MR Imaging of retinoblastoma subtypes: Aiding diagnostication, metastatic risk stratification and molecular subtype identification

Summary MR Imaging in retinoblastoma care Retinoblastoma is an oncologic disease presenting in very young children with a good prognosis in high income countries. Retinoblastoma is increasingly treated adopting eye-sparing treatment methods including intra-arterial chemotherapy. The eye sparing treatment methods necessitate diagnostic methods guiding treatment other than approaches requiring tumor tissue such as histopathological assessment. Moreover, the key to adequate personalized treatment is identifying tumor-subtypes for which detailed tumor characterization unravelling tumor-heterogeneity is warranted. MR imaging can be a technique facilitating this. MR imaging is a non-invasive technique without radiation damage, which is an important feature especially for the pediatric retinoblastoma population. MR imaging is generally used in high income countries in retinoblastoma care for supporting diagnosis, examining disease extent and screening the central nervous system for associated lesions and metastases. MR imaging for retinoblastoma diagnostication (part 1) It is important to differentiate retinoblastoma from the common (benign) retinoblastoma mimickers: persistent fetal vasculature (PFV) and Coats’ disease. In chapter 2 MR imaging of retinoblastoma with radiologically diffuse growth was compared with persistent fetal vasculature and Coat’s disease. A larger eye size (compared to contralateral), a sharp-V-shaped retinal detachment and vitreous seeding support retinoblastoma diagnosis. A smaller eye size, Y-shaped retinal detachment or intraretinal macrocyst supported retinoblastoma mimickers, either PFV or Coats’ disease. For a specific retinoblastoma mimicker diagnosis, ciliary body and lens deformations, optic nerve atrophy and a stalk between optic disc and lens were useful for diagnosis of PVF. For Coats’ disease, enhancing subfoveal nodules were found. Incorporating these imaging features in clinical MR imaging assessments can improve the accuracy of a diagnosis, diminishing the amount of children treated with the wrong regimen. MR imaging for retinoblastoma metastatic risk stratification (part 2) Optic nerve enhancement due to inflammation may be falsely interpreted as optic nerve tumor invasion, which is associated with an increased metastatic risk. In chapter 3 a retinoblastoma subtype with extensive inflammation was investigated: retinoblastoma-associated orbital cellulitis. Newly-identified contrast enhancement patterns could differentiate between tumor invasion and inflammation of the optic nerve head. Adopting these enhancement patterns, specificity increased for detection of optic nerve tumor invasion within context of orbital cellulitis from 32% to 89%. In chapter 4 we investigated the width of tumor invasion into the optic nerve, instead of the traditionally considered depth of invasion. Using this method, sensitivity for detecting optic nerve tumor invasion increased from 59% to 84%. Non-invasive molecular subtype identification using MR imaging in retinoblastoma (part 3) Increasingly, oncologic diseases are being classified and treated according to their molecular make-up. In chapter 5 a rare and aggressive retinoblastoma subtype was evaluated: retinoblastoma driven by MYCN amplification (MYCNARB1+/+) instead of RB1 pathogenic variants. MR imaging features that showed high specificity for MYCNARB1+/+ identification included: peripheral location, peritumoral hemorrhage, subretinal hemorrhage with a fluid level, and tumor-retinal folding with vitreous enclosure. Pre-treatment recognition of MYCNARB1+/+ retinoblastoma could aid selecting patients for trials for targeted treatmen. Radiogenomics – development of imaging biomarkers for molecular subtypes - has been receiving increasing scientific attention. Chapter 6 evaluated the spectrum of associations between imaging features and genomics in all oncologic diseases. In chapter 7 the use of radiogenomics in retinoblastoma was explored. MR imaging features were scored by a panel of radiologists and compared with whole genome expression data. The gene expression profile of ‘photorceptorness’ was evaluated, which’ low expression is associated with tumor progression and distinctive ex-vivo chemotherapy susceptibility. Low-photorceptorness was associated with advanced stage imaging features. These findings were (partly) validated in chapter 8. In an independent cohort with patients from three retinoblastoma referral centers.