Research ArticleClinical Investigation

Deep Learning–Enabled Quantification of 99mTc-Pyrophosphate SPECT/CT for Cardiac Amyloidosis

Robert J.H. Miller, Aakash Shanbhag, Anna M. Michalowska, Paul Kavanagh, Joanna X. Liang, Valerie Builoff, Nowell M. Fine, Damini Dey, Daniel S. Berman and Piotr J. Slomka
Journal of Nuclear Medicine July 2024, 65 (7) 1144-1150; DOI: https://doi.org/10.2967/jnumed.124.267542

Abstract

Transthyretin cardiac amyloidosis (ATTR CA) is increasingly recognized as a cause of heart failure in older patients, with 99mTc-pyrophosphate imaging frequently used to establish the diagnosis. Visual interpretation of SPECT images is the gold standard for interpretation but is inherently subjective. Manual quantitation of SPECT myocardial 99mTc-pyrophosphate activity is time-consuming and not performed clinically. We evaluated a deep learning approach for fully automated volumetric quantitation of 99mTc-pyrophosphate using segmentation of coregistered anatomic structures from CT attenuation maps. Methods: Patients who underwent SPECT/CT 99mTc-pyrophosphate imaging for suspected ATTR CA were included. Diagnosis of ATTR CA was determined using standard criteria. Cardiac chambers and myocardium were segmented from CT attenuation maps using a foundational deep learning model and then applied to attenuation-corrected SPECT images to quantify radiotracer activity. We evaluated the diagnostic accuracy of target-to-background ratio (TBR), cardiac pyrophosphate activity (CPA), and volume of involvement (VOI) using the area under the receiver operating characteristic curve (AUC). We then evaluated associations with the composite outcome of cardiovascular death or heart failure hospitalization. Results: In total, 299 patients were included (median age, 76 y), with ATTR CA diagnosed in 83 (27.8%) patients. CPA (AUC, 0.989; 95% CI, 0.974–1.00) and VOI (AUC, 0.988; 95% CI, 0.973–1.00) had the highest prediction performance for ATTR CA. The next highest AUC was for TBR (AUC, 0.979; 95% CI, 0.964–0.995). The AUC for CPA was significantly higher than that for heart-to-contralateral ratio (AUC, 0.975; 95% CI, 0.952–0.998; P = 0.046). Twenty-three patients with ATTR CA experienced cardiovascular death or heart failure hospitalization. All methods for establishing TBR, CPA, and VOI were associated with an increased risk of events after adjustment for age, with hazard ratios ranging from 1.41 to 1.84 per SD increase. Conclusion: Deep learning segmentation of coregistered CT attenuation maps is not affected by the pattern of radiotracer uptake and allows for fully automatic quantification of hot-spot SPECT imaging such as 99mTc-pyrophosphate. This approach can be used to accurately identify patients with ATTR CA and may play a role in risk prediction.

Transthyretin cardiac amyloidosis (ATTR CA) is an increasingly recognized cause of heart failure (1). Bone scintigraphy imaging with tracers such as 99mTc-pyrophosphate has been established as a highly accurate tool for diagnosing ATTR CA, with initial studies describing evaluation of planar images alone (2). However, subsequent studies have shown potential pitfalls from using this approach (3). Visual evaluation of SPECT images, in combination with screening for light chain amyloidosis, has now become the gold standard for noninvasive diagnosis (4). SPECT imaging improves diagnostic accuracy but is inherently subjective and requires expertise.

Hybrid SPECT/CT imaging facilitates more robust methods to quantify myocardial radiotracer activity. Scully et al. demonstrated high diagnostic accuracy for quantitation using SUVs from regions placed manually on SPECT/CT images (5). In our previous work, we demonstrated that quantitative measures have high diagnostic accuracy (6) and correlate with native myocardial T1 (7). However, these techniques have not translated into clinical practice, at least in part because they require time-consuming manual segmentation. Recent developments in foundational deep learning models have made it feasible to segment cardiac chambers and ventricular myocardium from noncontrast, ungated CT images (8). These algorithms can be leveraged to automate and objectify the manual processes involved in quantifying radiotracer uptake in the myocardium in a robust manner that is not affected by the tracer uptake pattern. We developed this automated workflow for 99mTc-pyrophosphate uptake quantification and then evaluated the potential clinical utility of different approaches to quantitation.

MATERIALS AND METHODS

Study Population

In this retrospective study, we identified patients who underwent 99mTc-pyrophosphate SPECT/CT imaging at the University of Calgary between March 2020 and October 2022. For each patient with more than one 99mTc-pyrophosphate study, only the first study was included. Additionally, we excluded 7 cases with significant registration errors identified during quality control processes. Significant offset in any direction between rib or spine uptake and anatomic landmarks was considered misregistration. This study was reviewed and approved by the Institutional Review Board at the University of Calgary (REB19-1448), and the requirement to obtain informed consent was waived.

Clinical Data

Demographics, past medical history, and red flags for ATTR CA at the time of imaging were determined through manual review of electronic medical records. Diagnosis of ATTR CA was based on the presence of, first, an endomyocardial biopsy positive for ATTR; second, diffuse myocardial uptake on SPECT/CT imaging after exclusion of monoclonal protein; or third, evidence of infiltrative cardiomyopathy on echocardiography or cardiac MRI with either a documented pathologic gene variant or a noncardiac biopsy demonstrating ATTR amyloidosis (4,9). One patient was diagnosed on the basis of the third criterion; this patient had undergone carpal tunnel biopsy (10) and showed mildly reduced biventricular function on cardiac MRI (genetic testing pending). At our center, coverage criteria for tafamidis initially required pathologic confirmation of disease, but this requirement was eventually removed to allow for noninvasive confirmation of diagnosis. An outline of patient diagnosis adjudication is shown in Supplemental Figure 1 (supplemental materials are available at http://jnm.snmjournals.org). All ATTR CA adjudication was performed with masking of the quantitative image results.

Imaging Protocols

Imaging with 99mTc-pyrophosphate was performed according to 1 of 2 protocols. All subjects underwent imaging on either an NM Discovery 870 (GE Healthcare, n = 233) or an NM Discovery 670 (GE Healthcare, n = 63) scanner. Patients received 740 MBq (20 mCi) of 99mTc-pyrophosphate, with planar images obtained at 3 h after injection over a 5- to 8-min duration, with a heart-centered field of view. SPECT images were acquired at 3 h with a 10% energy window and a 128 × 128 matrix. Patients imaged with the NM Discovery 870 had a total acquisition time of 12 min using a wide-energy, high-resolution collimator. Patients imaged with the NM Discovery 670 had a total acquisition time of 15 min using a low-energy high-resolution collimator. For both scanner systems, CT attenuation correction imaging was performed during an end-expiratory breath hold with no gating, in helical mode, with a slice thickness of 5 mm, tube voltage of 120 kVp and 20 mA, and a 512 × 512 matrix.

Expert Image Interpretation

Planar myocardial radiotracer retention was quantified from the 3-h images using heart-to-contralateral ratio at the time of clinical interpretation (4). Planar images at 3 h were also graded visually using Perugini grading (2). SPECT images at 3 h were visually assessed to determine whether diffuse radiotracer activity was present or absent in the myocardium (4).

CT Attenuation Correction Segmentation

A previously validated foundational deep learning model (Total Segmentator) was used to segment all heart chambers and myocardium (11). The model uses a no-new-Net (nnU-Net) architecture to automatically segment a variety of anatomic structures from CT images (11). A key advantage of this approach is the automatic optimization of the various hyperparameters, including an ensemble of 2- and 3-dimensional U-Net networks at multiple resolutions. For cardiac structures, the model was trained using expert annotations from contrast images, which were transferred to registered noncontrast images. We recently demonstrated excellent correlations between estimates from noncontrast CT and volumes manually segmented from gated, contrast-enhanced CT (left ventricular [LV] myocardium Spearman r = 0.947, LA volume Spearman r = 0.926) (12). The left atrium (LA) and LV myocardium, right atrium, and aorta regions derived automatically from the CT attenuation correction scans were then used to quantify the tracer on coregistered attenuation-corrected 99mTc-pyrophosphate SPECT images.

CT-Guided Quantification

All 99mTc-pyrophosphate quantification was performed using attenuation-corrected images. We used CT-derived LV myocardial boundaries to determine maximum myocardial counts. Maximum myocardial counts, corrected for radiotracer delay and camera sensitivity (13), were used to calculate the SUVmax from injected dose and weight (14). We also evaluated target-to-background ratio (TBR), as well as cardiac pyrophosphate activity (CPA) and volume of involvement (VOI), on the basis of our previous work (6). TBR is calculated as the maximum activity in the target (myocardium) divided by the mean activity of the background. VOI is quantified as the total volume of activity above threshold (maximum activity of the background) (6). CPA is quantified as the total activity (integrating volume and intensity) above threshold, divided by mean background activity (6).

For TBR, VOI, and CPA, LV myocardium was used as the target region, but we evaluated 5 methods for determining background activity. The first 3 methods were based on LA activity. First, we used the full cardiac segments to quantify LA 99mTc-pyrophosphate activity and LV myocardial 99mTc-pyrophosphate activity. These segmentations were used to quantify TBRfull, VOIfull, and CPAfull. Next, we evaluated a reduced-size left atrial region (reduced by 50%) to measure LA background activity. The reduced LA region was used to quantify TBRsmall, VOIsmall, and CPAsmall. Lastly, we considered 5 regions of interests (200 mm3 each) randomly sampled within the LA volume. This method was used to define TBRroi, VOIroi, and CPAroi measures. Similarly, we used maximal uptake in the right atrium to derive TBRra, VOIra, and CPAra and maximal uptake in the aorta derive TBRaorta, VOIaorta, and CPAaorta. Examples of the segmentation methods and overall deep learning workflow are shown in Figure 1.

FIGURE 1.
FIGURE 1.

Overview of segmentation and image quantification. Deep learning was used to segment noncontrast CT scans. LV myocardium segmentation was used without modification. We evaluated 5 methods for determining background activity. LA was sampled with full region, reduced region (shrunk 50% but centered in LA), and random regions of interest. Right atrium and aorta were sampled with full-size volumes. These segmentations were used to calculate TBR, VOI, and CPA. PYP = 99mTc-pyrophosphate; ROI = region of interest.

Outcomes

The primary outcome was diagnostic accuracy for ATTR CA. In the secondary analysis, we evaluated associations with the combined clinical outcome of cardiovascular death or heart failure hospitalization. Admissions for heart failure were ascertained from electronic medical records based on a primary admitting diagnosis of heart failure using the standardized codes from the International Classification of Diseases, 10th revision (15). Cardiovascular mortality was ascertained from electronic medical records and included death related to heart failure, myocardial infarction, or sudden death (16). Follow-up was censored at the last time that follow-up status could be verified. Outcomes were adjudicated with masking of the 99mTc-pyrophosphate results.

Statistical Analysis

Categorical variables were summarized as numbers and percentages and compared with a χ2 or Fisher exact test as appropriate. Continuous variables were summarized as mean and SD and compared using Student t test if normally distributed and summarized as median and interquartile range and compared using Wilcoxon rank sum test if not.

Prediction performance was evaluated using areas under the receiver operating characteristic curve (AUCs) and compared using the method of DeLong et al. (17). SPECT visual interpretation was not included in this analysis since it was used in part to adjudicate diagnosis. Unadjusted and age-adjusted associations with the combined outcome of cardiovascular death or heart failure hospitalization were assessed using Cox proportional hazards models. Multivariable models were limited to age-adjusted associations, given the low number of events (18). The proportional hazards assumption was assessed using Schoenfeld residuals, with no significant violations identified.

All statistical tests were 2-sided, with a P value of less than 0.05 considered statistically significant. All analyses were performed with STATA version 13 (StataCorp).

RESULTS

Patient Population

In total, 299 patients were included in the study, with a diagnosis of ATTR CA in 83 patients. The population characteristics are outlined in Table 1. Patients with ATTR CA were older (median age, 82 vs. 74 y; P < 0.001) and more likely to be male (81.9% vs. 56.5%; P < 0.001).

View this table:
TABLE 1.

Population Characteristics Stratified by Presence of ATTR CA

Cardiac chamber measurements took an average of 15.8 ± 0.2 s (including time for chamber segmentation and image quantification). Cases demonstrating chamber segmentation and image quantification are shown in Figure 2 for a patient with ATTR CA and in Figure 3 for a patient without ATTR CA. Imaging parameters are outlined in Table 2. The median SUVmax, TBR, VOI, and CPA, regardless of the quantification method, were significantly higher in patients with ATTR CA (all P < 0.001). Distributions of quantitative imaging variables are shown in Supplemental Figure 2.

FIGURE 2.
FIGURE 2.

Example patient with ATTR CA. (A and B) Attenuation correction imaging (A) segmented using deep learning (B). (C) Three-dimensional segmentations for all chambers. (D) Hybrid SPECT/CT imaging, with scale in SUV. (E) Chamber segmentations applied to coregistered SPECT to quantify radiotracer activity.

FIGURE 3.
FIGURE 3.

Example patient without ATTR CA. (A and B) Attenuation correction imaging (A) segmented using deep learning (B). (C) Three-dimensional segmentations for all chambers. (D) Hybrid SPECT/CT imaging, with scale in SUV. (E) Chamber segmentations applied to coregistered SPECT to quantify radiotracer activity.

View this table:
TABLE 2.

Imaging Characteristics Stratified by Presence of ATTR CA

Diagnostic Accuracy

Prediction performance for ATTR CA using clinical and quantitative measures, and using new automated measures with reduced LA volume as the background, is shown in Figure 4. CPAsmall (AUC, 0.989; 95% CI, 0.974–1.00) and VOIsmall (AUC, 0.988; 95% CI, 0.973–1.00) had the highest prediction performance for ATTR CA. The AUC for CPAsmall was significantly higher than that for the heart-to-contralateral ratio (AUC, 0.975; 95% CI, 0.952–0.998; P = 0.046) and SUVmax (AUC, 0.944; 95% CI, 0.919–0.970; P = 0.013).

FIGURE 4.
FIGURE 4.

Prediction performance for diagnosis of ATTR CA using clinical and quantitative methods (n = 83/299). SUVmax was quantified from LV myocardium region. Background for TBR, CPA, and VOI was reduced-size region of interest in LA. HCL = heart to contralateral.

Prediction performance for ATTR CA using all automated methods for quantifying CPA is shown in Supplemental Figure 3. Of the quantification methods, CPAsmall (AUC, 0.989; 95% CI, 0.974–1.00) and CPAra (AUC, 0.987; 95% CI, 0.970–1.00) had the highest prediction performance. Prediction performance for methods of quantifying TBR is shown in Supplemental Figure 4. Prediction performance using VOI, shown in Supplemental Figure 5, was similar to the results for CPA.

Clinical Outcomes

Among the 83 patients with ATTR CA, during a median follow-up of 1.6 y (interquartile range, 1.0–1.9 y), 15 patients experienced heart failure–related hospitalization and 8 patients experienced cardiovascular death as the first event. Unadjusted and age-adjusted associations with the combined outcome of cardiovascular death or heart failure hospitalization are shown in Figure 5. Most quantitative variables were associated with an increased risk of cardiovascular death or heart failure hospitalization in unadjusted analyses. After adjustment for age, all quantitative variables were associated with an increased risk of events, with hazard ratios ranging from 1.41 to 1.84 per SD increase.

FIGURE 5.
FIGURE 5.

Associations (per SD increase) with incidence of cardiovascular (CV) death or heart failure (HF) hospitalization in patients with ATTR CA.

DISCUSSION

We demonstrated the feasibility of fully automated quantification of 99mTc-pyrophosphate uptake using deep learning segmentation of coregistered CT attenuation maps. We leveraged CT anatomic information to identify structures that cannot be consistently identified using 99mTc-pyrophosphate imaging alone. The diagnostic accuracy for 99mTc-pyrophosphate quantification was consistently excellent, without the need for manual intervention. Quantification was independent of variability in 99mTc-pyrophosphate uptake, allowing quantification in cases with low uptake, where methods based on SPECT imaging alone may be subjective and unreliable. Lastly, we demonstrated the potential clinical importance of automated measures by evaluating associations with the combined outcome of cardiovascular mortality or hospitalization due to heart failure. Overall, our results suggest that fully automated quantitation of 99mTc-pyrophosphate is possible and could potentially be used to support physician interpretation or risk stratification.

There has been a significant increase in the use of 99mTc-pyrophosphate imaging in response to the high diagnostic accuracy of the test and increasing awareness regarding ATTR CA (19). As a result, 99mTc-pyrophosphate imaging is increasingly performed in community centers, where physicians may have less expertise in the technique. Although SPECT imaging is highly accurate when analyzed by experienced observers, experience is required to differentiate blood pool from myocardial radiotracer activity, particularly in cases with lower uptake. Dual-isotope imaging was proposed for quantifying myocardial uptake (20). However, this is currently impractical because of limited access to thallium, increased radiation exposure, and lack of a consistent definition of background activity. Hybrid SPECT/CT allows physicians to identify anatomic landmarks (21), so it is intuitive to use a similar approach for artificial intelligence approaches to image evaluation. We demonstrate the feasibility of this approach, using a deep learning model to segment CT attenuation imaging to facilitate automated quantification of SPECT tracer. A similar approach has been applied to whole-heart segmentation from CT attenuation correction imaging (22). Our method has the potential advantage of separating out regions of blood-pool activity from myocardial uptake and also providing the measures of overall disease burden such as CPA or VOI. Furthermore, the segmentation method for CT images was previously externally validated (12). Importantly, since the method is based on CT imaging, it does not rely on (or assume) any specific distribution of radiotracer. The model segments CT images rapidly (<16 s) and could be deployed as a preprocessing step, since the pipeline is fully automated. Therefore, it could be deployed in a clinical workflow without interruption to support physician interpretation, which may be particularly beneficial in centers without extensive 99mTc-pyrophosphate imaging expertise.

Several different methods for quantifying radiotracer 99mTc-pyrophosphate activity have been proposed to date. Scully et al. evaluated SUVpeak in myocardium as well as a retention index (5). Kessler et al. evaluated a method for whole-heart quantitation of radiotracer activity (23), which demonstrated high diagnostic accuracy. For quantifying cardiac amyloid activity, Dorbala et al. proposed a method that relied on manually defining myocardium on CT attenuation imaging and then transferring those contours to associated 99mTc-pyrophosphate imaging (24). Importantly, these measurements correlated strongly with cardiovascular MR markers of disease burden (24). In our previous work, we demonstrated that CPA and VOI (manually derived) have high diagnostic accuracy (6) and correlate with native myocardial T1 (7). However, all of those studies relied on tedious manual segmentation, which is a barrier to clinical translation.

We evaluated a large number of quantitation schemes facilitated by automated segmentation of CT structures. The diagnostic accuracy of most quantitative variables was excellent, with only a few significant differences identified. SUVmax had slightly lower diagnostic accuracy, which may be a reflection of inherent limitations in quantifying SUV from SPECT imaging (25) or of the fact that SUVmax is defined by a single pixel value. There was a trend toward higher prediction performance for methods using small LA volumes or RA volumes as background. All of the methods had diagnostic accuracy similar or superior to that of planar interpretation, which was informed by SPECT visual interpretation (since we relied on clinically reported planar interpretation). All of the methods for quantifying 99mTc-pyrophosphate activity from SPECT/CT were associated with the combined clinical outcome after adjustment for age. Further large-scale comparisons may be needed to differentiate more precisely between different methods of determining the background counts.

Identification of diffuse myocardial uptake on SPECT images is part of the diagnostic criteria for ATTR CA (4) but requires expertise to accurately differentiate blood pool from myocardium. Therefore, quantitative analysis of SPECT/CT may be particularly valuable in smaller centers, with less expertise, as a diagnostic tool. Furthermore, risk prediction remains an important clinical need, particularly with the emergence of additional therapies for ATTR CA (26). Objective and automated SPECT/CT quantification may identify patients at highest risk and also provide precise assessment of changes in radiotracer uptake in response to therapy (27). Artificial intelligence–based volumetric quantification of 99mTc-pyrophosphate activity, and potentially other bone-seeking radiotracers, may be a particularly precise way to follow changes in response to disease-modifying therapies (28). This could potentially be used to identify patients who are not responding to therapy and could be transitioned to other therapies.

Our study had a few important limitations. We quantified all SPECT/CT images at 3 h using 99mTc-pyrophosphate, and the results may not extrapolate to centers performing imaging only at 1 h (3) or using other radiotracers (29). Endomyocardial biopsy was not performed on most patients. Although a pathologic diagnosis would provide greater certainty regarding adjudication of ATTR CA, this is not reflective of clinical practice, which has shifted to increasingly noninvasive diagnosis (3). Additionally, we had a limited number of patients with serial studies and therefore were not able to assess changes over time or in response to therapy (28). However, this would be an important next step in evaluating this approach. Lastly, we relied on spatial registration between CT and 99mTc-pyrophosphate images. Expert oversight is needed to verify if the registration is correct, which may be difficult in cases with limited uptake. Nevertheless, correct SPECT/CT alignment is required for attenuation correction of the 99mTc-pyrophosphate counts, even if CT is not used for the region definition. Thus, any quantitative approach for attenuation-corrected 99mTc-pyrophosphate SPECT scans must ensure correct SPECT/CT registration.

CONCLUSION

The anatomic information available from CT attenuation correction imaging can be leveraged by deep learning to automatically segment and quantify 99mTc-pyrophosphate imaging regardless of the intensity or distribution of radiotracer uptake. This approach can be used to support physician interpretation of 99mTc-pyrophosphate images, which may be particularly beneficial in community centers, or to quantify radiotracer activity, which may potentially play a role in assessment of therapies or risk stratification.

DISCLOSURE

This research was supported in part by grant R35HL161195 from the National Heart, Lung, and Blood Institute/National Institutes of Health (NHLBI/NIH) (principal investigator, Piotr Slomka). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Robert Miller reports research and consulting support from Pfizer. Nowell Fine reports research and consulting support from Pfizer, Akcea, Alnylam, and Eidos. Paul Kavanagh, Daniel Berman, and Piotr Slomka participate in software royalties for QPS software at Cedars-Sinai Medical Center. Piotr Slomka has received research grant support from Siemens Medical Systems and consulting fees from Synektik. Daniel Berman has served as a consultant for GE Healthcare. No other potential conflict of interest relevant to this article was reported.

KEY POINTS

QUESTION: Can deep learning use anatomic information from CT attenuation maps to automatically quantify 99mTc-pyrophosphate imaging?

PERTINENT FINDINGS: Using 299 patients, we demonstrated the feasibility of a deep learning–based approach for quantifying 99mTc-pyrophosphate imaging, with a mean processing time of less than 16 s. Measures of 99mTc-pyrophosphate radiotracer uptake have high diagnostic accuracy for ATTR CA and were associated with the composite outcome of cardiovascular death or heart failure–related hospitalization.

IMPLICATIONS FOR PATIENT CARE: CT-guided quantification of 99mTc-pyrophosphate imaging can be used to automate analysis, support physician interpretation, and provide quantitative risk stratification.

Footnotes

  • ↵* Contributed equally to this work.

  • Published online May 9, 2024.

REFERENCES

  1. 1.↵
    1. Ruberg FL,
    2. Grogan M,
    3. Hanna M,
    4. Kelly JW,
    5. Maurer MS
    . Transthyretin amyloid cardiomyopathy. J Am Coll Cardiol. 2019;73:2872–2891.
    OpenUrlCrossRefPubMed
  2. 2.↵
    1. Perugini E,
    2. Guidalotti PL,
    3. Salvi F,
    4. et al
    . Noninvasive etiologic diagnosis of cardiac amyloidosis using 99mTc-3,3-diphosphono-1,2-propanodicarboxylic acid scintigraphy. J Am Coll Cardiol. 2005;46:1076–1084.
    OpenUrlFREE Full Text
  3. 3.↵
    1. Masri A,
    2. Bukhari S,
    3. Ahmad S,
    4. et al
    . Efficient 1-hour technetium-99m pyrophosphate imaging protocol for the diagnosis of transthyretin cardiac amyloidosis. Circ Cardiovasc Imaging. 2020;13:e010249.
    OpenUrl
  4. 4.↵
    1. Dorbala S,
    2. Ando Y,
    3. Bokhari S,
    4. et al
    . ASNC/AHA/ASE/EANM/HFSA/ISA/SCMR/SNMMI expert consensus recommendations for multimodality imaging in cardiac amyloidosis. J Nucl Cardiol. 2020;27:659–673.
    OpenUrlCrossRef
  5. 5.↵
    1. Scully PR,
    2. Morris E,
    3. Patel KP,
    4. et al
    . DPD quantification in cardiac amyloidosis. JACC Cardiovasc Imaging. 2020;13:1353–1363.
    OpenUrlAbstract/FREE Full Text
  6. 6.↵
    1. Miller RJH,
    2. Cadet S,
    3. Mah D,
    4. et al
    . Diagnostic and prognostic value of technetium-99m pyrophosphate uptake quantitation for transthyretin cardiac amyloidosis. J Nucl Cardiol. 2021;28:1835–1845.
    OpenUrl
  7. 7.↵
    1. Roshankar G,
    2. White GC,
    3. Cadet S,
    4. et al
    . Quantitative technetium pyrophosphate and cardiovascular magnetic resonance in patients with suspected cardiac amyloidosis. J Nucl Cardiol. 2022;29:2679–2690.
    OpenUrl
  8. 8.↵
    1. Han D,
    2. Shanbhag A,
    3. Miller RJH,
    4. et al
    . Artificial intelligence-based automated left ventricular mass quantification from non-contrast cardiac CT scans: correlation with contrast CT and cardiac MRI. medRxiv website. https://www.medrxiv.org/content/10.1101/2024.01.12.24301169v1. Published January 13, 2024. Accessed April 18, 2024.
  9. 9.↵
    1. Garcia-Pavia P,
    2. Rapezzi C,
    3. Adler Y,
    4. et al
    . Diagnosis and treatment of cardiac amyloidosis. A position statement of the European Society of Cardiology Working Group on Myocardial and Pericardial Diseases. Eur J Heart Fail. 2021;23:512–526.
    OpenUrl
  10. 10.↵
    1. Khayambashi S,
    2. Elzinga K,
    3. Hahn C,
    4. et al
    . Amyloidosis tissue confirmation for tafamidis eligibility using transverse carpal ligament and tenosynovium biopsy. Can J Cardiol. 2022;38:1643–1646.
    OpenUrl
  11. 11.↵
    1. Wasserthal J,
    2. Breit H-C,
    3. Meyer MT,
    4. et al
    . TotalSegmentator: robust segmentation of 104 anatomic structures in CT images. Radiol Artif Intell. 2023;5:e230024.
    OpenUrlCrossRef
  12. 12.↵
    1. Miller RJ,
    2. Shanbhag A,
    3. Killekar A,
    4. et al
    . AI-defined cardiac anatomy improves risk stratification of hybrid perfusion imaging. JACC Cardiovasc Imaging. March 6, 2024 [Epub ahead of print].
  13. 13.↵
    1. Ito T,
    2. Matsusaka Y,
    3. Onoguchi M,
    4. et al
    . Experimental evaluation of the GE NM/CT 870 CZT clinical SPECT system. J Appl Clin Med Phys. 2021;22:165–177.
    OpenUrl
  14. 14.↵
    1. Kinahan PE,
    2. Fletcher JW
    . Positron emission tomography-computed tomography standardized uptake values in clinical practice and assessing response to therapy. Semin Ultrasound CT MR. 2010;31:496–505.
    OpenUrlCrossRefPubMed
  15. 15.↵
    1. Southern DA,
    2. Norris CM,
    3. Quan H,
    4. et al
    . An administrative data merging solution for dealing with missing data in a clinical registry. BMC Med Res Methodol. 2008;8:1.
    OpenUrlCrossRefPubMed
  16. 16.↵
    1. Hicks KA,
    2. Tcheng JE,
    3. Bozkurt B,
    4. et al
    . 2014 ACC/AHA key data elements and definitions for cardiovascular endpoint events in clinical trials. J Am Coll Cardiol. 2015;66:403–469.
    OpenUrlFREE Full Text
  17. 17.↵
    1. DeLong ER,
    2. DeLong DM,
    3. Clarke-Pearson DL
    . Comparing the areas under two or more correlated receiver operating characteristic curves. Biometrics. 1988;44:837–845.
    OpenUrlCrossRefPubMed
  18. 18.↵
    1. Austin PC,
    2. Steyerberg EW
    . Events per variable (EPV) and the relative performance of different strategies for estimating the out-of-sample validity of logistic regression models. Stat Methods Med Res. 2017;26:796–808.
    OpenUrlCrossRefPubMed
  19. 19.↵
    1. Poterucha TJ,
    2. Elias P,
    3. Bokhari S,
    4. et al
    . Diagnosing transthyretin cardiac amyloidosis by technetium Tc 99m pyrophosphate: a test in evolution. JACC Cardiovasc Imaging. 2021;14:1221–1231.
    OpenUrlCrossRef
  20. 20.↵
    1. Tamarappoo B,
    2. Otaki Y,
    3. Manabe O,
    4. et al
    . Simultaneous Tc-99m PYP/Tl-201 dual-isotope SPECT myocardial imaging in patients with suspected cardiac amyloidosis. J Nucl Cardiol. 2020;27:28–37.
    OpenUrl
  21. 21.↵
    1. Sperry BW,
    2. Bateman TM,
    3. Akin EA,
    4. et al
    . Hot spot imaging in cardiovascular diseases. J Nucl Med. 2022;63:1722–1740.
    OpenUrlAbstract/FREE Full Text
  22. 22.↵
    1. Bhattaru A,
    2. Rojulpote C,
    3. Vidula M,
    4. et al
    . Deep learning approach for automated segmentation of myocardium using bone scintigraphy SPECT/CT in patients with suspected cardiac amyloidosis. J Nucl Cardiol. 2024;33:101809.
    OpenUrl
  23. 23.↵
    1. Kessler L,
    2. Fragoso Costa P,
    3. Kersting D,
    4. et al
    . Quantitative 99mTc-DPD-SPECT/CT assessment of cardiac amyloidosis. J Nucl Cardiol. 2023;30:101–111.
    OpenUrl
  24. 24.↵
    1. Dorbala S,
    2. Park MA,
    3. Cuddy S,
    4. et al
    . Absolute quantitation of cardiac (99m)Tc-pyrophosphate using cadmium-zinc-telluride-based SPECT/CT. J Nucl Med. 2021;62:716–722.
    OpenUrlAbstract/FREE Full Text
  25. 25.↵
    1. Dickson JC,
    2. Armstrong IS,
    3. Gabiña PM,
    4. et al
    . EANM practice guideline for quantitative SPECT-CT. Eur J Nucl Med Mol Imaging. 2023;50:980–995.
    OpenUrl
  26. 26.↵
    1. Maurer MS,
    2. Kale P,
    3. Fontana M,
    4. et al
    . Patisiran treatment in patients with transthyretin cardiac amyloidosis. N Engl J Med. 2023;389:1553–1565.
    OpenUrl
  27. 27.↵
    1. Fontana M,
    2. Martinez-Naharro A,
    3. Chacko L,
    4. et al
    . Reduction in CMR derived extracellular volume with patisiran indicates cardiac amyloid regression. JACC Cardiovasc Imaging. 2021;14:189–199.
    OpenUrl
  28. 28.↵
    1. Papathanasiou M,
    2. Kessler L,
    3. Bengel FM,
    4. et al
    . Regression of myocardial 99mTc-DPD uptake after tafamidis treatment of cardiac transthyretin amyloidosis. J Nucl Med. 2023;64:1083–1086.
    OpenUrlAbstract/FREE Full Text
  29. 29.↵
    1. Hutt DF,
    2. Quigley AM,
    3. Page J,
    4. et al
    . Utility and limitations of 3,3-diphosphono-1,2-propanodicarboxylic acid scintigraphy in systemic amyloidosis. Eur Heart J Cardiovasc Imaging. 2014;15:1289–1298.
    OpenUrlCrossRefPubMed
  • Received for publication January 29, 2024.
  • Accepted for publication April 3, 2024.
Back to top

In this issue

Print
Download PDF
Article Alerts
Email Article
Citation Tools
Bookmark this article

Jump to section

Keywords