Evolving role of cardiac magnetic resonance in cardiac amyloidosis

Abstract

Cardiac amyloidosis, characterized by misfolded transthyretin (ATTR) or immunoglobulin amyloid light-chain (AL) proteins, has shifted from a rarely diagnosed, near fatal or fatal condition, to a recognized and treatable cause of heart failure. This transformation has been enabled in large part by advances in non-invasive multimodality imaging, with cardiac magnetic resonance (CMR) occupying a prominent diagnostic and prognostic role. Late gadolinium enhancement, native T1 and T2 mapping, and extracellular volume quantification allow detection of amyloid infiltration with high sensitivity, often before overt phenotypic expression. CMR also provides robust and reproducible prognostic information that complements biomarkers and bone-avid scintigraphy. This review summarizes the established CMR diagnostic yield, its role within contemporary diagnostic algorithms for ATTR and AL amyloidosis, the prognostic value of quantitative tissue characterization, and the expanding evidence base for using CMR to monitor response to the new generation of disease modifying therapies. It also examines emerging evidence, including artificial intelligence assisted interpretation, non-contrast diagnostic strategies, extracardiac tissue mapping, and refinement of tissue mapping thresholds, that is likely to shape CMR practice in amyloidosis over the coming years.

Key Words: Cardiac magnetic resonance imaging; Cardiac amyloidosis; Tissue characterization; T1 mapping; T2 mapping; Delayed enhancement; Extracellular volume; Diagnosis; Prognosis

Core Tip: Cardiac magnetic resonance imaging role in cardiac amyloidosis has evolved over time. Initially, cine imaging, chamber quantification, and characteristic delayed enhancement findings were used to identify this condition. With evolving technology, T1, T2 mapping and extra cellular volume have added to diagnostic accuracy. Further data has shown potential of differentiating transthyretin amyloidosis from amyloid light-chain amyloidosis. Extracellular volume and delayed enhancement imaging carry robust prognostic information, and serial extracellular volume is under active investigation as a treatment-monitoring biomarker.

INTRODUCTION

Cardiac amyloidosis (CA) results from deposition of misfolded protein fibrils within the myocardium, producing progressive ventricular wall thickening, diastolic dysfunction, and eventually restrictive physiology and heart failure[1]. The two clinically dominant subtypes are transthyretin amyloidosis (ATTR-CM), which may be wild-type (wtATTR, age related) or hereditary (hATTR, due to pathogenic TTR gene variants), and immunoglobulin light-chain amyloidosis (AL-CM), arising from a plasma cell or lymphoproliferative clonal disorder[1,2]. A third, rarer subtype involving serum amyloid A protein may also result in cardiac involvement in the context of chronic inflammatory disease, but myocardial involvement is extremely rare. Historically considered rare and uniformly fatal, CA is now understood to be substantially underdiagnosed. Autopsy series have revealed wtATTR amyloid deposits in up to 25% of individuals over the age of 80, and prospective screening studies in patients with heart failure with preserved ejection fraction and unexplained left ventricular hypertrophy (LVH) report rates of previously unrecognized CA ranging from 10% to over 20%[1,3].

The therapeutic landscape has changed dramatically over the past decade. Transthyretin stabilizers like tafamidis, approved by the United States Food and Drug Administration in 2019 following the pivotal ATTR-ACT trial, and acoramidis, approved in November 2024 following the ATTRibute-CM trial, and the RNA interference agent vutrisiran (HELIOS-B trial, 2024) have each demonstrated meaningful reductions in mortality and cardiovascular hospitalization in ATTR-CM[4,5]. Meanwhile, AL amyloidosis management has been transformed by anti-plasma cell regimens incorporating daratumumab, with haematological complete response now achievable in a substantial proportion of patients and associated with cardiac organ response and improved survival. This therapeutic progress has in turn increased the clinical urgency of early, non-invasive diagnosis, considering disease modifying benefit appears greatest in earlier stage disease, and of reliable tools capable of tracking disease trajectory and confirming therapeutic benefit over time[4,6]. Cardiac magnetic resonance (CMR) has become indispensable to this effort because of its unique capacity for direct myocardial tissue characterization, independent of geometric assumptions, and its ability to quantify interstitial amyloid burden in absolute terms[1-3].

Several reviews have addressed CMR tissue characterization in CA; this review attempts to consolidate four bodies of evidence that have emerged largely since 2024 and are not yet synthesized together elsewhere: (1) Systematic review and meta-analysis data on artificial intelligence (AI)-assisted CMR interpretation for amyloidosis detection; (2) The serial extracellular volume (ECV) literature evaluating tafamidis treatment response; (3) Newer data on CMR changes in both acoramidis and vutrisiran users; and (4) Extracardiac (hepatic and splenic) ECV mapping as an extension of standard CMR protocols. We also incorporate the 2025 American College of Cardiology (ACC) concise clinical guidance, published after most prior reviews on this subject.

METHODOLOGY

A literature search was conducted in PubMed/MEDLINE and Google Scholar for articles published between January 2016 and June 2026, using combinations of the terms “cardiac amyloidosis”, “cardiac magnetic resonance”, “extracellular volume”, “T1 mapping”, “T2 mapping”, “late gadolinium enhancement”, “ATTR”, “AL amyloidosis”, “tafamidis”, and “artificial intelligence” for this narrative review. Society guideline documents, [ACC, European Society of Cardiology (ESC), multisociety consensus statements] were reviewed directly from publisher sources. Reference lists of retrieved systematic reviews and meta-analyses were hand-searched for additional relevant primary studies. Non-English-language articles and abstract-only publications were excluded. 1 preprint was included due to its unique nature and is highlighted in text. As a narrative rather than systematic review, article selection prioritized methodological quality, recency, and relevance over exhaustive inclusion.

CMR TISSUE CHARACTERIZATION: THE DIAGNOSTIC TOOLKIT

Cine imaging and morphology

Standard cine CMR provides a comprehensive morphologic and functional assessment of both ventricles and atria, and is the starting point for any CA evaluation[1,3]. The classic morphologic findings include concentric biventricular hypertrophy with preserved or mildly reduced left ventricular ejection fraction in earlier disease, normal or small ventricular cavities reflecting restrictive filling physiology, biatrial enlargement, thickened atrioventricular valves, and small pericardial or pleural effusions[1-3]. The right ventricle (RV) is frequently involved, with free-wall hypertrophy and, in advanced disease, impaired function contributing to the biventricular restrictive phenotype. Global longitudinal strain analysis applied to cine images frequently demonstrates an apical sparing pattern, in which relative apical longitudinal strain is preserved compared with basal and mid ventricular segments, a finding first described in echocardiographic series that has since been confirmed on CMR[2,3]. While not entirely specific for CA, given that apical sparing strain can occasionally be seen in other infiltrative conditions, it raises suspicion and warrants tissue characterization even when wall thickness is only borderline elevated[3].

Late gadolinium enhancement

Late gadolinium enhancement (LGE) remains the most widely used and historically established CMR feature of CA[1-3]. The classic appearance is diffuse, circumferential subendocardial enhancement that progresses to a transmural pattern as amyloid burden increases. This enhancement distribution reflects the predilection of amyloid fibrils for the subendocardial interstitium in early disease, with progressive outer-wall involvement correlating with advancing amyloid burden[2,3]. The abnormal extracellular expansion produced by amyloid deposition also alters gadolinium kinetics in a characteristic way: The blood pool loses contrast agent faster than normal myocardium and achieving adequate inversion of normal myocardium on standard look-locker inversion recovery sequences becomes difficult or impossible. The use of phase-sensitive inversion recovery sequences, which are relatively insensitive to the choice of inversion time, is therefore strongly preferred in clinical practice[2,3]. Beyond the left ventricular myocardium, LGE in the atrial walls, right ventricular free wall, interatrial septum, and atrioventricular valves can be observed and provides additional supportive evidence[1]. LGE extent typically quantified as the percentage of left ventricle (LV) myocardial volume, correlates with biomarker-based disease staging (including NT-proBNP and troponin) and independently predicts all-cause mortality across both ATTR and AL subtypes[1,2].

Native T1 mapping

Native T1 mapping measures the longitudinal relaxation time of myocardium without requiring gadolinium administration and is markedly prolonged in CA owing to amyloid infiltration of the interstitium[3,7]. Native T1 values in confirmed CA are typically elevated well beyond normal ranges at both 1.5T (> 1100-1150 ms) and 3T (> 1300-1400 ms), though exact thresholds are vendor- and sequence-specific[1,3,7]. A key clinical advantage of native T1 mapping is its sensitivity at early and low-burden disease stages. T1 elevation is among the first detectable tissue changes in CA, often preceding overt LGE and structural wall thickening[8]. This makes native T1 particularly valuable in the screening and early detection context, where a normal LGE study could otherwise be falsely reassuring. In a study of 125 patients with LVH of varying etiologies, native T1 showed marked elevation in CA compared with hypertensive heart disease and hypertrophic cardiomyopathy, with diagnostic accuracy comparable to or exceeding that of LGE[7]. Because native T1 requires no contrast, it is especially advantageous in patients with chronic kidney disease, a population overrepresented in amyloidosis cohorts, where gadolinium use requires greater caution[2,3].

Extracellular volume fraction

ECV, derived from pre and post contrast T1 values of myocardium and blood pool corrected for haematocrit, directly estimates the interstitial (extracellular) compartment of the myocardium and is regarded as the single most informative quantitative CMR parameter in CA[1,3]. ECV quantifies expansion of the extracellular compartment. In CA this expansion is dominated by amyloid deposition, but edema and interstitial fibrosis also contribute, so ECV is best regarded as the closest available non-invasive surrogate for amyloid burden rather than a direct measurement of it. Normal myocardial ECV is approximately 25%-28%. In CA, values are frequently in the 40%-60% range, and in severe or advanced disease may exceed 60%, a degree of interstitial expansion not seen in other causes of LVH or cardiomyopathy[1,3,9]. ECV elevation mirrors amyloid fibril burden directly and has therefore attracted interest as a surrogate for disease staging, monitoring, and treatment response assessment. Multiple cohort studies and a systematic review and meta-analysis of tissue mapping parameters in CA have confirmed that higher native T1 and higher ECV are each independently associated with increased all-cause mortality, with ECV showing the strongest prognostic signal of any single CMR measurement[9-11]. In a prospective cohort of AL-CM patients, ECV carried a hazard ratio of approximately 2.1 per 10% increment (95%CI: 1.38-3.16) for all-cause mortality after multivariable adjustment, and native T1 a hazard ratio of approximately 2.4 per 100 ms increment (95%CI: 1.38-4.32)[10].

T2 mapping and ancillary sequences

Native T2 mapping is generally not markedly elevated in CA under stable clinical conditions, in contrast to the profound T1/ECV abnormalities, and is therefore used less for primary diagnosis of amyloidosis[1,3]. Its role in this context is to evaluate for superimposed myocardial edema, acute inflammatory cardiomyopathy, or other infiltrative processes (such as cardiac sarcoidosis) that may enter the differential diagnosis of unexplained LVH or restrictive physiology[2,3], however, there is growing evidence that T2 mapping can be used to differentiate between ATTR and AL amyloidosis[12,13]. Experimental techniques being investigated in smaller CA research cohorts include diffusion tensor imaging (DTI) to probe myocardial fibre disarray[14], magnetic resonance (MR) elastography to assess myocardial stiffness, a direct correlate of amyloid-related impaired distensibility, and proton MR spectroscopy. These remain confined to research settings and are discussed further in the emerging evidence section[2,11]. Table 1 summarises tissue characterization in CMR.

Table 1 Core cardiac magnetic resonance imaging tissue characterization parameters in cardiac amyloidosis.

Parameter	Normal range	Cardiac amyloidosis	Clinical utility	Ref.	Strength of supporting evidence
Native T1 (1.5T)	Approximately 950-1000 ms	Markedly elevated, often > 1100-1150 ms	Among the highest sensitivity/specificity of any CMR marker; rises before LGE in early disease	[7,12]	Moderate-High, multiple concordant cohort studies and meta-analyses
Native T1 (3T)	Approximately 1200-1300 ms	Markedly elevated, often > 1300-1400 ms	Same diagnostic role; absolute thresholds are field-strength and sequence dependent	[7,12]	Moderate-High, multiple concordant cohort studies and meta-analyses
ECV	Approximately 25%-28%	Frequently > 40%, can exceed 50%-60%	Strongest single prognostic CMR parameter; tracks amyloid burden and treatment response	[8,9,11]	High, consistently the strongest single prognostic/diagnostic parameter across multiple meta-analyses
LGE pattern	Absent	Diffuse subendocardial → transmural; atrial, RV, valvular involvement	Historic diagnostic anchor; correlates with disease stage and biomarkers	[2,3,15]	High, long established diagnostic anchor across large cohorts
Native myocardial T2	Normal	Usually normal to mildly elevated unless acute decompensation	Helps exclude active myocarditis/edema as an LVH mimic	[13]	Low-Moderate, smaller, more heterogeneous evidence base, primarily single/few-centre studies

Normal and disease-range values shown are representative of published cohorts; absolute thresholds vary by scanner vendor, field strength, and mapping sequence and should be interpreted against each site’s local reference ranges. CMR: Cardiac magnetic resonance; LGE: Late gadolinium enhancement; ECV: Extracellular volume; RV: Right ventricle; LVH: Left ventricular hypertrophy.

DISTINGUISHING ATTR FROM AL AMYLOIDOSIS ON CMR

While the morphologic and tissue-mapping signature of CA is broadly similar across subtypes, several quantitative and temporal differences have been described that may provide diagnostic clues in individual patients[1,3]. The delayed enhancement pattern in the two subtypes tends to be different. ATTR-CM demonstrates more extensive LGE, with transmural enhancement occurring in approximately 90% of ATTR patients compared to only 37% of AL patients[15]. In contrast, AL-CM more commonly shows subendocardial LGE without transmural extension. ATTR patients also uniformly demonstrate right ventricular free wall enhancement, which is less common in AL. The Query Amyloid Late Enhancement scoring system, which quantifies LGE extent including RV involvement, also helps differentiate subtypes[16]. AL-CM tends to present with proportionally greater elevations in native T1 and ECV relative to the absolute degree of wall thickening, reflecting the more rapidly progressive and diffuse interstitial expansion characteristic of light-chain fibril deposition, and carries a substantially worse short-term prognosis at any given ECV compared with ATTR-CM[10,11]. In contrast, ATTR-CM often presents with greater absolute wall thickness and more extensive LGE for a given ECV level, consistent with the more indolent, longer-established amyloid infiltration typical of wild-type disease in particular[1,3]. T2 mapping has also shown evidence to differentiate between the two CA subtypes. AL-CM showed significantly elevated myocardial T2 times compared to ATTR-CM[12,13]. A 2024 study even suggested a cutoff of 61 ms on T2 mapping, greater than 61 ms was likely AL, and less than 61 ms was likely ATTR[17]. It should be emphasized that this 61 ms threshold derives from a single-center cohort and has not been externally validated. T2-based cutoffs should therefore be at this time be regarded as hypothesis generating rather than diagnostic, and native T1 and ECV remain the primary quantitative parameters on CMR for potential subtype discrimination. As with native T1 and ECV, absolute T2 values are influenced by vendor, field strength, and mapping sequence; the cutoff above should not be applied interchangeably across platforms without local validation. Atrial LGE may be proportionally more prominent in AL-CM, and the presence of marked atrial wall enhancement in a patient with a hypertrophic phenotype should raise suspicion for this subtype. Despite these tendencies, CMR findings overlap substantially between subtypes and cannot reliably distinguish ATTR from AL at the individual patient level. CMR findings must therefore always be interpreted in conjunction with bone scintigraphy (technetium-99m pyrophosphate, technetium-99m–labeled 3,3-diphosphono-1,2-propanedicarboxylic acid (DPD), or technetium-99m–labeled hydroxymethylene diphosphonate (HMDP) and comprehensive screening for a monoclonal protein (serum and urine immunofixation electrophoresis and serum free light chain ratio), which together form the backbone of modern non-invasive subtype diagnosis[1,6].

The four principal tissue characterization parameters differ in their relative utility between the subtypes. LGE is present in both, but transmural enhancement occurs in approximately 90% of ATTR-CM compared with 37% of AL-CM, and right ventricular free wall enhancement is near-universal in ATTR-CM, so LGE extent carries greater discriminatory weight in ATTR-CM while remaining strongly prognostic in both[15,16]. Native T1 is elevated in both, but disproportionately so in AL-CM relative to the degree of wall thickening, making it particularly valuable for early detection in AL-CM, in which wall thickness may still be near normal[7,10,11]. ECV is the strongest prognostic parameter in both subtypes, but the prognostic gradient is steeper in AL-CM, where a given value carries a worse short term outcome than the same value in ATTR-CM. ECV is also the parameter of interest for treatment monitoring in ATTR-CM, whereas AL cardiac response is conventionally assessed using NT-proBNP-based criteria[9-11]. T2 shows the most subtype specific behaviour, being significantly higher in AL-CM and underlying the proposed cutoffs discussed above, while in ATTR-CM it is usually normal and serves principally to exclude superimposed pathology[12,13,17].

CMR WITHIN THE CONTEMPORARY DIAGNOSTIC ALGORITHM

Professional society guidance, including the 2025 ACC concise clinical guidance on ATTR-CM[6,18] and the multi society expert consensus recommendations on multimodality imaging in CA[1], positions CMR as a key second tier test once clinical suspicion is raised by echocardiography, electrocardiography, or unexplained heart failure with a hypertrophic or restrictive phenotype. Echocardiographic red flags, including granular sparkling myocardial texture, increased wall thickness out of proportion to electrocardiogram voltages, low-flow low-gradient aortic stenosis, and strain with apical sparing, are typically what prompts CMR referral. Once CMR demonstrates characteristic tissue findings, the diagnostic pathway bifurcates based on the bone scintigraphy and clonal screening results[1,6]. Grade 2-3 cardiac radiotracer uptake in the absence of any detectable monoclonal protein is considered sufficient for a non-biopsy tissue diagnosis of ATTR-CM under current international guidelines[1,6,19]. Any evidence of clonal disease mandates hematologic evaluation and, in most cases, histologic confirmation of amyloid type, via abdominal fat aspirate, minor salivary gland biopsy, or endomyocardial biopsy[6]. The role and comparison of different imaging modalities in CA is summarized in Table 2[20-23]. A simplified educational pathway for CMR role in diagnosis, monitoring and prognosis is summarized in Table 3. This table is not a guideline algorithm. It differs from formal guidance in some respects. The 2025 ACC Concise Clinical Guidance addresses ATTR-CM only and does not cover AL-CM, whereas Table 3 spans both. The 2023 ESC guidelines position bone-tracer scintigraphy rather than CMR as the pivotal subtype-determining investigation and assign contrast-enhanced CMR a Class IIa recommendation for serial follow-up, whereas Table 3 places CMR earlier as the trigger for subtype triage. The 2021 multisociety consensus is organized by imaging technique rather than clinical pathway and issues no sequential algorithm.

Table 2 Comparison of different imaging modalities in cardiac amyloidosis.

Modality	Sensitivity	Specificity	Cohort and reference standard	Primary role	Key limitation
Bone scintigraphy - Perugini grade ≥ 2[19]	90%	97%	1217 patients with suspected cardiac amyloidosis at specialist centres: 857 with histologically proven amyloid (374 by endomyocardial biopsy) and 360 with confirmed non-amyloid cardiomyopathy. Figures are for grade 2-3 uptake alone; adding a negative monoclonal screen raises specificity and positive predictive value to 100% (95%CI: 98.0-100). Addresses discrimination of ATTR from AL in patients already suspected of amyloidosis, not detection in an unselected population	Non-biopsy diagnosis of ATTR-CM	Valid only with negative serum and urine immunofixation and free light chains. Uptake of any grade is more sensitive (> 99%) but far less specific (86%), with false positives almost exclusively from AL-CA
Bone scintigraphy - pooled diagnostic accuracy, all tracers[20]	92.2% (95%CI: 89%-95%)	95.4% (95%CI: 77%-99%)	Bivariate meta-analysis of 6 studies comprising 529 patients with suspected cardiac ATTR. Differs from the row above because it pools unselected cohorts without the negative clonal screen requirement	Screening and diagnosis of ATTR-CM	Accuracy varies by tracer (DPD > PYP > HMDP); wide specificity confidence interval reflects few non-amyloid comparators; poor discriminator for AL-CA
CMR - ECV[21]	90%	91%	Meta-analysis of 18 diagnostic studies comprising 2015 patients	Diagnosis, quantification, prognostication, monitoring	Requires gadolinium; least validated in significant chronic kidney disease. Diagnostic odds ratio 84.6 (95%CI: 30.3-236.2), significantly higher than LGE
CMR - native T1 mapping[21]	87%	88%	Same meta-analysis as the row above; sensitivity and specificity did not differ significantly from LGE	Contrast-free diagnostic alternative	Vendor- and sequence-dependent (MOLLI/ShMOLLI/SASHA); absolute values are not interchangeable across field strengths
CMR - LGE[21]	85%	92%	Same meta-analysis as the two rows above; diagnostic odds ratio 20.1 (95%CI: 9.1-44.1) vs 84.6 (95%CI: 30.3-236.2) for ECV	Diagnosis, prognostication	Qualitative and pattern-based; myocardial nulling becomes difficult in advanced disease
Echocardiography - apical sparing of longitudinal strain[22]	72%	66%	544 patients with confirmed cardiac amyloidosis (375 ATTR, 169 AL) and 200 controls in whom amyloidosis was suspected but excluded, from 12 centres in 4 countries, plus 174 healthy subjects. Diagnosis by endomyocardial biopsy with mass spectrometry, or grade 2-3 Tc99-PYP uptake without monoclonal gammopathy; controls excluded by biopsy or by negative scintigraphy and light chains. Apical sparing ratio cut-off 1.67, AUC 0.74	First-line screening (widely available, low cost)	Specificity, rather than sensitivity, is the principal weakness: Apical sparing was present in 32% of clinically similar control patients and in 6% of healthy subjects, giving a negative predictive value of 0.50. Accuracy is further reduced in chronic kidney disease
Cardiac amyloid PET (11C-PiB, 18F-florbetapir/florbetaben)[23]	78% (91% for 11C-PiB)	95% (97% for 11C-PiB)	Derived from the PET subset of a meta-analysis of 31 articles comprising 37 studies and 2585 patients across CMR, SPECT and PET; the pooled PET figure is not drawn from all 31 articles	Emerging; AL-CA detection, particularly with negative or equivocal scintigraphy	Varies substantially by tracer; the best-performing tracer (11C-PiB) has a 20-minute half-life limiting routine use; not yet in guideline algorithms; no FDA-approved cardiac indication

ATTR-CM: Transthyretin amyloid cardiomyopathy; AL-CM: Immunoglobulin light-chain amyloid cardiomyopathy; ATTR: Transthyretin amyloidosis; AL: Immunoglobulin light-chain amyloidosis; CI: Confidence interval; DPD: ^99mTc-3,3-diphosphono-1,2-propanodicarboxylic acid; PYP: ^99mTc-pyrophosphate; HMDP: ^99mTc-hydroxymethylene diphosphonate; CMR: Cardiac magnetic resonance; ECV: Extracellular volume; LGE: Late gadolinium enhancement; MOLLI: Modified Look-Locker inversion recovery; ShMOLLI: Shortened modified Look-Locker inversion recovery; SASHA: Saturation recovery single-shot acquisition; AUC: Area under the curve; PET: Positron emission tomography; ¹¹C-PiB: Carbon-11 Pittsburgh compound B; ¹⁸F: Fluorine-18; SPECT: Single-photon emission computed tomography; FDA: United States Food and Drug Administration.

Table 3 Simplified educational diagnostic and monitoring pathway incorporating cardiac magnetic resonance.

Step	CMR finding	Next step
1 Unexplained LVH/HFpEF	Concentric LVH, reduced strain, small pericardial effusion	Proceed to tissue characterization
2 Tissue characterization	Diffuse subendocardial/transmural LGE + elevated native T1/ECV	High suspicion for cardiac amyloidosis
3 Subtype triage	Bone-avid tracer scintigraphy (Tc-99m PYP/DPD/HMDP) + serum/urine free light chains	Grade 2-3 uptake + negative clonal screen = ATTR (no biopsy needed); any clonal positivity requires hematologic work-up ± biopsy
4 Baseline severity/prognosis	Quantify ECV, LGE extent, strain, NT-proBNP/troponin	Risk-stratify and stage per ATTR or AL staging systems
5 Therapy and follow-up	Serial CMR (ECV) on disease-modifying therapy	Track treatment response; stable/falling ECV favors disease stabilization

CMR: Cardiac magnetic resonance; LVH: Left ventricular hypertrophy; HFpEF: Heart failure with preserved ejection fraction; LGE: Late gadolinium enhancement; ECV: Extracellular volume; ATTR: Transthyretin amyloidosis; Tc-99m: Technetium-99m; PYP: ^99mTc-pyrophosphate; DPD: 3,3-diphosphono-1,2-propanodicarboxylic acid; HMDP: Hydroxymethylene diphosphonate; NT-proBNP: N-terminal pro-B-type natriuretic peptide.

Society guidelines do have some differences, creating potential variability in real world diagnostic practice. Areas of divergence include the interpretation of Grade 1 scintigraphic uptake, the threshold for advancing to endomyocardial biopsy in the presence of an incidental monoclonal gammopathy of uncertain significance, and the diagnostic terminology used to convey certainty (suspected, probable, or definite amyloidosis)[1]. The integration of advanced modalities, particularly quantitative T1/ECV mapping thresholds and amyloid-specific positron emission tomography tracers, also varies across society documents and is an active focus of harmonization efforts[1,6].

In settings where CMR access is limited, a stepwise approach using clinical suspicion, cardiac biomarkers, electrocardiography, and comprehensive echocardiography (including speckle-tracking strain) is increasingly used to triage patients before committing to advanced imaging, consistent with multi society consensus guidance recommending strain echocardiography as a first line diagnostic and prognostic tool. This approach can reduce unnecessary CMR referrals in resource-constrained environments, though it does not substitute for the tissue characterization specificity that CMR provides once amyloidosis is suspected. Echocardiographic strain patterns such as apical sparing are not entirely specific to CA and are better regarded as a trigger for further evaluation than as a standalone diagnostic replacement for CMR[22]. Where CMR is unavailable, bone-avid scintigraphy with monoclonal protein screening together permits a non-biopsy diagnosis of ATTR-CM without any CMR input[1,6,19]. Where CMR is available but mapping sequences are not, LGE with phase-sensitive inversion recovery remains diagnostically informative, but the absence of native T1 and ECV removes both the contrast-free early detection capability and the strongest single prognostic parameter, so LGE extent and echocardiographic strain must carry the prognostic weight. Where bone scintigraphy is unavailable, subtype assignment cannot be completed non-invasively and tissue diagnosis by abdominal fat aspirate, minor salivary gland biopsy or endomyocardial biopsy becomes necessary in patients with a detectable monoclonal protein or an ambiguous phenotype; CMR can establish the presence and severity of infiltration but cannot substitute for subtype determination. Finally, where disease-modifying therapy is unavailable or unreimbursed, the argument for serial CMR monitoring largely falls away and the priority shifts to accurate initial diagnosis using the least resource-intensive combination of investigations that achieves it.

PROGNOSTIC VALUE OF CMR

Beyond establishing diagnosis, CMR-derived tissue characterization is one of the most robust non-invasive prognostic tools available in CA, providing information complementary to and in some analyses incremental to established biomarker based staging systems[9-11]. A systematic review and meta-analysis of tissue mapping parameters across 21 cohort studies confirmed that higher native T1, higher ECV, and a lower myocardial to skeletal T2 ratio are each associated with significantly increased mortality risk, with ECV demonstrating the strongest and most consistent prognostic signal[9]. In AL-CM, two independent cohort studies enrolling 38 and 89 patients respectively demonstrated that native T1 and ECV remained independently predictive of all-cause mortality after multivariable adjustment for cardiac biomarkers, which themselves form the basis of the Mayo Clinic staging system for AL amyloidosis[10,11]. ECV hazard ratios of approximately 2.1 per 10% increment and native T1 hazard ratios of approximately 2.4 per 100 ms increment suggest that CMR parameters capture dimensions of disease severity not fully reflected by circulating biomarkers alone. In ATTR-CM, LGE extent and ECV similarly track with established disease stage and functional class[1-3]. Emerging data suggest that ECV may identify a subset of patients with low amyloid burden who have near perfect separation from hypertrophic phenocopies [area under the curve (AUC) approximately 0.97 for ECV alone], with lower ECV burden associated with better survival outcomes, raising the possibility that CMR could contribute to identifying candidates who may benefit most from early initiation of disease modifying therapy[8].

CMR parameters have also been proposed as endpoints in clinical trials of novel therapies. Because ECV reflects interstitial amyloid burden directly and changes predictably over time in untreated disease, it provides a plausible surrogate for demonstrating biological activity of anti-amyloid treatments. Incorporating ECV as a co-primary or secondary endpoint in future trials of emerging agents, including RNA interference therapies, gene editing approaches, and amyloid fibril disruptors, is an area of active interest[4,18,24].

CMR FOR THERAPY MONITORING IN THE ERA OF DISEASE-MODIFYING TREATMENT

The approval of multiple disease-modifying therapies for ATTR-CM has created a practical clinical need to assess treatment response over time and to identify patients whose disease continues to progress despite therapy. TTR stabilizers reduce the concentration of amyloidogenic TTR tetramers available for fibril formation but do not directly dissolve existing amyloid deposits; the expected radiologic correlate of effective therapy is therefore stabilization of ECV rather than regression[4,5]. RNA interference agents such as vutrisiran go further by suppressing hepatic TTR production, reducing the substrate for amyloid formation more completely, though whether this translates into measurable ECV reduction over feasible follow-up periods is under active investigation[4].

A 2024 systematic review and meta-analysis pooling data from 246 ATTR-CM patients across six studies provided the most comprehensive evidence to date for CMR-ECV as a tafamidis monitoring biomarker[25]. In tafamidis treated patients, mean ECV change from baseline to follow-up was approximately 0.3% (95%CI: -1.83 to 2.49%), which was not statistically significant, while untreated comparators showed a mean ECV increase of approximately 4.2% (95%CI: 0.44 to 8.02%), which was statistically significant. Follow-up duration across the six included studies ranged from 6 months to 1 year, with five of the six studies using a 1-year interval. These findings support the interpretation that tafamidis halts the otherwise expected progressive accumulation of interstitial amyloid, rather than reversing pre-existing fibril deposition. No significant difference in ECV change was identified between wild-type and hereditary ATTR-CM subgroups.

A complementary cohort study using serial CMR in tafamidis treated ATTR-CM patients published in 2025 found that serial ECV change between baseline and follow-up CMR was the sole CMR or clinical metric independently associated with adverse cardiovascular outcomes in multivariable analysis, outperforming clinical, laboratory, and other imaging metrics[24]. Patients with rising ECV on treatment had substantially worse event free survival than those with stable or declining ECV, positioning serial ECV as a potential tool for identifying inadequate therapeutic response and prompting consideration of treatment escalation or trial enrolment for patients on tafamidis. This positions serial CMR-ECV as a promising biomarker for tafamidis based treatment monitoring, complementing functional measures such as six-minute walk distance and cardiopulmonary exercise testing, and quality of life instruments, all of which are referenced in current ACC clinical guidance[18]. Practical barriers, including the cost of repeat gadolinium enhanced CMR, scanner time, and logistical challenges for frail elderly patients, currently limit serial CMR monitoring outside specialized amyloidosis centres and clinical trial programmes[4,18].

The pivotal trials establishing efficacy of acoramidis (ATTRibute-CM) and vutrisiran (HELIOS-B) in ATTR-CM were powered around functional and biomarker endpoints. However, sub studies and sub analysis have demonstrated CMR changes on treatment. Acoramidis sub study showed favourable trends in LV mass, stroke volume, ejection fraction, and strain. ECV increased modestly in both the acoramidis and placebo groups, however, ECV defined amyloid regression (≥ 5% absolute reduction) was observed in 12.5% of acoramidis treated patients and in none of the placebo-treated patients[26]. The vutrisiran sub analysis showed stronger response. It demonstrated stabilization of myocardial ECV with vutrisiran compared with progressive ECV expansion with placebo, with a modeled treatment effect of 6.56 percentage points. A subset of vutrisiran-treated patients also demonstrated ECV-defined amyloid regression[27].

It is important to distinguish the strength of evidence underlying these applications of CMR. The diagnostic and prognostic value of native T1, ECV, and LGE rests on a substantial and largely concordant body of observational cohort and meta-analytic data. By contrast, the evidence supporting serial ECV as a treatment-monitoring tool is considerably earlier-stage. It derives from a pooled meta-analysis of observational cohorts and one single-centre outcomes cohort, neither of which was designed to test whether acting on serial ECV changes, for example, switching or escalating therapy, improves patient outcomes. Serial ECV-guided treatment adjustment should therefore be regarded as an investigational strategy under active evaluation, appropriate within specialized amyloidosis centres or clinical trial protocols, rather than a validated basis for routine treatment switching decisions in general cardiology practice.

EMERGING EVIDENCE

AI and deep learning

A rapidly growing body of literature applies machine learning and deep learning methods to CA imaging across modalities, fueled by the recognition that CA is underdiagnosed and that automated pattern recognition could accelerate case detection[28-31]. A systematic review and meta-analysis of AI-assisted diagnostic models for CA (preprint) pooled four CMR based deep learning studies and reported a robust pooled AUC of 96.8% (95%CI: 95.1-98.6%) with low heterogeneity and a diagnostic accuracy of 91.7% (95%CI: 89.1-94.3%)[29]. These performance metrics place AI assisted CMR interpretation on par with, and in some comparisons ahead of, AI models trained on echocardiography (AUC approximately 94%), while remaining below the exceptional performance of bone scintigraphy based AI models benefiting from very large training datasets[29].

Deep learning networks trained on LGE and cine images have been used to classify CMR studies as amyloidosis present or absent, sometimes achieving performance comparable to experienced readers[32]. In early proof of concept work conducted outside the amyloidosis field, a progressive sequential causal generative adversarial network was applied to plain cine magnetic resonance imaging (MRI) images to synthesize LGE equivalent images with simultaneous automated cardiac tissue segmentation, achieving an overall segmentation accuracy of approximately 97% and a structural similarity index of 0.78 for synthesized images[33]. This work was done in ischemic heart disease, but the same concept can be applied for amyloidosis for substitution of gadolinium based delayed enhancement imaging. If validated prospectively, such an approach could reduce reliance on gadolinium contrast for diagnostic screening, with particular relevance for patients with renal impairment. Despite these encouraging metrics, AI tools for CMR in CA remain investigational. Multiple methodological barriers also exist. First, training cohorts for published AI-CMR amyloidosis models have generally been drawn from single or few centres, raising concern for selection bias and limited generalizability. Second, most deep-learning classifiers function as ‘black boxes’, post hoc explainability methods such as saliency/Grad-CAM mapping are frequently reported but have not themselves been clinically validated as faithful representations of model reasoning. Third, cross-vendor and cross-sequence generalizability remains a substantial open problem. T1/T2 mapping sequence implementation differs meaningfully across vendors and techniques [e.g., Modified Look-Locker Inversion recovery (MOLLI) and Shortened MOLLI (ShMOLLI), SASHA], and models trained on data from a single vendor or center frequently show performance degradation on external data. Finally, there is minimal published prospective evidence on the real-world workflow impact of AI-assisted interpretation in this setting, including effects on reading time, false-positive referral rates, and downstream testing, all necessary before AI tools move from an investigational to a clinically adopted role. Current systematic reviews and expert commentary explicitly call for prospective, multicentre, and multivendor validation studies, attention to algorithmic equity across demographic subgroups, and clear frameworks for real world clinical deployment before these tools can be integrated into guideline directed practice[28,31].

Non-contrast and abbreviated diagnostic strategies

Because native T1 elevation is detectable without gadolinium, there is growing interest in non-contrast or abbreviated CMR protocols for both initial diagnosis and serial surveillance, reducing procedural risk, scan time, cost, and the practical barriers to repeated imaging[3,7]. The clinical rationale is particularly compelling for patients with chronic kidney disease. Comparative studies report that native T1 alone can achieve diagnostic accuracy approaching that of full LGE and ECV protocols in selected patient populations with established CA and clear clinical phenotypes[3,7]. However, ECV continues to provide incremental prognostic information beyond native T1 alone, and purely non contrast protocols are unlikely to replicate the treatment monitoring capability of serial ECV quantification[9,10]. The optimal strategy may be a risk-stratified approach: Non-contrast T1-based screening at initial evaluation for frail or renally impaired patients, with full contrast-enhanced imaging reserved for cases where ECV based staging or longitudinal treatment monitoring is the appropriate strategy[3].

Extracardiac and whole-body tissue mapping

The systemic nature of both AL and ATTR amyloidosis raises the possibility of extending CMR tissue characterization beyond the myocardium to capture multiorgan amyloid burden within a single imaging session[1,3]. Because the liver and spleen are routinely included within the CMR field of view during cardiac imaging, hepatic and splenic ECV can be calculated from the same pre and post contrast T1 mapping data used for myocardial ECV quantification, without additional scan time[3]. Reference ranges for hepatic ECV (approximately 24%-32%) and splenic ECV (approximately 25%-33%) have been established in small normative cohorts, and preliminary data suggest that hepatic ECV is elevated in systemic amyloidosis, particularly in AL and hATTR forms with prominent liver involvement[3]. This whole organ tissue characterization approach is conceptually appealing as a means of staging systemic disease burden and potentially tracking response to therapy across multiple organ compartments within a single examination. However, clinical validation in adequately powered prospective cohorts and integration into staging algorithms remain at an early stage[1,3].

Refinement of tissue mapping thresholds and field strength standardization

A practical impediment to wider adoption of quantitative T1 and ECV mapping in CA is the absence of universally agreed diagnostic thresholds applicable across different scanner manufacturers, field strengths, and mapping sequences[1,3,7]. Absolute native T1 values are not interchangeable between 1.5T and 3T scanners, nor between MOLLI and ShMOLLI or SASHA sequences, even on the same platform. This limits the direct comparability of results across centres and over time when scanner or protocol changes occur, which is a significant practical problem for multicentre trials aiming to use ECV as a treatment response endpoint. Multi society consensus statements increasingly emphasize the importance of standardized CMR acquisition protocols, site qualification, and centralized image analysis as prerequisites for the reproducibility required to reliably detect small but clinically meaningful ECV changes over the course of therapy[1,6]. Phantom based calibration, cross vendor harmonization studies, and the development of scanner agnostic ECV reference ranges are active areas of methodological work likely to underpin the next generation of CMR based therapeutic trials[1,3].

Experimental tissue and functional techniques

DTI, myocardial MR elastography, and proton MR spectroscopy each offer complementary information about amyloid related myocardial changes not fully captured by standard T1/ECV mapping or LGE[2,11]. DTI measures the directional diffusivity of water molecules within myocardium and can characterize fibre orientation and organisation; amyloid deposition is postulated to disrupt normal helical fibre architecture in ways that may be detectable by DTI before significant T1 or structural changes occur[11]. MR elastography quantifies myocardial stiffness non-invasively by analysing shear wave propagation, providing a direct mechanical correlate of the amyloid mediated increase in myocardial rigidity that underlies diastolic dysfunction in CA. Early feasibility studies are encouraging but sample sizes remain very small. Proton MR spectroscopy can assess myocardial lipid and metabolite content and may provide insight into the metabolic consequences of amyloid infiltration on cardiomyocyte energetics[2,11]. All of these techniques face significant practical barriers, including long acquisition times, motion sensitivity, post processing complexity, and the absence of standardized protocols. They are most likely to remain research tools for the immediate future, though they may contribute important mechanistic insights into disease biology and novel biomarker development[2]. Figure 1 summarises clinical techniques and their readiness for CA.

Figure 1 Clinical readiness of cardiac magnetic resonance techniques in cardiac amyloidosis. Readiness tiers represent the author’s synthesis of the evidence reviewed in this article. They do not correspond to formal recommendation classes or levels of evidence issued by societies. Green: Established clinical use, supported by concordant cohort data and meta-analyses and incorporated into society imaging guidance[1,9,21,34,35]. Amber: Emerging, supported by limited or single-centre evidence and carrying no formal guideline recommendation[3,13,17,24,25,27]. Red: Research use only, confined to small feasibility cohorts[2,11,14].

GUIDELINES

The comprehensive multi society AHA/ACC/SCMR/ASE 2021 expert consensus emphasizes central role of CMR in the non-invasive diagnosis of CA, which includes evaluating cardiac structure, function, and myocardial tissue characterization[34,35]. It also emphasises the role in prognosis as described above, including LGE presence and pattern, native T1, post-contrast T1, and multiple morphologic parameters. The guidelines also mention using LV wall thickness, LV mass, and ECV as emerging tools to assess disease progression and response to therapy[34,35]. European guidelines also offer similar perspective. 2023 ESC Guidelines for the Management of Cardiomyopathies suggests contrast enhanced CMR should be considered for the serial follow-up and assessment of therapeutic response in patients with CA (Class IIa)[36].

Despite overlap, there are key differences among these guidelines. The 2025 ACC Concise Clinical Guidance on transthyretin CA evaluation and management provides a diagnostic algorithm and practical guidance on selecting among the approved ATTR therapies, but does not extend to AL-CA[18]. The 2023 ESC Guidelines address amyloidosis as one phenotype within a broader cardiomyopathy framework, with dedicated screening and diagnosis algorithms. ESC designates bone-tracer scintigraphy [^99mTc-pyrophosphate (PYP)/DPD/HMDP] as the diagnostic gold standard for ATTR-CM and, together with a negative monoclonal screen, permits a non-biopsy diagnosis. The 2021 AHA/ACC/SCMR/ASE Multisociety Expert Consensus is technique oriented rather than disease pathway oriented, providing detailed multimodality imaging guidance spanning echocardiography, nuclear scintigraphy, and CMR for both AL- and ATTR-CA, including acquisition and interpretation standards. None of these three documents currently assigns ECV a formal recommendation class as a therapy monitoring endpoint. ECV remains in all three frameworks, a research and prognostic parameter.

LIMITATIONS OF CMR IN CARDIAC AMYLOIDOSIS

Contrast-based protocols (LGE, ECV) are contraindicated or require caution in advanced renal impairment, a population enriched for amyloidosis; estimated glomerular filtration rate thresholds and shared decision-making frameworks should guide gadolinium use in this group[2,3].

Absolute native T1 and ECV values are field strength, vendor, and sequence dependent, complicating direct comparison across centres and over time if scanner hardware or acquisition protocols change[1,3,7]. ECV quantification requires a concurrently measured hematocrit for blood T1 correction; errors or temporal mismatch between hematocrit sampling and CMR acquisition, or true physiological hematocrit fluctuation between serial scans, can introduce apparent ECV change unrelated to true interstitial amyloid burden.

Despite encouraging evidence, CMR cannot reliably distinguish ATTR from AL amyloidosis in isolation and must always be interpreted alongside bone avid radionuclide scintigraphy and haematologic screening for a monoclonal protein[1,6].

Atrial fibrillation, premature ventricular contractions, and poor breath holding capacity, all prevalent in this elderly, can degrade cine and mapping image quality, reducing diagnostic confidence[3].

Cost, scanner availability, scan time, and the burden of repeated gadolinium administration limit the feasibility of routine serial CMR for treatment monitoring outside specialist amyloidosis centres[4,18].

AI assisted CMR interpretation tools for CA are not yet validated for routine independent clinical use, are predominantly derived from small retrospective single centre cohorts, and require prospective multicentre validation before clinical deployment[28,30,31].

A further limitation is the absence of formal health economic evaluation of serial ECV based monitoring. Guideline directed disease modifying therapy for ATTR-CM disproportionately affects an elderly population with a high prevalence of chronic kidney disease, in whom repeated gadolinium-based contrast administration carries both practical and safety considerations. Current macrocyclic gadolinium-based contrast agents carry substantially lower risk of nephrogenic systemic fibrosis than earlier linear agents and are generally considered acceptable in moderate renal impairment, but caution remains warranted in severe renal impairment, and cumulative tissue retention with repeated dosing remains incompletely characterized. To our knowledge, no published cost-effectiveness analysis or health technology assessment has directly compared serial contrast enhanced CMR against lower-cost, non-nephrotoxic alternatives (serial biomarkers, echocardiographic strain, or non-contrast native T1 mapping) for longitudinal monitoring of treatment response in ATTR-CM. This represents a clear priority for future health-services research as the treated population and treatment duration both expand.

CONCLUSION

Cardiac MRI has evolved from an adjunctive research tool into an important diagnostic and prognostic tool for CA, providing structural, functional, and direct myocardial tissue characterization information that cannot be obtained by any other non-invasive modality[1,3,9]. LGE, native T1 mapping, and ECV quantification are complementary techniques: LGE provides robust visual pattern recognition and extent quantification; native T1 offers gadolinium free sensitivity for early disease; and ECV delivers the most precise and clinically actionable estimate of amyloid burden for staging and treatment monitoring[2,3,9]. The role of CMR is expanding in parallel with the therapeutic landscape. Serial ECV quantification is emerging as the most promising CMR biomarker for documenting treatment response to TTR stabilizers and silencers, with meta-analytic evidence now supporting its ability to distinguish treated from untreated disease trajectories[18,24], though its implementation as a routine monitoring tool is contingent on standardization of acquisition protocols, reduction of practical barriers, and prospective validation in larger, multicenter cohorts[6,18]. Also worth noting that no study has yet tested whether altering therapy on the basis of serial ECV improves patient outcomes, and serial ECV-guided treatment adjustment therefore remains investigational. AI, non-contrast diagnostic strategies, extracardiac tissue mapping, and advanced microstructural imaging techniques represent the frontier of CMR innovation in amyloidosis[28-30]. Each holds promise for expanding the diagnostic reach, clinical accessibility, and mechanistic depth of CMR in this disease, but each also requires the rigorous prospective validation and methodological standardization that will ultimately determine whether they translate from research settings into guideline directed practice[6,28].

Looking ahead, numerous actionable research projects can be explored to further consolidate current and emerging findings. These includes: (1) Establishment of standardized ECV thresholds and minimum follow-up intervals required for ECV to be validated as a surrogate clinical trial endpoint, supported by prospective correlation with hard clinical outcomes; (2) Design and execution of prospective, multi center, non-inferiority trials directly comparing non-contrast/abbreviated CMR protocols against standard gadolinium-based protocols for both diagnostic and monitoring indications; and (3) Development of regulatory validation benchmarks for AI-assisted CMR interpretation tools, including defined multicenter, multivendor performance standards and post-market surveillance requirements, prior to incorporation into clinical guideline pathways.