Application of a multi-task learning-based light-weight convolution neural network for the automatic segmentation of organs at risk in thorax / 中华放射肿瘤学杂志

Objective:To evaluate the application of a multi-task learning-based light-weight convolution neural network (MTLW-CNN) for the automatic segmentation of organs at risk (OARs) in thorax.Methods:MTLW-CNN consisted of several layers for sharing features and 3 branches for segmenting 3 OARs. 497 cases with thoracic tumors were collected. Among them, the computed tomography (CT) images encompassing lung, heart and spinal cord were included in this study. The corresponding contours delineated by experienced radiation oncologists were ground truth. All cases were randomly categorized into the training and validation set ( n=300) and test set ( n=197). By applying MTLW-CNN on the test set, the Dice similarity coefficients (DSCs) of 3 OARs, training and testing time and space complexity (S) were calculated and compared with those of Unet and DeepLabv3+ . To evaluate the effect of multi-task learning on the generalization performance of the model, 3 single-task light-weight CNNs (STLW-CNNs) were built. Their structures were totally the same as the corresponding branches in MTLW-CNN. After using the same data and algorithm to train STLW-CNN, the DSCs were statistically compared with MTLW-CNN on the testing set. Results:For MTLW-CNN, the averages (μ) of lung, heart and spinal cord DSCs were 0.954, 0.921 and 0.904, respectively. The differences of μ between MTLW-CNN and other two models (Unet and DeepLabv3+ ) were less than 0.020. The training and testing time of MTLW-CNN were 1/3 to 1/30 of that of Unet and DeepLabv3+ . S of MTLW-CNN was 1/42 of that of Unet and 1/1 220 of that of DeepLabv3+ . The differences of μ and standard deviation (σ) of lung and heart between MTLW-CNN and STLW-CNN were approximately 0.005 and 0.002. The difference of μ of spinal cord was 0.001, but σof STLW-CNN was 0.014 higher than that of MTLW-CNN.Conclusions:MTLW-CNN spends less time and space on high-precision automatic segmentation of thoracic OARs. It can improve the application efficiency and generalization performance of the models.

A study of prediction model of lung dose in early stage non-small cell lung cancer with stereotactic body radiotherapy / 中华放射肿瘤学杂志

Objective@#To study a lung dose prediction method for the early stage non-small cell lung cancer (NSCLC) treated with stereotactic body radiotherapy based on machine learning algorithm, and to evaluate the feasibility of application in planning quality assurance.@*Methods@#A machine learning algorithm was utilized to achieve DVH prediction. First, an expert plan dataset with 125 cases was built, and the geometric features of ROI, beam angle and dose-volume histogram(DVH) parameters in the dataset were extracted. Following a correlation model was established between the features and DVHs. Second, the geometric and beam features from 10 cases outside the training pool were extracted, and the model was adopted to predict the achievable DVHs values of the lung. The predicted DVHs values were compared with the actual planned results.@*Results@#The mean squared errors of external validation for the 10 cases in mean lung dose (MLD)MLD and V20 of the lung were 91.95 cGy and 3.12%, respectively. Two cases whose lung doses were higher than the predicted values were re-planned, and the results showed that the the lung doses were reduced.@*Conclusion@#It is feasible to utilize the anatomy and beam angle features to predict the lung DVH parameters for plan evaluation and quality assurance in early stage NSCLC patients treated with stereotactic body radiotherapy

Study of three-dimensional dose distribution prediction model in radiotherapy planning based on full convolution network / 中华放射肿瘤学杂志

Objective:To explore a three-dimensional dose distribution prediction method for the left breast cancer radiotherapy planning based on full convolution network (FCN), and to evaluate the accuracy of the prediction model.Methods:FCN was utilized to achieve three-dimensional dose distribution prediction. First, a volumetric modulated arc therapy (VMAT) plan dataset with 60 cases of left breast cancer was built. Ten cases were randomly chosen from the dataset as the test set, and the remaining 50 cases were used as the training set. Then, a U-Net model was built with the organ structure matrix as inputs and dose distribution matrix as outputs. Finally, the model was adopted to predict the dose distribution of the cases in the test set, and the predicted 3D doses were compared with actual planned results.Results:The mean absolute differences of PTV, ipsilateral lung, heart, whole lung and spinal cord for 10 cases were (119.95±9.04) cGy, (214.02±9.04) cGy, (116.23±30.96) cGy, (127.67±69.19) cGy, and (37.28±18.66) cGy, respectively. The Dice similarity coefficient (DSC) of the prediction dose and the planned dose in the 80% and 100% prescription dose range were 0.92±0.01 and 0.92±0.01. The γ rate of 3 mm/3% in the area of 80% and 10% prescription dose range were 0.85±0.03 and 0.84±0.02. Conclusion:FCN can be used to predict the three-dimensional dose distribution of left breast cancer patients undergoing VMAT.

A study of prediction model of lung dose in early stage non-small cell lung cancer with stereotactic body radiotherapy / 中华放射肿瘤学杂志

Objective To study a lung dose prediction method for the early stage non-small cell lung cancer (NSCLC) treated with stereotactic body radiotherapy based on machine learning algorithm,and to evaluate the feasibility of application in planning quality assurance.Methods A machine learning algorithm was utilized to achieve DVH prediction.First,an expert plan dataset with 125 cases was built,and the geometric features of ROI,beam angle and dose-volume histogram(DVH) parameters in the dataset were extracted.Following a correlation model was established between the features and DVHs.Second,the geometric and beam features from 10 cases outside the training pool were extracted,and the model was adopted to predict the achievable DVHs values of the lung.The predicted DVHs values were compared with the actual planned results.Results The mean squared errors of external validation for the 10 cases in mean lung dose (MLD) MLD and V20 of the lung were 91.95 cGy and 3.12％,respectively.Two cases whose lung doses were higher than the predicted values were re-planned,and the results showed that the the lung doses were reduced.Conclusion It is feasible to utilize the anatomy and beam angle features to predict the lung DVH parameters for plan evaluation and quality assurance in early stage NSCLC patients treated with stereotactic body radiotherapy

Evaluation of different volumetric-modulated arc therapy for the large planning target volume in T3 lung cancer / 中国基层医药

Objective To investigate the feasibility and dosimetric characteristics of using dual - arc volumetric modulated arc therapy and multiple partial-arc VMAT for T3 lung cancer.Methods From June 2016 to May 2018,thirteen lung cancer patients with large planning target volume were replanned with dual full arcs VMAT (F-VMAT) and six partial-arc s VMAT( P-VMAT) on RayStation v4.5 RayArc function.PTV volume median was 550.9cm3(ranged 402.2-834.8cm3 ) and to a prescribed dose of 60 Gy in 30 fractions.Equivalent target coverage was required for all plans,and clinical goals were evaluated using various dose-volume metrics.These included PTV dose conformity,mean lung/heart dose,lung V5 ,V10 ,V20 ,V30 ,heart V30 and V40 ,and Dmax of spinal canal.The total monitor units ( MUs) were also examined. Results All VMAT plans satisfied the treatment criteria. F - VMAT achieved better homogeneity index ( HI) and MUs than P -VMRT( t = -3.904,P =0.002),and the conformal number(CN) of tumor volumes was likely clinically indistinguishable.However,F-VMAT significantly reduced lung V5 ,V10 and mean lung dose[V5:(51.31 ± 5.36)% vs.(43.44 ± 5.28)%,t=6.908,P=0.00;V10:(38.34 ± 3.26)% vs.(34.05 ± 3.74)%,t=4.632,P=0.001;Dmean:(1 449 ± 117.19)cGy vs.(1 375.38 ± 148.98)cGy, t=4.93, P =0.00 ], and heart dosimetric parameters were also observed in favor of P - VMRT [ V30 : (20.6 ± 10.4)% vs.(16.4 ± 8.9)%,t =3.822,P =0.02;V40:(14.6 ± 7.5)% vs.(11.88 ± 7.1)%,t =3.096,P =0.009;Dmean:(1 442.9 ± 651.2)cGy vs.(1 263.5 ± 605.6)cGy,t=3.986,P=0.02],and there were no statisti-cally significant differences in lung V20,V30 and spinal cord Dmax between the two groups(all P>0.05).Conclusion VMAT is an effective treatment for stage T3 lung cancer patients. The primary advantage of P - VMAT was the reduction in low dose area and decreased risk of symptomatic radioactive lung injury.It may be a priority for pulmonary malignancy patients with the large planning target volume.

Evaluation of different volumetric-modulated arc therapy for the large planning target volume in T3 lung cancer / 中国基层医药

Objective@#To investigate the feasibility and dosimetric characteristics of using dual-arc volumetric modulated arc therapy and multiple partial-arc VMAT for T3 lung cancer.@*Methods@#From June 2016 to May 2018, thirteen lung cancer patients with large planning target volume were replanned with dual full arcs VMAT(F-VMAT) and six partial-arc s VMAT(P-VMAT)on RayStation v4.5 RayArc function.PTV volume median was 550.9cm3(ranged 402.2-834.8cm3) and to a prescribed dose of 60 Gy in 30 fractions.Equivalent target coverage was required for all plans, and clinical goals were evaluated using various dose-volume metrics.These included PTV dose conformity, mean lung/heart dose, lung V5, V10, V20, V30, heart V30 and V40, and Dmax of spinal canal.The total monitor units (MUs) were also examined.@*Results@#All VMAT plans satisfied the treatment criteria.F-VMAT achieved better homogeneity index(HI) and MUs than P-VMRT(t=-3.904, P=0.002), and the conformal number(CN) of tumor volumes was likely clinically indistinguishable.However, F-VMAT significantly reduced lung V5, V10 and mean lung dose[V5: (51.31±5.36)% vs.(43.44±5.28)%, t=6.908, P=0.00; V10: (38.34±3.26)% vs.(34.05±3.74)%, t=4.632, P=0.001; Dmean: (1 449±117.19)cGy vs.(1 375.38±148.98)cGy, t=4.93, P=0.00], and heart dosimetric parameters were also observed in favor of P-VMRT[V30: (20.6±10.4)% vs.(16.4±8.9)%, t=3.822, P=0.02; V40: (14.6±7.5)% vs.(11.88±7.1)%, t=3.096, P=0.009; Dmean: (1 442.9±651.2)cGy vs.(1 263.5±605.6)cGy, t=3.986, P=0.02], and there were no statistically significant differences in lung V20, V30 and spinal cord Dmax between the two groups(all P>0.05).@*Conclusion@#VMAT is an effective treatment for stage T3 lung cancer patients.The primary advantage of P-VMAT was the reduction in low dose area and decreased risk of symptomatic radioactive lung injury.It may be a priority for pulmonary malignancy patients with the large planning target volume.

Digital signal processing of a novel neuron discharge model stimulation strategy for cochlear implants / 南方医科大学学报

<p><b>OBJECTIVE</b>To apply the classic leakage integrate-and-fire models, based on the mechanism of the generation of physiological auditory stimulation, in the information processing coding of cochlear implants to improve the auditory result.</p><p><b>METHODS</b>The results of algorithm simulation in digital signal processor (DSP) were imported into Matlab for a comparative analysis.</p><p><b>RESULTS</b>Compared with CIS coding, the algorithm of membrane potential integrate-and-fire (MPIF) allowed more natural pulse discharge in a pseudo-random manner to better fit the physiological structures.</p><p><b>CONCLUSION</b>The MPIF algorithm can effectively solve the problem of the dynamic structure of the delivered auditory information sequence issued in the auditory center and allowed integration of the stimulating pulses and time coding to ensure the coherence and relevance of the stimulating pulse time.</p>