A machine learning framework with anatomical prior for online dose verification using positron emitters and PET in proton therapy

Abstract

We developed a machine learning framework in order to establish the correlation between dose and activity distributions in proton therapy. A recurrent neural network was used to predict dose distribution in three dimensions based on the information of proton-induced positron emitters. Hounsfield Unit (HU) information from CT images and analytically derived stopping power (SP) information were incorporated as auxiliary inputs. Four different scenarios were investigated: Activity only, Activity + HU, Activity + SP and Activity + HU + SP. The performance was quantitatively studied in terms of mean absolute error (MAE) and mean relative error (MRE), under different signal-to-noise ratios (SNRs). In addition to the first dataset of mono-energetic beams, three additional datasets were validated to help evaluate the generalization capability of our proposed model: a dataset of a lower SNR, five reconstructed PET images, and a dataset of spread-out Bragg peaks. Good verification accuracy of dose verification in three dimensions is demonstrated. The inclusion of anatomical information improves both accuracy and generalization. For an activity profile with an SNR of 4 (the mono-energetic case), the framework is able to obtain an MRE of ∼ 0.99% over the whole range and a range uncertainty of ∼ 0.27 mm. The machine learning-based framework may emerge as a useful tool to allow for online dose verification and quality assurance in proton therapy.

Direct mapping from PET coincidence data to proton-dose and positron activity using a deep learning approach

“…Large-size ring PET solutions have been proposed for dose mapping, but our study proposes the use of a small palm-sized PET detector solution leveraging the deep learning framework to optimize for the minimum number of coincidences needed for reconstruction. Our dual-head setup has a geometrically active area 100 times smaller than the PET solution proposed in the previous study (Hu et al 2020). In Hu et al, there were 88,000 crystals, whereas we have only 1024 crystal elements and a complete detector with 10 times smaller geometric size and 85 times less calibration load.…”

“…However, the (Liu et al 2019) study was based on the intrinsic isotopic distributions that cannot be measured and hence cannot be verified experimentally. The second sequence of the RNN model-based study combined the PET detector model, proton beam stopping power, and anatomical information using the reconstructed dose-mapping profile (Hu et al 2020). Another study (Zhang et al 2021) used deep learning methods to predict proton dose in 3D.…”

Machine learning based oxygen and carbon concentration derivation using dual‐energy CT for PET‐based dose verification in proton therapy

“…For quantitative analysis, mean relative error (MRE) was used to evaluate the difference between the simulated activity and their counterparts in the ground truth, in the same manner as in our previous work 18,19,26 : $$\begin{equation}{\rm{MRE}} = \frac{1}{N}\mathop \sum \limits_i \frac{{{\rm{\;}}\left| {{\rm{\;}}\hat y_i - y_i{\rm{\;}}} \right|}}{{\mathop {{\rm{max\;}}}\limits_i (\hat y_i)}}\end{equation}$$where N is the number of pixels included for calculation.$$\;{y_i}$$ refers to the predicted outcome (dose/activity), $${\hat y_i}$$ refers to the ground truth (dose/activity). MRE95 corresponds to the result of those pixels of no less than 95% of the peak value in each profile (activity/dose).…”

“…Our group recently published several papers focusing on the use of machine learning-based approaches for range and dose verification. [16][17][18][19] Tissue composition, in particular the fraction of carbon and oxygen, directly affects the production yield of positron emitters in proton therapy. On one hand, their concentrations, together with cross-section values (i.e., reaction channels), impact the yield of positron emitters and thus the shape of activity profiles.…”

Long short‐term memory networks for proton dose calculation in highly heterogeneous tissues

“…33 Naturally, these limitations regarding fluence extend to the general beam setup, that is, the couch and gantry angle. To the best of our knowledge, only the approach by Liu et al 34,35 focuses on the prediction of individual pencil beams, however, only considering a one-dimensional (1D) mapping of the activity distribution of the positron emitters to the 1D pencil beam dose distribution.…”