SPC365

CT-Radiomic features of Epicardial Adipose Tissue in AF recurrence after catheter ablation

Session: Sessão de Posters 46 - TC Cardíaca
Speaker: Inês Amorim Cruz

Congress: CPC 2024
Topic: B. Imaging
Theme: 03. Imaging
Subtheme: 03.6 Cross-Modality and Multi-Modality Imaging Topics
Session Type: Cartazes
FP Number: ---
Authors: Inês Amorim Cruz; Sílvia O. Diaz; Fábio Sousa-Nunes; João Pedrosa; Inês Neves; Rafael Teixeira; Francisca Saraiva; João Almeida; João Primo; Francisco Sampaio; António S. Barros; Ricardo Fontes-Carvalho
<p style="text-align:justify"><span style="font-size:medium"><span style="font-family:&quot;Times New Roman&quot;,serif"><span style="color:#000000"><strong><span style="font-size:10.5pt">Background:</span></strong><span style="font-size:10.5pt"> Obesity is </span><span style="font-size:10.5pt">an </span><span style="font-size:10.5pt">important risk </span><span style="font-size:10.5pt">factor</span><span style="font-size:10.5pt"> for atrial fibrillation (AF).</span> <span style="font-size:10.5pt">Advances </span><span style="font-size:10.5pt">in</span><span style="font-size:10.5pt"> imaging techniques </span><span style="font-size:10.5pt">have enabled the exploration of </span><span style="font-size:10.5pt">regional body fat distribution.</span> <span style="font-size:10.5pt">Epicardial adipose tissue (EAT) is a metabolically active tissue unique </span><span style="font-size:10.5pt">for</span><span style="font-size:10.5pt"> its unobstructed proximity to the heart, which </span><span style="font-size:10.5pt">might</span><span style="font-size:10.5pt"> influence the risk of AF or its recurrence.</span> <span style="font-size:10.5pt">Radiomics is an emerging image analysis</span><span style="font-size:10.5pt"> technique that leverages noninvasive</span><span style="font-size:10.5pt"> tissue analysis and </span><span style="font-size:10.5pt">is </span><span style="font-size:10.5pt">potentially useful for </span><span style="font-size:10.5pt">AF </span><span style="font-size:10.5pt">risk stratification.</span> <span style="font-size:10.5pt">We aimed to explore </span><span style="font-size:10.5pt">the </span><span style="font-size:10.5pt">radiomic features of EAT, which could improve </span><span style="font-size:10.5pt">the </span><span style="font-size:10.5pt">risk stratification of AF recurrence after catheter ablation.</span></span></span></span></p>

<p style="text-align:justify"><span style="font-size:medium"><span style="font-family:&quot;Times New Roman&quot;,serif"><span style="color:#000000"><strong><span style="font-size:10.5pt">Methods:</span></strong><span style="font-size:10.5pt"> We included all consecutive patients </span><span style="font-size:10.5pt">who underwent</span><span style="font-size:10.5pt"> AF ablation (2017-2021) who performed a CT scan prior to the procedure.</span><span style="font-size:10.5pt">The EAT was segmented using a U-Net framework, which is a convolutional neural network (CNN) designed for image segmentation, with </span><span style="font-size:10.5pt">noncontrast acquisition automatically </span><span style="font-size:10.5pt">applied.</span> <span style="font-size:10.5pt">The process of extracting radiomic features</span><span style="font-size:10.5pt"> was </span><span style="font-size:10.5pt">carried out</span><span style="font-size:10.5pt"> using the Pyradiomics </span><span style="font-size:10.5pt">software, resulting </span><span style="font-size:10.5pt">in </span><span style="font-size:10.5pt">a </span><span style="font-size:10.5pt">total </span><span style="font-size:10.5pt">of </span><span style="font-size:10.5pt">851 features</span><span style="font-size:10.5pt">. The three primary categories into which they</span><span style="font-size:10.5pt"> can be </span><span style="font-size:10.5pt">classified are</span><span style="font-size:10.5pt">: (i) shape </span><span style="font-size:10.5pt">attributes, </span><span style="font-size:10.5pt">(ii) first</span><span style="font-size:10.5pt">-</span><span style="font-size:10.5pt">order (intensity) </span><span style="font-size:10.5pt">features, </span><span style="font-size:10.5pt">and (iii) texture </span><span style="font-size:10.5pt">characteristics. </span><span style="font-size:10.5pt">We </span><span style="font-size:10.5pt">conducted</span><span style="font-size:10.5pt"> an exploratory analysis</span><span style="font-size:10.5pt"> by employing</span><span style="font-size:10.5pt"> a univariate Cox regression model, </span><span style="font-size:10.5pt">adjusting for</span><span style="font-size:10.5pt"> age, </span><span style="font-size:10.5pt">sex</span><span style="font-size:10.5pt">, BMI</span><span style="font-size:10.5pt">,</span><span style="font-size:10.5pt"> and type of AF</span><span style="font-size:10.5pt">,</span><span style="font-size:10.5pt"> and using non-transformed radiomic features.</span></span></span></span></p>

<p style="text-align:justify"><span style="font-size:medium"><span style="font-family:&quot;Times New Roman&quot;,serif"><span style="color:#000000"><strong><span style="font-size:10.5pt">Results:</span></strong><span style="font-size:10.5pt"> A total of 533 patients </span><span style="font-size:10.5pt">were included in the study, of whom </span><span style="font-size:10.5pt">36% </span><span style="font-size:10.5pt">were </span><span style="font-size:10.5pt">women, </span><span style="font-size:10.5pt">with a </span><span style="font-size:10.5pt">median age </span><span style="font-size:10.5pt">of </span><span style="font-size:10.5pt">58 years </span><span style="font-size:10.5pt">(interquartile range of </span><span style="font-size:10.5pt">49</span><span style="font-size:10.5pt">-</span><span style="font-size:10.5pt">65</span><span style="font-size:10.5pt">),</span><span style="font-size:10.5pt"> and 20% had persistent AF</span><span style="font-size:10.5pt">.</span><span style="font-size:10.5pt"> During a median follow-up of 26 months [IQR 19-36], 130 patients (24%) developed AF recurrence. Univariate Cox regression showed that only measures of texture heterogeneity of EAT were higher in patients with </span><span style="font-size:10.5pt">a </span><span style="font-size:10.5pt">higher risk of recurrence</span><span style="font-size:10.5pt">,</span><span style="font-size:10.5pt"> including contrast-related features (GLSZM &ndash; Gray-Level Non-Uniformity, HR 1.23 [95% CI, 1.01-1.51] p = 0.043) and non-uniform </span><span style="font-size:10.5pt">gray</span><span style="font-size:10.5pt">-level matrices (Size-Zone, Run-Length</span><span style="font-size:10.5pt">,</span><span style="font-size:10.5pt"> and Dependence Non-Uniformity). Features related to </span><span style="font-size:10.5pt">the </span><span style="font-size:10.5pt">shape and EAT attenuation distribution (global signal intensity) were similar between </span><span style="font-size:10.5pt">the </span><span style="font-size:10.5pt">groups.</span></span></span></span></p>

<p style="text-align:justify"><span style="font-size:medium"><span style="font-family:&quot;Times New Roman&quot;,serif"><span style="color:#000000"><strong><span style="font-size:10.5pt">Conclusion:</span></strong><span style="font-size:10.5pt"> In this </span><span style="font-size:10.5pt">group</span><span style="font-size:10.5pt"> of patients with </span><span style="font-size:10.5pt">atrial fibrillation (</span><span style="font-size:10.5pt">AF</span><span style="font-size:10.5pt">), the</span><span style="font-size:10.5pt"> qualitative </span><span style="font-size:10.5pt">characteristics (</span><span style="font-size:10.5pt">heterogeneity</span><span style="font-size:10.5pt"> of tissue</span><span style="font-size:10.5pt">) of epicardial adipose tissue, </span><span style="font-size:10.5pt">as opposed to </span><span style="font-size:10.5pt">quantitative features </span><span style="font-size:10.5pt">like</span><span style="font-size:10.5pt"> volume, were </span><span style="font-size:10.5pt">found to be linked to a faster </span><span style="font-size:10.5pt">recurrence </span><span style="font-size:10.5pt">of AF following</span><span style="font-size:10.5pt"> catheter ablation.</span></span></span></span></p>