisation biases.

Methodology
The study utilized a custom phantom study design to assess QSM and $R2^*$ relaxation rate mapping at 3 T.

Phantom Fabrication: Twelve phantoms were constructed, each containing co-embedded synthetic HA and CaOx particles (1.0–2.0 mm) within a single 50 mL tube. The particles were embedded in three distinct alginate gel environments: pure alginate, adipose-mimicking, and fibroglandular tissue (FGT)-mimicking. A critical methodological adaptation involved using barium chloride (BaCl $_2$ ) as the crosslinker instead of the standard calcium chloride (CaCl $_2$ ) to prevent the dissolution of CaOx particles by excess Ca $^{2+}$ ions.
Acquisition: Scans were performed on a 3 T Siemens MAGNETOM Prisma scanner using a multi-echo gradient echo (ME-GRE) sequence at two isotropic resolutions (0.70 mm and 0.86 mm).
Processing Pipeline: A custom Python-based QSM pipeline was developed, aligned with the ISMRM QSM Consensus. It included ROMEO for phase unwrapping, V-SHARP for background field removal, and Total Variation (TV)-regularised dipole inversion via ADMM. $R2^*$ maps were generated via voxel-wise mono-exponential fitting of the magnitude data.
Validation: A "digital twin" simulation was constructed at sub-voxel resolution to quantify the specific contributions of PVE and TV regularisation to susceptibility underestimation. This simulation modeled realistic particle morphology (ellipsoids) and compared Level A (local field only) and Level B (full pipeline including air-tissue interfaces) reconstructions against ground truth.

Key Results

QSM Detection: QSM successfully detected HA particles in 18 of 24 measurements (75%), with the highest reliability in adipose-mimicking and pure alginate gels. However, CaOx particles were undetectable in all 24 measurements (0%), with susceptibility values indistinguishable from the surrounding gel.
Susceptibility Recovery: Measured HA susceptibility values recovered only 5–6% of the literature ground truth ($-7.0$ ppm). The digital twin analysis revealed that TV regularisation was the dominant source of signal loss (57.5%), exceeding the loss from partial volume effects alone (24.3%). The interaction between PVE (creating gentle gradients) and TV regularisation (interpreting these gradients as noise) caused compounding signal attenuation.
$R2^*$ Detection: In contrast to QSM, $R2^*$ mapping detected HA in 23/24 measurements (96%) and CaOx in 22/24 measurements (92%). Differentiation was clearest in pure alginate, where low baseline $R2^$ allowed for robust fitting across the echo train.
Classification Strategy: The study established a two-parameter classification scheme:
1. QSM-positive with elevated $R2^*$ : Consistent with HA (malignancy-associated).
2. QSM-negative with moderate $R2^*$ : Consistent with CaOx (benign-associated).
3. Neither elevated: No calcification detected.

Significance and Claims
The paper claims that at 3 T, QSM enables categorical detection of HA microcalcifications, while $R2^*$ provides complementary sensitivity to CaOx. This combination allows for a two-parameter classification of breast calcification composition from a single multi-echo acquisition. The study demonstrates that compositional differentiation is achievable at clinical field strength not through ratio-based quantification (which is compromised by signal loss), but through a categorical detection paradigm where the presence of a focal diamagnetic source indicates HA, and its absence (coupled with $R2^*$ elevation) indicates CaOx.

The authors emphasize that the non-detection of CaOx on QSM is a fundamental sensitivity limitation of TV-regularised dipole inversion for sub-voxel sources at 3 T, rather than a processing error. They conclude that while in-vivo differentiation requires further validation (particularly regarding lipid content and sub-millimetre particle resolution), the reliable detection of HA clusters (1–2 mm) suggests that malignancy-associated calcifications may already be detectable using existing clinical acquisitions. The work also highlights the necessity of non-calcium crosslinkers (like BaCl $_2$ ) for future phantoms involving calcium oxalate in alginate matrices.

Technical Summary: Quantitative Susceptibility Mapping for Differentiating Hydroxyapatite and Calcium Oxalate Breast Calcifications at 3T

More like this