339 - 3D Topological Analysis for Assessing Therapeutic Responses to Radiation and Chemotherapy with and without Anti-CD40 Immunotherapy in Local Advanced Rectal Cancer
Presenter(s)
H. Zhang1, J. Liu1, N. N. Sanford2, G. Khatri3, R. D. Timmerman4, T. A. Aguilera5, and H. Peng6; 1UT Southwestern Medical Center, Dallas, TX, 2Harvard Radiation Oncology Program, Boston, MA, 3UT Southwestern, Dallas, TX, 4Department of Radiation Oncology, University of Texas Southwestern Medical Center, Dallas, TX, 5Department of Radiation Oncology, UT Southwestern Medical Center, Dallas, TX, 6University of Texas Southwestern Medical Center, Dallas, TX
Purpose/Objective(s): This study aims to integrate topological analysis with standard radiomics from T2-weighted MRI to evaluate therapeutic responses and quantify treatment-induced changes following neoadjuvant short course radiotherapy followed by 3 months of chemotherapy (SCRT/Chemo) ± anti-CD40 immunotherapy.
Materials/Methods: Pre- and post-treatment MRIs from 21 patients in INNATE trial (NCT04130854) (Arm1: SCRT/Chemo + anti-CD40, n=12; Arm2: SCRT/Chemo alone, n=9) were analyzed. Grayscale intensities in gross tumor volumes (GTVs), independently verified by a board-certified radiologist, were normalized to [0,1]. Sublevel set filtrations were adopted to construct persistent homology representations. Persistence diagrams and multi-threshold Betti numbers were calculated to quantify topological features, including connected components (H0), loops (H1), and voids (H2). Gaussian modeling of Betti2 dynamics employed fitted parameters (a, µ, s, R²) as higher-order descriptors. Standard 3D radiomics features were extracted GTV, while 2D focusing on pre-/post-treatment axial slices registered to pelvic landmarks for consistent texture analysis. Delta features (difference between pre- and post- treatment) were calculated to assess treatment-induced changes. Statistical analyses (Welch’s t-test and Spearman correlation) were conducted between two arms and response groups (complete response [CR], partial response [PR]).
Results: Arm1 (with anti-CD40) exhibits more pronounced GTV reductions (?Volume: -47.64±30.24 cm³ vs. Arm2: -18.56±22.44 cm³, p<0.01). Elongation analysis reveals that higher ?Elongation in PR vs. CR patients (P < 0.05). In 2D radiomics, only five significant texture features are identified, showing high inter-feature correlation (>0.9). ?GLDM_DependenceNonUniformity is significantly lower in Arm1, reflecting enhanced tumor homogeneity post-treatment. 3D topological analysis identifies 16 significant features, with ?Betti2 (t=0.2) and ?Persistence Entropy (dim2) demonstrating the strongest subgroup differentiation. In PR_Arm2 (PR patients without anti-CD40), minimal delta changes in persistence entropy (dim2) (?=-0.52 ± 0.44 vs. ?=-2.31 in others, p<0.0005) indicate limited structural remodeling, consistent with poor outcome. CR_Arm1(CR patients with anti-CD40) demonstrate significant post-treatment reductions in ?Betti2 (t=0.2) (?=-143.0 ± 138.72 vs. ?=0.53 ± 10.13 in others, p<0.0002), independent of GTV changes (Spearman ?=0.35). Additionally Gaussian modeling based on pre-treatment data can distinguish responders (75% CR with R² < 0.95) from non-responders (77% PR with R² = 0.95).
Conclusion: In rectal cancer, 3D topological features can be used to complement conventional radiomics, providing deeper insight into tumor heterogeneity, invasiveness, and response to treatment.