Medical physics最新文献

Background: Plan review is critical for safely delivering radiation dose to a patient under radiotherapy and mainly performed by medical physicist in routine clinical practice. Recently, the deep-learning models have been used to assist this manual process. As black-box models the reason for their predictions are unknown. Thus, it is important to improve the model interpretability to make them more reliable for clinical deployment.

Purpose: To alleviate this issue, a deep generative model, adversarial autoencoder networks (AAE), was employed to automatically detect anomalies in intensity-modulated radiotherapy plans.

Methods: The typical plan parameters (collimator position, gantry angle, monitor unit, etc.) were collected to form a feature vector for the training sample. The reconstruction error was the difference between the output and input of the model. Based on the distribution of reconstruction errors of the training samples, a detection threshold was determined. For a test plan, its reconstruction error obtained by the learned model was compared with the threshold to determine its category (anomaly or regular). The model was tested with four network settings. It was also compared with the vanilla AE and the other six classic models. The area under receiver operating characteristic curve (AUC) along with other statistical metrics was employed for evaluation.

Results: The AAE model achieved the highest accuracy (AUC = 0.997). The AUCs of the other seven classic methods are 0.935 (AE), 0.981 (K-means), 0.896 (principle component analysis), 0.978 (one-class support vector machine), 0.934 (local outlier factor), and 0.944 (hierarchical density-based spatial clustering of applications with noise), and 0.882 (isolation forest). This indicates that AAE model could detect more anomalous plans with less false positive rate.

Conclusions: The AAE model can effectively detect anomaly in radiotherapy plans for lung cancer patients. Comparing with the vanialla AE and other classic detection models, the AAE model is more accurate and transparent. The proposed AAE model can improve the interpretability of the results for radiotherapy plan review.

Background: Obtaining accurate segmentation regions for liver cancer is of paramount importance for the clinical diagnosis and treatment of the disease. In recent years, a large number of variants of deep learning based liver cancer segmentation methods have been proposed to assist radiologists. Due to the differences in characteristics between different types of liver tumors and data imbalance, it is difficult to train a deep model that can achieve accurate segmentation for multiple types of liver cancer.

Purpose: In this paper, We propose a balance Dice Loss(BD Loss) function for balanced learning of multiple categories segmentation features. We also introduce a comprehensive method based on BD Loss to achieve accurate segmentation of multiple categories of liver cancer.

Materials and methods: We retrospectively collected computed tomography (CT) screening images and tumor segmentation of 591 patients with malignant liver tumors from West China Hospital of Sichuan University. We use the proposed BD Loss to train a deep model that can segment multiple types of liver tumors and, through a greedy parameter averaging algorithm (GPA algorithm) obtain a more generalized segmentation model. Finally, we employ model integration and our proposed post-processing method, which leverages inter-slice information, to achieve more accurate segmentation of liver cancer lesions.

Results: We evaluated the performance of our proposed automatic liver cancer segmentation method on the dataset we collected. The BD loss we proposed can effectively mitigate the adverse effects of data imbalance on the segmentation model. Our proposed method can achieve a dice per case (DPC) of 0.819 (95%CI 0.798-0.841), significantly higher than baseline which achieve a DPC of 0.768(95%CI 0.740-0.796).

Conclusions: The differences in CT images between different types of liver cancer necessitate deep learning models to learn distinct features. Our method addresses this challenge, enabling balanced and accurate segmentation performance across multiple types of liver cancer.

Background: Volumetric modulated arc therapy (VMAT) revolutionizes cancer treatment by precisely delivering radiation while sparing healthy tissues. Fluence maps generation, crucial in VMAT planning, traditionally involves complex and iterative, and thus time consuming processes. These fluence maps are subsequently leveraged for leaf-sequence. The deep-learning approach presented in this article aims to expedite this by directly predicting fluence maps from patient data.

Purpose: To accelerate VMAT treatment planning by quickly predicting fluence maps from a 3D dose map. The predicted fluence maps can be quickly leaf sequenced because the network was trained to take into account the machine constraints.

Methods: We developed a 3D network which we trained in a supervised way using a combination of $L_1$ and $L_2$ losses, and radiation therapy (RT) plans generated by Eclipse and from the REQUITE dataset, taking the RT dose map as input and the fluence maps computed from the corresponding RT plans as target. Our network predicts jointly the 180 fluence maps corresponding to the 180 control points (CP) of single arc VMAT plans. In order to help the network, we preprocess the input dose by computing the projections of the 3D dose map to the beam's eye view (BEV) of the 180 CPs, in the same coordinate system as the fluence maps. We generated over 2000 VMAT plans using Eclipse to scale up the dataset size. Additionally, we evaluated various network architectures and analyzed the impact of increasing the dataset size.

Results: We are measuring the performance in the 2D fluence maps domain using image metrics (PSNR and SSIM), as well as in the 3D dose domain using the dose-volume histogram (DVH) on a test set. The network inference, which does not include the data loading and processing, is less than 20 ms. Using our proposed 3D network architecture as well as increasing the dataset size using Eclipse improved the fluence map reconstruction performance by approximately 8 dB in PSNR compared to a U-Net architecture trained on the original REQUITE dataset. The resulting DVHs are very close to the one of the input target dose.

Conclusions: We developed a novel deep learning approach for ultrafast VMAT planning by predicting all the fluence maps of a VMAT arc in one single network inference. The small difference of the DVH validate this approach for ultrafast VMAT planning.

Background: The nuclear Overhauser enhancement (NOE)-mediated saturation transfer effect at -1.6 ppm, termed NOE(-1.6 ppm), has demonstrated potential for detecting ischemic stroke. However, the quantification of the NOE(-1.6 ppm) effect usually relies on a multiple-pool Lorentzian fit method, which necessitates a time-consuming acquisition of the entire chemical exchange saturation transfer (CEST) Z-spectrum with high-frequency resolution, thus hindering its clinical applications.

Purpose: This study aims to assess the feasibility of employing asymmetry analysis, a rapid CEST data acquisition and analysis method, for quantifying the NOE(-1.6 ppm) effect in an animal model of ischemic stroke.

Methods: We examined potential contaminations from guanidinium/amine CEST, NOE(-3.5 ppm), and asymmetric magnetization transfer (MT) effects, which could reduce the specificity of the asymmetry analysis of NOE(-1.6 ppm). First, a Lorentzian difference (LD) analysis was used to mitigate direct water saturation and MT effects, providing separate estimations of the contributions from the guanidinium/amine CEST and NOE effects. Then, the asymmetry analysis of the LD fitted spectrum was compared with the asymmetry analysis of the raw CEST Z-spectrum to evaluate the contribution of the asymmetric MT effect at -1.6 ppm.

Results: Results show that the variations of the LD quantified NOE(-1.6 ppm) in stroke lesions are much greater than that of the CEST signals at +1.6 ppm and NOE(-3.5 ppm), suggesting that NOE(-1.6 ppm) has a dominating contribution to the asymmetry analysis at -1.6 ppm compared with the guanidinium/amine CEST and NOE(-3.5 ppm) in ischemic stroke. The NOE(-1.6 ppm) variations in the asymmetry analysis of the raw CEST Z-spectrum are close to those in the asymmetry analysis of the LD fitted spectrum, revealing that the NOE(-1.6 ppm) dominates over the asymmetric MT effects.

Conclusion: Our study demonstrates that the asymmetry analysis can quantify the NOE(-1.6 ppm) contrast in ischemic stroke with high specificity, thus presenting a viable alternative for rapid mapping of ischemic stroke.