Click here to

Session: Imaging for Treatment Assessment and Outcome Modeling [Return to Session]

Patient-Specific Adjuvant Chemotherapy Decision Making Using Radiomics for Locally-Advanced NPC: A Multicenter Retrospective Study

X Teng1*, J Zhang1, Z Ma1, X Han1, S Lam1, H Xiao1, C Liu1, W Li1, Y Huang1, F Lee2, W Yip3, A Cheung1, H Lee4, J Cai1, (1) The Hong Kong Polytechnic University, Hong Kong (2) Queen Elizabeth Hospital, Hong Kong (3) Hong Kong Sanatorium and Hospital (4) The University Of Hong Kong


SU-F-206-7 (Sunday, 7/10/2022) 2:00 PM - 3:00 PM [Eastern Time (GMT-4)]

Room 206

Purpose: Concurrent chemo-radiation therapy (CCRT) is the mainstream treatment strategy for locally advanced nasopharyngeal (NPC) patients, but the benefit of adding adjuvant chemotherapy (ACT) is still unclear. This study proposed a radiomics-based ACT net benefit index to guide patient-specific clinical decision-making in ACT by developing and comparing treatment-specific prognostic models for three-year progression-free survival (PFS).

Methods: 703 locally advanced NPC patients in two local hospitals from 2012 to 2015 were screened for discovery and validation. In addition to clinicopathological information, the contrast enhanced computed tomography images and primary tumor volumes were collected for radiomic feature extraction. Low repeatable radiomic features were screened out before modeling. Separate Cox regression models were developed for CCRT and ACT using the single most prognostic radiomic feature combined with clinicopathological factors. Patients with higher ACT net benefit, which is proposed as the hazard difference of CCRT model to ACT model, are considered to benefit from ACT.

Results: In the discovery dataset, the CCRT model had a C-index of 0.71 (p-value < 0.0001) on patients received CCRT alone, and the ACT model achieved 0.81 (p-value = 0.0001) for those received ACT. Within the high benefit patients stratified by the median ACT net benefit index, ACT resulted in a statistically higher (p-value=0.0112) three-year PFS rate (91%) than CCRT treatment alone (70%). Such difference was also statistically valid (p-value < 0.05) in internal bootstrapping. In the external validation cohort, the benefit improved from 82% (ACT) vs. 87% (CCRT) (p-value=0.98) to 80% vs. 75% (p=0.2054) after stratification.

Conclusion: The proposed radiomics-based ACT net benefit index successfully identified ACT beneficial patients with a higher three-year PFS rate in the internal dataset, but less significant using the same index externally. Further assessment of the proposed index on more patients is warranted.

Funding Support, Disclosures, and Conflict of Interest: ITS/080/19, Project of Strategic Importance (P0035421), The Hong Kong Polytechnic University and Shenzhen-Hong Kong-Macau S&T Program (Category C) (SGDX20201103095002019), Shenzhen Basic Research Program (R2021A067)


CT, Quantitative Imaging, Image Guidance


IM- CT: Radiomics

Contact Email