The Feasibility of Integrating Resting-State fMRI Networks into Radiotherapy Treatment Planning

Abstract

Background

Functional magnetic resonance imaging (fMRI) presents the ability to selectively protect functionally significant regions of the brain when primary brain tumors are treated with radiation therapy. Previous research has focused on task-based fMRI of language and sensory networks; however, there has been limited investigation on the inclusion of resting-state fMRI into the design of radiation treatment plans.

Methods and materials

In this pilot study of 9 patients with primary brain tumors, functional data from the default mode network (DMN), a network supporting cognitive functioning, was obtained from resting-state fMRI and retrospectively incorporated into the design of radiation treatment plans. We compared the dosimetry of these fMRI DMN avoidance treatment plans with standard of care treatment plans to demonstrate feasibility. In addition, we used normal tissue complication probability models to estimate the relative benefit of fMRI DMN avoidance treatment plans over standard of care treatment plans in potentially reducing memory loss, a surrogate for cognitive function.

Results

On average, we achieved 20% (P = 0.002) and 12% (P = 0.002) reductions in the mean and maximum doses, respectively, to the DMN without compromising the dose coverage to the planning tumor volume or the dose-volume constraints to organs at risk. Normal tissue complication probability models revealed that when the fMRI DMN was considered during radiation treatment planning, the probability of developing memory loss was lowered by more than 20%.

Conclusion

In this pilot study, we demonstrated the feasibility of including rs-MRI data into the design of radiation treatment plans to spare cognitively relevant brain regions during radiation therapy. These results lay the groundwork for future clinical trials that incorporate such treatment planning methods to investigate the long-term behavioral impact of this reduction in dose to the cognitive areas and their neural networks that support cognitive performance.

Résumé

Contexte

L'imagerie par résonance magnétique fonctionnelle (IRMf) permet de protéger de manière sélective des régions fonctionnellement importantes du cerveau lorsque des tumeurs cérébrales primaires sont traitées par radiothérapie. Des recherches antérieures ont mis l'accent sur l’IRMf basée sur les tâches des réseaux langagiers et sensoriels; il y a cependant eu peu d’études sur l'inclusion de l’IRMf au repos (IRMf-rs) dans la conception des plans de traitement en radiothérapie.

Méthodologie et matériel

Dans cette étude pilote mené auprès de neuf patients atteints de tumeurs cérébrales primaires, les données fonctionnelles du réseau du mode par défaut (MPD), un réseau qui soutient les fonctions cognitives, ont été obtenues par IRMf-rs et incorporées de manière rétrospective dans les plans de traitement. Nous avons comparé la dosimétrie de ces plans de traitement d’IRMf évitant le MPD avec les plans de traitement de la norme de soins afin d'en démontrer la faisabilité. De plus, nous avons utilisé des modèles de probabilité de complication des tissus sains (NTCP) pour estimer les avantages relatifs des plans de traitement sur les plans de traitement de la norme de soins dans la réduction potentielle de la perte de mémoire, un indicateur de la fonction cognitive.

Résultats

En moyenne, nous avons obtenu des réductions de 20% (P = 0,002) et de 12% (P = 0,002) de la dose moyenne et maximale, en regard du MDN, sans compromettre la couverture de dose du volume tumoral de planification ou les contraintes de dose-volume pour les organes à risque. Les modèles de NTCP montrent que lorsque l’IRMf évitant le MPD était envisagée durant la planification du traitement de radiothérapie, la probabilité de perte de mémoire était réduite de plus de 20%.

Conclusion

Dans cette étude pilote, nous avons démontré la faisabilité d'inclure les données d’IRMf dans la conception des plans de traitement en radiothérapie afin d’épargner les régions cognitivement pertinente du cerveau durant la radiothérapie. Ces résultats établissent les bases pour des essais cliniques futurs incorporant de telles méthodes de planification de traitement afin d’étudier les effets à long terme sur le comportement de cette réduction de dose sur les aires cognitives et les réseaux neuronaux qui appuient le rendement cognitif.

Introduction

Functional magnetic resonance imaging (fMRI) presents the ability to selectively protect functionally significant regions of the brain during radiation therapy (RT) of primary brain tumors [1]. Previous investigations have focused on incorporating task-based fMRI information from sensory networks (motor, visual, auditory) or language networks into conventional treatment plans [2], [3], [4], [5], [6], [7]. In addition, multiple investigators have demonstrated the feasibility of incorporating tissue-specific functional information derived from advanced imaging modalities into treatment plans. These studies include functional task-based information from magnetoencephalography [8], metabolic information from magnetic resonance spectroscopy [2], [3], as well as structural information from diffusion tensor imaging data [4], [7]. These studies provide evidence that the radiation dose to areas demonstrating functional significance on neuroimaging studies can be reduced while maintaining clinically acceptable doses of radiation to the tumor and organs at risk (OARs). However, in spite of the extensive research identifying cognitive impairments as a complication after radiotherapy [9], [10], [11], [12], there has been little focus on preserving functionally significant regions or networks that support cognitive processing. It stands to reason that if we are able to limit the dose to networks that are associated with cognitive performance, it may be possible to minimize treatment-related loss of cognitive functioning, ultimately improving long-term outcome and quality of life for these patients.

While brain tissue function data derived from task-based fMRI has great utility, the acquisition of task-based fMRI may not always be clinically feasible. Task-based fMRI requires that the subject be willing to participate in the task as well as be cognitively capable of performing the requested task. This task presents challenges for the population of patients with primary brain tumors, leading to inadequate compliance. Furthermore, it requires additional equipment and software that are used to display the task within the MRI environment and such equipment may not be readily available in a standard clinical facility. In recent years, the application of resting-state fMRI (rs-fMRI) has provided potential answers to the challenges associated with the acquisition of task-based fMRI data within a clinical setting. rs-fMRI does not require that participants perform a specific task during the MR scan, but rather that they lay at rest.

Referred to as resting-state functional connectivity, this technique is able to quantify the strength of functional interactions between brain regions based on coactivations of temporal fluctuations in the blood oxygen level–dependent (BOLD) signal that are present during resting conditions [13], [14]. Preliminary work in this field demonstrated that the regions of the motor network identified from task-based fMRI studies presented similar patterns of coactivation in the BOLD signal during resting-state conditions [15]. Further work in the field has allowed rs-fMRI to address questions regarding the strength of neural communication within and between various large-scale neural networks in the absence of a task. These resting-state networks are consistently replicated across studies and include networks that are associated with sensory processing (auditory, visual, somatosensory, and motor) and those associated with higher order cognitive processing [16].

One of the most highly investigated resting-state networks is the default mode network (DMN). The DMN is a set of cortical regions that demonstrate reduced neural activity during task-based activities, but increased activity during resting conditions [17]. The network includes nodes in the lateral parietal cortex, posterior cingulate cortex (PCC), and medial prefrontal cortex [18], [19]. The DMN has been associated with self-reflection, internally directed thoughts, and mind wandering [20]. Damage to this network has been associated with reduced cognitive performance in both healthy populations [21] as well as in multiple patient populations, including those with traumatic brain injury [22], [23], [24] and Alzhemier's disease [25]. Therefore, we propose that it may be important to specifically protect this neural network during radiation treatment planning and delivery.

The aim of this study was to retrospectively include rs-fMRI–guided protection of the DMN into the design of radiation treatment plans of 9 patients with primary brain tumors. A normal tissue complication probability (NTCP) model was used to estimate the relative benefit (by lowering the risk of developing cognitive impairments, eg, memory loss) derived from the fMRI DMN avoidance treatment plans compared with standard of care plans.