SU105 - The Use of Image Biomarker Standardization Initiative Features, to Develop Radiomics Classifiers for Neoadjuvant Response in Locally Advanced Rectal Cancer

Sunday, June 2, 2024

11:55 AM – 12:01 PM ET, USA

Location: Room 314-317

Abstract Podium Presenter(s)

Ross K. McMahon, MRCS MBChB BMSc (he/him/his)

Clinical Research Fellow & Surgical Registrar
University of Glasgow
Glasgow Royal Infirmary, Academic Unit of Surgery
Glasgow, Scotland, United Kingdom

All relevant financial disclosures have been managed and will be listed below.
This individual has no financial relationships with ineligible companies.

Author(s)

LG

Laura Grocutt, n/a

Clinical Scientist
University of Glasgow
Glasgow, Scotland, United Kingdom

Disclosure information not submitted.
CS

Colin Steele, n/a

Consultant Surgeon
FRCS
Glasgow, Scotland, United Kingdom

Disclosure information not submitted.
MM

Martina Mori, n/a

Medical Physics-Researcher
IRCCS San Raffaele Scientific Institute
Milan, Lombardia, Italy

Disclosure information not submitted.
JP

Jonathan Platt, n/a

Consultant Radiologist
Glasgow Royal Infirmary
Glasgow, Scotland, United Kingdom

Disclosure information not submitted.
MD

Michael Digby, n/a

Consultant Radiologist
Glasgow Royal Infirmary
Glasgow, Scotland, United Kingdom

Disclosure information not submitted.
PH

Paul Horgan, n/a

Professor of Surgery
NHS Greater Glasgow and Clyde, University of Glasgow
Glasgow, Scotland, United Kingdom

Disclosure information not submitted.
ES

Emiliano Spezi, n/a

Full Professor of Healthcare Engineering and Director of Research
Cardiff University
Cardiff, Wales, United Kingdom

Disclosure information not submitted.
SO

Sean M. O'Cathail, DPhil FRCP FRCR

Senior Research Fellow
University of Glasgow
Glasgow, Scotland, United Kingdom

Disclosure information not submitted.
CR

Campbell S. Roxburgh, FRCS

Consultant
NHS Greater Glasgow and Clyde, University of Glasgow
Glasgow, Scotland, United Kingdom

Disclosure information not submitted.

Purpose/Background: A major challenge in locally advanced rectal cancer (LARC) is heterogeneity of response to neoadjuvant radiotherapy - with no reliable predictors. Those who achieve pathological complete response (pCR), or Clinical complete response (cCR) have excellent outcomes. In cCR, organ preservation is attainable and avoids the morbidity of surgery. This has brought into focus the need to better understand the processes that govern response.
Radiomics allows the extraction of quantitative features from standard imaging which could identify potential response biomarkers and enhance tumour phenotyping. Here, we explore a radiomic model of tumour response in a large cohort of LARC patients.

Methods/Interventions: We included consecutive LARC patients from a single cancer centre who received curative-intent NAT between; June 2016–July 2021. Region-of-interests around the gross tumour volume (pGTV) were contoured on true-axial slices of radiotherapy planning CTs. The study aimed to define radiomic features (RF) in association with the following dichotomous endpoints; ‘complete response’ (pCR and cCR) vs ‘incomplete response’ (all others), and ‘good regression’ (pCR or TRG0-1 or cCR) vs ‘poor regression’ (TRG2-3).
Radiomics was performed using SPAARC software that includes 164 Image Biomarker Standardisation Initiative (IBSI) shape and texture features. Based on validated methodology, the extracted RFs were combined into multivariate logistic regression (MVLR) on training cohort using minimum redundancy and a bootstrap-based machine learning approach to select the most significant RF and develop models. The resulting Prognostic Index (PI) was then tested on the validation cohort.

Results/Outcomes: 222 LARC patients were included divided between training (n=150) and validation (n=72) cohorts. 54 (24.3%) achieved ‘complete response’. The ‘response model’ PI combined Morphology_Diameter and cm_inv_diff_norm_3D_avg which measures Grey-level homogeneity. This model associated with ‘complete response’ (AUC=0.70, 95%CL 0.62-0.78, p= 0.0001) in training with results confirmed in validation (AUC =0.70, 95%CI 0.60-0.80, p=0.0242). On analysis of tumour regression, 139 (63%) achieved ‘good regression’. The ‘regression model’ included morph_vol_approx szm_szhge_3D. According to IBSI, the latter emphasizes zone counts in the lower left quadrant of the Gray-Level-Size-Zone-Matrix,. This model associated with ‘poor regression’ in training (AUC= 0.64, 95%CI 0.56-0.72, p= 0.0039) and validation (AUC=0.66, 95%CI 0.53-0.84, p=0.0398).

Conclusion/Discussion: In a large cohort of LARC patients, we report, in training and validation cohorts, a strong association between radiomics features and treatment response. This study is strengthened by the fact IBSI standardized features were used highlighting the potential value of non-invasive radiomic classifiers. Further expanded analysis on the interpretation of the textural features included in the models are warranted as is external validation. Further work is planned to incorporate biological data from circulating and tissue biomarkers, which we envisage will strengthen our understanding of radiobiology and support novel approaches to tumour phenotyping.