A dose prediction model for treatment planning was generated using U-Net architecture. The model was generated for advanced stage cancer patients. The U- Net architecture was created with depth=6 and kernel=6. The model architecture was successful to reduce the input image size (192X192) to feature map (6X6) which helped to extract the low level features. The dose prediction of the model was trained with depth=6, kernel=6, MSE loss, Adam optimizer, 1000 epochs and a batch size of 4. The predicted dose was rescaled for gamma analysis to quantify accuracy of the model. The renormalized predicted dose was quantified using gamma analysis with a 3mm, 3% dose tolerance. The gamma map was generated to visualize the regions where dose distributions failed. The gamma percentage values obtained on the training set were acceptable. The mean and standard deviation values of gamma pass percentage obtained on training dataset were 97.5% and 1.24% respectively, which concluded that training process was successful and was an almost perfect match of true dose and predicted dose. However, gamma pass percentage values obtained on validation set was not a good representation of the true dose. Nevertheless, the validation dataset was able to predict the approximate highest dose region. A gamma analysis with a 5mm, 5% dose tolerance was performed to test the the level of discrepancy between the predicted and true dose in the validation set. This increased the gamma pass percentage compared to the 3mm, 3% analysis to a mean gamma pass percentage of 26.2 ± 7.47%. Although the predicted dose was not of sufficient accuracy for clinical use, there technique studied in this work show promise for further development. / Thesis / Master of Science (MSc)

Machine learning has shown great potential as a step in automating radiotherapy treatment planning. It can be used for dose prediction and a popular deep learning architecture for this purpose is the U-net. Since it was proposed in 2015, several modifications and extensions have been proposed in the literature. In this study, three promising modifications are reviewed and implemented for dose prediction on a prostate cancer data set and compared with a 3D U-net as a baseline. The tested modifications are residual blocks, densely connected layers and attention gates. The different models are compared in terms of voxel error, conformity, homogeneity, dose spillage and clinical goals. The results show that the performance was similar in many aspects for the models. The residual blocks model performed similar or better than the baseline in almost all evaluations. The attention gates model performed very similar to the baseline and the densely connected layers were uneven in the results, often with low dose values in comparison to the baseline. The study also shows the importance of consistent ground truth data and how inconsistencies affect metrics such as isodose Dice score and Hausdorff distance.

Cancer is a group of diseases that are characterized by abnormal cell growth and is considered a leading cause of death globally. There are a number of different cancer treatment modalities, one of which is radiation therapy. In radiation therapy treatment planning, it is important to make sure that enough radiation is delivered to the tumor and that healthy organs are spared, while also making sure to account for uncertainties such as misalignment of the patient during treatment. To reduce the workload on clinics, data-driven automated treatment planning can be used to generate treatment plans for new patients based on previously delivered plans. In this thesis, we propose a novel method for robust automated treatment planning where a deep learning model is trained to deform a dose in accordance with a set of potential scenarios that account for the different uncertainties while maintaining certain statistical properties of the input dose. The predicted scenario doses are then used in a robust optimization problem with the goal of finding a treatment plan that is robust to these uncertainties. The results show that the proposed method for deforming doses yields realistic doses of high quality and that the proposed pipeline can potentially generate doses that conform better to the target than the current state of the art but at the cost of dose homogeneity. / Cancer är ett samlingsnamn för sjukdomar som karaktäriseras av onormal celltillväxt och betraktas som en ledande dödsorsak globalt. Det finns olika typer av cancerbehandling, varav en är strålterapi. Inom strålterapiplanering är det viktigt att säkerställa att tillräckligt med strålning ges till tumören, att friska organ skonas, och att osäkerheter som felplacering av patienten under behandlingen räknas med. För att minska arbetsbelastningen på kliniker används data-driven automatisk strålterapiplanering för att generera behandlingsplaner till nya patienter baserat på tidigare levererade behandlingar. I denna uppsats föreslår vi en ny metod för robust automatisk strålterapiplanering där en djupinlärningsmodell tränas till att deformera en dos i enlighet med en mängd potentiella scenarion som motsvarar de olika osäkerheterna medan vissa statistiska egenskaper bibehålls från originaldosen. De predicerade scenariodoserna används sedan i ett robust optimeringsproblem där målet är att hitta en behandlingsplan som är robust mot dessa osäkerheter. Resultaten visar att den föreslagna metoden för dosdeformation ger realistiska doser av hög kvalitet, vilket i sin tur kan leda till robusta doser med högre doskonformitet än tidigare metoder men på bekostnad av doshomogenitet.

The ability to quantify uncertainties associated with neural network predictions is crucial when they are relied upon in decision-making processes, especially in safety-critical applications like radiation therapy. In this paper, a single-model estimator of both epistemic and aleatoric uncertainties in a regression 3D U-net used for radiation dose prediction is presented. To capture epistemic uncertainty, Monte Carlo Dropout is employed, leveraging dropout during test-time inference to obtain a distribution of predictions. The variability among these predictions is used to estimate the model’s epistemic uncertainty. For quantifying aleatoric uncertainty quantile regression, which models conditional quantiles of the output distribution, is used. The method enables the estimation of prediction intervals of a user-specified significance level, where the difference between the upper and lower bound of the interval quantifies the aleatoric uncertainty. The proposed approach is evaluated on two datasets of prostate and breast cancer patient geometries and corresponding radiation doses. Results demonstrate that the quantile regression method provides well-calibrated prediction intervals, allowing for reliable aleatoric uncertainty estimation. Furthermore, the epistemic uncertainty obtained through Monte Carlo Dropout proves effective in identifying out-of-distribution examples, highlighting its usefulness for detecting anomalous cases where the model makes uncertain predictions. / Förmågan att kvantifiera osäkerheter i samband med neurala nätverksprediktioner är avgörande när de åberopas i beslutsprocesser, särskilt i säkerhetskritiska tillämpningar såsom strålterapi. I denna rapport presenteras en en-modellsimplementation för att uppskatta både epistemiska och aleatoriska osäkerheter i ett 3D regressions-U-net som används för att prediktera stråldos. För att fånga epistemisk osäkerhet används Monte Carlo Dropout, som utnyttjar dropout under testtidsinferens för att få en fördelning av prediktioner. Variabiliteten mellan dessa prediktioner används för att uppskatta modellens epistemiska osäkerhet. För att kvantifiera den aleatoriska osäkerheten används kvantilregression, eller quantile regression, som modellerar de betingade kvantilerna i outputfördelningen. Metoden möjliggör uppskattning av prediktionsintervall med en användardefinierad signifikansnivå, där skillnaden mellan intervallets övre och undre gräns kvantifierar den aleatoriska osäkerheten. Den föreslagna metoden utvärderas på två dataset innehållandes geometrier för prostata- och bröstcancerpatienter och korresponderande stråldoser. Resultaten visar på att kvantilregression ger välkalibrerade prediktionsintervall, vilket tillåter en tillförlitlig uppskattning av den aleatoriska osäkerheten. Dessutom visar sig den epistemiska osäkerhet som erhålls genom Monte Carlo Dropout vara användbar för att identifiera datapunkter som inte tillhör samma fördelning som träningsdatan, vilket belyser dess lämplighet för att upptäcka avvikande datapunkter där modellen gör osäkra prediktioner.