Objective <p>We quantified volume and shape variability in lung microwave ablation (LMWA) zones, comparing them with expected ablation zones, exploring the correlation with tissue contraction.</p> Materials and methods <p>After Institutional Review Board approval, we retrospectively included patients who underwent LMWA between January 2015 and January 2019. Exclusion criteria were ablations with multiple burns/probes, overlapping ablation zones, or indistinguishable background lung parenchyma. Ablation zones were oriented along an applicator-centric coordinate system. We used Pyradiomics to generate volumes and the Euler characteristic transform for three-dimensional shape space analysis. Wilcoxon paired signed-rank tests compared the expected <i>versus</i> the actual ablation zone. Tissue contraction was quantified using paired anatomical landmarks before and after computed tomography scans.</p> Results <p>We included 111 ablations in 72 patients (31 male, 41 female; median age 59). Median ablation power was 65 watts (range 20‒65), median ablation time 5 min (range 1‒10). Total energy correlated with volume and width (<i>p</i> = 0.007, <i>p</i> = 0.003, respectively). Ablation volume did not differ from vendor predictions (<i>p</i> = 0.452), whereas length (<i>p</i> &lt; 0.001) and maximum width (<i>p</i> &lt; 0.001) were greater than predicted. Ablation shapes were more elongated (<i>p</i> &lt; 0.001), less spherical (<i>p</i> &lt; 0.001), asymmetric (wider in back than in front, <i>p</i> = 0.007), and diverged from expected ellipsoids. There was no correlation between tissue contraction and volume, power, or time.</p> Conclusion <p>The provided vendor model offers a reasonable estimate of mean ablation volume, but wide variability in volume and shape necessitates a more bespoke method of ablation zone prediction. Tissue contraction did not correlate with ablation zone variability.</p> Relevance statement <p>High variability in actual ablation volume, shape, length, width, and position along the needle and divergence from ellipsoids necessitates the development of improved ablation zone prediction models.</p> Key Points <p><UnorderedList Mark="Bullet"> <ItemContent> <p>While median ablation volumes do not differ from vendor predictions, ablation length (<i>p</i> &lt; 0.001) and maximum width (<i>p</i> &lt; 0.001) are greater than vendor predictions, and volumes are variable.</p> </ItemContent> <ItemContent> <p>Ablation shapes diverge from ellipsoids, are more elongated (<i>p</i> &lt; 0.001), less spherical (<i>p</i> &lt; 0.001), and asymmetric (<i>p</i> = 0.007).</p> </ItemContent> <ItemContent> <p>Tissue contraction does not account for the volume variability of ablation zones.</p> </ItemContent> </UnorderedList></p> Graphical Abstract <p></p>