Microdosimetry

At a scale comparable to a cell nucleus, where the biological damage induced by radiation occurs, the energy deposition is affected by stochastic fluctuations and cannot be accurately described with macroscopic mean values, such as the dose or the linear energy transfer (LET). Microdosimetry has been proposed as a methodology for characterizing radiation field quality in micrometer-sized volumes and represents the bridge between physical characteristics of the radiation and biological outcomes. It has been exploited by radiobiological models, such as the Microdosimetric Kinetic Model (MKM), for predicting cell survival and relative biological effectiveness (RBE). We are using microdosimetry to characterize the in- and out-of-field radiation field in particle therapy, and use it as a tool for assessing biological outcomes. We are also developing a new hybrid microdosimeter (HDM), which has unique features than existing detectors and will allow for a superior characterization of radiation quality in particle therapy.

Radiobiological Models

Radiobiological models for cell survival and RBE prediction are a key part of treatment planning systems (TPS) for calculating the dose delivered. An inaccurate determination of the RBE can lead to an underdosage to the tumor, limiting treatment success, or an overdosage to normal tissue, increasing the probability of complications. The main limitation shared by all existing models is the assumption that all variables follow a Poisson distribution and thus can be described by their mean. This assumption neglects stochastic fluctuations of energy deposition both from cell to cell and from dose fractionation (time variable), which can be especially significant in highly mixed radiation fields that occur at the beam edges and in the distal region. We have developed the general stochastic microdosimetry-based kinetic model (GSM2), that consider the stochastic behavior of both energy deposition and cell inactivation, providing the radiation response of the irradiated tissue, and predict both cell survival and RBE.

Artificial Intelligence applied to Radiation Oncology

Among the major revolution of our century, artificial intelligence and data analysis are without any doubt among the most relevant. Machine and Deep Learning (ML and DL) models are nowadays systematically used in many fields. We recently started gaining interest in application of machine and deep learning techniques applied to medical physics and radiation oncology. We have implemented a machine learning model to track particles in microdosimetry to overcome experimental limitations of our new detector HDM. We have also developed a fully machine and deep learning algorithm to predict biological effectiveness of a wide range of ions relevant both for radiotherapy and for radioprotection (ANAKIN).

Proton therapy

Proton therapy has emerged as an attractive alternative for cancer treatment because of its superior spatial dose distribution in the patient. This unique characteristic translates into a lower radiation exposure of normal tissue surrounding the tumor, and thus leads to a reduced incidence of side effects.

Treatment Plan Optimization

The current clinical practice assumes a constant RBE, disregarding any dependence on dose, radiation quality and biological endpoint, and might results in an incorrect estimate of the dose delivered during treatment, especially to normal tissue. Taking advantage of our new models (GSM2 and ANAKIN), we are working to implement variable RBE into commercial TPS. Furthermore, we are developing machine learning-based normal tissue complications probability (NTCP) models for various toxicities, which can be also implemented into TPSs to assess the risk of complications for individual patients, and further optimize the dose plans to reduce complications.

Treatment Plan Verification

To further improve treatment effectiveness, we are investigating a new approach to monitor in-patient proton range. TPS rely on imaging approaches to locate tumors, but there are sizable uncertainties at the time of irradiation due to anatomical modifications, patient alignment, beam delivery and dose calculation. A mispositioning potentially translates into an under-dosage of the tumor as well as an over-dosage of the normal tissue, which can significantly hinder treatment efficacy. We developed a novel strategy for real-time range and dose verification. The methodology is based on the detection of prompt gammas, whose production is enhanced with a non-radioactive element transported selectively to the tumor with a drug carrier. Nuclear interactions of this element with protons generate a signature gamma spectrum, whose intensity is correlated with beam range, and from which the tumor position can be reconstructed.