ESTRO 2024 - Abstract Book

S1747

Clinical - Lung

ESTRO 2024

2597

Digital Poster

RADIOPATHOMICS MODEL TO PREDICT EARLY RESPONSE IN LA-NSCLC PATIENTS UNDERGOING CHEMORADIATION

Carlo Greco 1,2 , Edy Ippolito 1,2 , Fiore Michele 2,1 , Aurelia Iurato 2 , Marianna Miele 2 , Luca Eolo Trodella 2 , Tacconi Claudia 1 , Palumbo Vincenzo 1 , Rosa Sicilia 3 , Matteo Tortora 3 , Paolo Soda 4 , Sara Ramella 1,2 1 Università Campus Bio-Medico di Roma, Via Alvaro del Portillo, 21 - 00128, Radiation Oncology, Rome, Italy. 2 Fondazione Policlinico Universitario Campus Bio-Medico, Via Alvaro del Portillo, 200 - 00128, Radiation Oncology, Rome, Italy. 3 Università Campus Bio-Medico di Roma, Via Alvaro del Portillo, 21 - 00128, Unit of Computer Systems and Bioinformatics, Department of Engineering, Rome, Italy. 4 Università Campus Bio-Medico di Roma, Via Alvaro del Portillo, 21 - 00128, 1Unit of Computer Systems and Bioinformatics, Department of Engineering, Rome, Italy

Purpose/Objective:

We aimed to investigate the power of pre-treatment CT-based radiopathomic features to predict early response during chemoradiation. The idea was to convert the multimodal dataset composed of medical images, radiomics and pathomics data, and the other useful information into a signature that, feeding AI-based decision support systems, will offer the possibility to personalize therapy

Material/Methods:

Patients with stage unresectable stage III NSCLC enrolled in a prospective protocol and treated with concurrent definitive chemoradiation with an adaptive approach at our institution from 2012 to 2014. Different data including clinical data, CT scans, histological slides were evaluated and different features extracted: 1) Semantic features divided into personal data (age, sex and smoking attitude), staging scores of the tumour (T, N and tumor stage), and histology and gene mutations evaluation; 2) Radiomics features, derived from CT scans collected prior to the start of concomitant chemoradiation therapy treatment; the radiomics features were extracted from 3D ROIs given by the Clinical Target Volume (CTV) that was preferred over GTV as in a previous work it showed to be more accurate; 3) Pathomics features: biopsy hystological slides of lung cancer tissue stained with haematoxylin and eosin (HE) were digitised obtaining whole slide images (WSI) using NanoZoomer 2.0 RT (Hamamatsu) at a magnification power of 20x at a resolution of 0.5 μm per pixel. WSIs were loaded on QuPath software and re -evaluated by an expert lung pathologist that selected representative number of tumor areas, called Regions of Interest (ROIs). Tumor areas within the ROIs, defined as “crops”, were manually segmented, where segmentation consists in contouring the tumor edge avoiding fibrosis, necrosis or histological artifacts. Within this context different multimodal late fusion rules and two patient-wise aggregation rules leveraging the richness of information given by CT images, whole-slide scans and clinical data were explored

Results:

Overall 50 patients with locally advanced NSCLC were enrolled in the adaptive protocol. Among these patients, 35 patients had available histological slides and were included in this analysis. In this latter group, 13 (37.1%) patients