Back

Artificial Intelligence-Derived Extracellular Volume Fraction for Diagnosis and Prognostication in Patients with Light-Chain Cardiac Amyloidosis

Hwang, I.-C.; Chun, E. J.; Kim, P. K.; Kim, M.; Park, J.; Choi, H.-M.; Yoon, Y. E.; Cho, G.-Y.; Choi, B. W.

2024-03-22 cardiovascular medicine
10.1101/2024.03.20.24304642 medRxiv
Show abstract

AimsT1 mapping on cardiac magnetic resonance (CMR) imaging is useful for diagnosis and prognostication in patients with light-chain cardiac amyloidosis (AL-CA). We conducted this study to evaluate the performance of T1 mapping parameters for detection of cardiac amyloidosis (CA) in patients with left ventricular hypertrophy (LVH) and their prognostic values in patients with AL-CA, using a semi-automated deep learning algorithm. Methods and ResultsA total of 300 patients who underwent CMR for differential diagnosis of LVH were analyzed. CA was confirmed in 50 patients (39 with AL-CA and 11 with transthyretin amyloidosis), hypertrophic cardiomyopathy in 198, hypertensive heart disease in 47, and Fabry disease in 5. A semi-automated deep learning algorithm (Myomics-Q) was used for the analysis of the CMR images. The optimal cutoff extracellular volume fraction (ECV) for the differentiation of CA from other etiologies was 33.6% (diagnostic accuracy 85.6%). he artificial intelligence (AI)-derived ECV showed a significant prognostic value for a composite of cardiovascular death and heart failure hospitalization in patients with AL-CA (revised Mayo stage III or IV) (adjusted hazard ratio 4.247 for ECV [≥]40%, 95% confidence interval 1.215-14.851, p-value=0.024). Incorporation of AI-derived ECV into the revised Mayo staging system resulted in better risk stratification (integrated discrimination index 27.9%, p=0.013; net reclassification index 13.8%, p=0.007). ConclusionsAI-assisted T1 mapping on CMR imaging allows for improved diagnosis of CA from other etiologies of LVH. Furthermore, AI-derived ECV has significant prognostic value in patients with AL-CA, suggesting its clinical usefulness. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=138 SRC="FIGDIR/small/24304642v1_ufig1.gif" ALT="Figure 1"> View larger version (46K): org.highwire.dtl.DTLVardef@1f6d35forg.highwire.dtl.DTLVardef@1af21c1org.highwire.dtl.DTLVardef@d1120corg.highwire.dtl.DTLVardef@1f8089d_HPS_FORMAT_FIGEXP M_FIG C_FIG

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.