Current clinical practice in radiotherapy is predominantly based on the physical optimisation of the dose distribution according to the anatomical information of the tumour and normal tissues. This often results in treatment strategies where homogenous dose distributions are planned to be delivered to the targets of interest. In most cases, clinical practice does not account for the radiosensitivity of individual patients, nor does it for the spatial and temporal heterogeneities of the tumour radioresistance before or during treatment. However, it is well known that the biological heterogeneity of the tumour may occur at multiple levels - as inter- and intra- patient variations - and evidence is accumulating that this may be one of the (if not the) major causes of tumour recurrence and consequent treatment failure. The genetic heterogeneities of the tumour, the varying metabolism of the tumour cells, different cellular densities, and tumour hypoxia lead to varying levels of radiation sensitivity within the same tumour mass. Dose escalation strategies targeting those tumour sub-volumes may increase local control, provided that meaningful radiobiological targets are defined. Functional imaging, such as positron emission tomography (PET), magnetic resonance imaging, or combined multiparametric modalities, may have a crucial role not only in the identification of those tumour sub-regions of increased radioresistance for a biologically guided radiotherapy strategy but also in the monitoring of the responsiveness of the tumour to the treatment in a framework of response-adaptive radiotherapy.
In the case of PET imaging of hypoxia, several dose sculpting strategies have been proposed to escalate the dose in radioresistant tumour sub-volumes. They include empirical approaches exclusively based on clinical experience, methods based on a linear conversion of radiotracer uptake into dose levels, and more sophisticated strategies guided by radiobiological modelling where the computation of the dose distribution required to achieve a desired level of tumour control probability (TCP) is considered. This talk will overview the proposed dose painting strategies, going from the "classical" dose painting by numbers, dose painting by contours and dose redistribution methods to alternative dose clustering methods more recently proposed. To date, the potential of hypoxia-PET imaging for quantitatively characterising the microenvironment and determining the required dose escalation level has yet to be fully explored. Indeed, one of the main challenges when dealing with hypoxic treatment resistance is the inherent multiscale feature of the problem where the functional imaging of hypoxia is limited to the millimetre regime, while the oxygen diffusion varies on a sub-millimetre scale. Consequently, the limitation of the PET resolution risks overshadowing the ability of this imaging technique to represent the microenvironment faithfully. Hence, developing robust dose sculpting methods to bridge the resolution gap would be timely and desirable to improve current treatment standards. As an example, two different dose prescription strategies based on hypoxia PET imaging for both photons and protons will be compared in terms of TCP by considering an in-silico modelled underlying oxygen distribution at a sub-millimetre scale.
If pre-treatment functional imaging investigations could provide initial information on the dose levels required to counteract the specific radioresistance of the tumour for each patient, subsequent examinations early during the treatment could then provide information on tumour responsiveness. This information may subsequently be used to determine the need for treatment adaptation and allow for individual adjustments of the radiation dose, also taking into account the accumulated delivered dose distribution at the repeated imaging time. While biologically guided radiotherapy and individualisation of cancer treatment remain two of the ultimate aims of modern radiotherapy, their translation into clinical practice remains challenging.