Search

Fast and accurate delineation of organs on treatment-fraction images is critical in magnetic resonance imaging-guided adaptive radiotherapy (MRIgART). This study proposes a personalized auto-segmentation (AS) framework to assist online delineation of prostate cancer using MRIgART.Image data from 26 patients diagnosed with prostate cancer and treated using hypofractionated MRIgART (5 fractions per patient) were collected retrospectively. Daily pretreatment T2-weighted MRI was performed using a 1.5-T MRI system integrated into a Unity MR-linac. First-fraction image and contour data from 16 patients (80 image-sets) were used to train the population AS model, and the remaining 10 patients composed the test set. The proposed personalized AS framework contained two main steps. First, a convolutional neural network was employed to train the population model using the training set. Second, for each test patient, the population model was progressively fine-tuned with manually checked delineations of the patient's current and previous fractions to obtain a personalized model that was applied to the next fraction.Compared with the population model, the personalized models substantially improved the mean Dice similarity coefficient from 0.79 to 0.93 for the prostate clinical target volume (CTV), 0.91 to 0.97 for the bladder, 0.82 to 0.92 for the rectum, and 0.91 to 0.93 for the femoral heads, respectively.The proposed method can achieve accurate segmentation and potentially shorten the overall online delineation time of MRIgART.

Purpose To develop a deep-learning method to predict patient-specific dose volume histograms (DVHs) for radiotherapy planning. Methods Patient data included 180 cases with nasopharyngeal cancer, of which 153 cases were used for training and 27 for testing. A network (named 'DVHnet') based on a convolutional neural network (CNN) was designed for directly predicting DVHs of organs at risk (OARs). Two-channel images with contoured structures were generated as the inputs for training the model. A one-dimensional array consisting of 256 continuous volume percentages on a DVH curve for each slice was calculated as the corresponding output. The combined DVH was then calculated. Sixteen OARs were modeled in the study. Prediction accuracy was evaluated against the corresponding DVH curve of ground truth (GT) plans. A global DVH analysis and critical dosimetry metrics for each OAR were calculated for quantitative evaluation. The performance of DVHnet also was evaluated against two baselines: DosemapNet (developed by our research group) and commercial RapidPlan software. Results The predicted mean differences in average dose of all OARs using DVHnet were 0.30 ± 0.95Gy. And the predicted differences in D2% and D50 can be control within 2.32Gy and 0.69Gy. For most OARs, there were no obvious differences between the dosimetric metrics of the predicted and GT values for both DVHnet and DosemapNet (p≥0.05). Only the predicted D2% of the optic organs for DVHnet, and of brain stem PRV for DosemapNet displayed statistically significant differences. Except for the optic organs, DVHnet performs better than or comparably with RapidPlan. The mean difference in proportion of points of interest were 3.59% ± 7.78%. Conclusions A deep-learning network model was developed to automatically extract useful features for accurate prediction of patient-specific DVH curves directly. The performance of DVHnet was comparable to DosemapNet and RapidPlan.

Purpose To develop a deep learning-based model to predict achievable dose-volume histograms (DVHs) of organs at risk (OARs) for automation of inverse planning. Methods The model was based on a connected residual deconvolution network (CResDevNet) and compared with UNet as a baseline. The DVHs of OARs are dependent on patient anatomical features of the planning target volumes and OARs, and these spatial relationships can be learned automatically from prior high-quality plans. The contours of planning target volumes and OARs were parsed from the plan database and used as the input to the model, and the dose-area histograms (DAHs) of the OARs were output from the model. The model was trained from scratch by correlating anatomical features with DAHs of OARs, then accumulating these histograms to obtain the final predicted DVH for each OAR. Helical tomotherapy plans for 170 nasopharyngeal cancer patients were used to train and validate the model. An additional 60 patient treatment plans were used to test the predictive accuracy of the model. The DVHs and dose-volume indices (DVIs) of clinical interest for each OAR in the testing dataset were predicted to evaluate the accuracy of the models. The mean absolute errors in the DVIs for each OAR were calculated using each model and statistically compared using a paired-samples t-test. Dice similarity coefficients for areas of the DVHs were also evaluated. Results Dose-volume histograms of 21 OARs in nasopharyngeal cancer were predicted using the models. For each patient, 63 DVIs for all OARs were calculated. Using the 60 patient treatment plans in the testing dataset, 79% and 73% of the DVIs predicted using the CResDevNet and UNet models, respectively, were within 5% of the clinical values. The median value of the DVIs' mean absolute errors was 3.2 ± 2.5% and 3.7 ± 2.9% for the CResDevNet and UNet models, respectively. The average dice similarity coefficient for all OARs was 0.965 using the CResDevNet model and 0.958 using the UNet model. Conclusions A deep learning model was developed for predicting achievable DVHs of OARs. The prediction accuracy of the CResDevNet model was evaluated using a planning database of nasopharyngeal cancer cases and shown to be more accurate than the UNet model. Prediction accuracy was also higher for larger-volume OARs. The model can be used for automation of inverse planning and quality assessment of individual treatment plans.

Purpose To develop a deep learning method for prediction of three-dimensional (3D) voxel-by-voxel dose distributions of helical tomotherapy (HT). Methods Using previously treated HT plans as training data, a deep learning model named U-ResNet-D was trained to predict a 3D dose distribution. First, the contoured structures and dose volumes were converted from plan database to 3D matrix with a program based on a developed visualization toolkit (VTK), then transferred to U-ResNet-D for correlating anatomical features and dose distributions at voxel-level. One hundred and ninety nasopharyngeal cancer (NPC) patients treated by HT with multiple planning target volumes (PTVs) in different prescription patterns were studied. The model was typically trained from scratch with weights randomly initialized rather than using transfer-learning method, and used to predict new patient's 3D dose distributions. The predictive accuracy was evaluated with three methods: (a) The dose difference at the position r, δ(r, r) = Dc (r) - Dp (r), was calculated for each voxel. The mean (μδ(r,r) ) and standard deviation (σδ(r,r) ) of δ(r, r) were calculated to assess the prediction bias and precision; (b) The mean absolute differences of dosimetric indexes (DIs) including maximum and mean dose, homogeneity index, conformity index, and dose spillage for PTVs and organ at risks (OARs) were calculated and statistically analyzed with the paired-samples t test; (c) Dice similarity coefficients (DSC) between predicted and clinical isodose volumes were calculated. Results The U-ResNet-D model predicted 3D dose distribution accurately. For twenty tested patients, the prediction bias ranged from -2.0% to 2.3% and prediction error varied from 1.5% to 4.5% (relative to prescription) for 3D dose differences. The mean absolute dose differences for PTVs and OARs are within 2.0% and 4.2%, and nearly all the DIs for PTVs and OARs had no significant differences. The averaged DSC ranged from 0.95 to 1 for different isodose volumes. Conclusions The study developed a new deep learning method for 3D voxel-by-voxel dose prediction, and shown to be able to produce accurately dose predictions for nasopharyngeal patients treated by HT. The predicted 3D dose map can be useful for improving radiotherapy planning design, ensuring plan quality and consistency, making clinical technique comparison, and guiding automatic treatment planning.

Purpose To develop a method for predicting optimal dose distributions, given the planning image and segmented anatomy, by applying deep learning techniques to a database of previously optimized and approved Intensity‐modulated radiation therapy treatment plans. Methods Eighty cases of early‐stage nasopharyngeal cancer (NPC) were included in the study. Seventy cases were chosen randomly as the training set and the remaining as the test set. The inputs were the images with structures, with each target and organs at risk (OARs) assigned a unique label. The outputs were dose maps, including coarse dose maps and converted fine dose maps (FDM) from convolution. Two types of input images with structures were used in the model building. One type of input included the images (with associated structures) without manipulation. The second type of input involved modifying the image gray label with information from radiation beam geometry. ResNet101 was chosen as the deep learning network for both. The accuracy of predicted dose distributions was evaluated against the corresponding dose as used in the clinic. A global three‐dimensional gamma analysis was calculated for the evaluation. Results The proposed model trained with the two different sets of input images and structures could both predict patient‐specific dose distributions accurately. For the out‐of‐field dose distributions, the model obtained from the input with radiation geometry performed better (dose difference in %, 4.7 ± 6.1% vs 5.5 ± 7.9%, P 0.05), except for the bilateral optic nerves and the optic chiasm. Conclusions The proposed system with radiation geometry added to the input is a promising method to generate patient‐specific dose distributions for radiotherapy. It can be applied to obtain the dose distributions slice‐by‐slice for planning quality assurance and for guiding automated planning.

Purpose Delineation of the clinical target volume (CTV) and organs at risk (OARs) is very important for radiotherapy but is time-consuming and prone to inter-observer variation. Here, we proposed a novel deep dilated convolutional neural network (DDCNN)-based method for fast and consistent auto-segmentation of these structures. Methods Our DDCNN method was an end-to-end architecture enabling fast training and testing. Specifically, it employed a novel multiple-scale convolutional architecture to extract multiple-scale context features in the early layers, which contain the original information on fine texture and boundaries and which are very useful for accurate auto-segmentation. In addition, it enlarged the receptive fields of dilated convolutions at the end of networks to capture complementary context features. Then, it replaced the fully connected layers with fully convolutional layers to achieve pixel-wise segmentation. We used data from 278 patients with rectal cancer for evaluation. The CTV and OARs were delineated and validated by senior radiation oncologists in the planning computed tomography (CT) images. A total of 218 patients chosen randomly were used for training, and the remaining 60 for validation. The Dice similarity coefficient (DSC) was used to measure segmentation accuracy. Results Performance was evaluated on segmentation of the CTV and OARs. In addition, the performance of DDCNN was compared with that of U-Net. The proposed DDCNN method outperformed the U-Net for all segmentations, and the average DSC value of DDCNN was 3.8% higher than that of U-Net. Mean DSC values of DDCNN were 87.7% for the CTV, 93.4% for the bladder, 92.1% for the left femoral head, 92.3% for the right femoral head, 65.3% for the intestine and 61.8% for the colon. The test time was 45 s per patient for segmentation of all the CTV, bladder, left and right femoral heads, colon and intestine. We also assessed our approaches and results with those in the literature: our system showed superior performance and faster speed. Conclusions These data suggest that DDCNN can be used to segment the CTV and OARs accurately and efficiently. It was invariant to the body size, body shape, and age of the patients. DDCNN could improve the consistency of contouring and streamline radiotherapy workflows. This article is protected by copyright. All rights reserved.

Purpose: To evaluate the influence of carbon fiber couchtop and patient positioning devices on cone beam CT(CBCT)image quality and develop an effective method to correct the influence. Methods: A standard CT phantom (Catphan 500) was used to evaluate the influence of iBeam evo carbon fiber couchtop on the quality of CBCTimage obtained from an Elekta synergy machine. The evaluation indices were contrast-to-noise ratio(CNR),spatial resolution,image uniformity, and imagenoise. With using the Beer–Lambert law and the energy-response function of the flat-panel imager, a method was applied to deduct the image signal of the couchtop (and the positioning devices) from each projection image of a phantom/patient, and then used all corrected projection images to reconstruct a CBCTimage. The performance of the correction method was evaluated using the CBCTimages of a Catphan 500 phantom, a head-and-neck cancer patient, and a prostate cancer patient. Results: In two phantom studies (the phantom to simulate a human head and neck and the one to simulate a human body), the CNR of the CBCTimages obtained with the couchtop reduced 18.1% and 29.8%, respectively with respect to those obtained without the couchtop; meanwhile, the image uniformity reduced 16.4% and 24.1% due to the use of the carbon fiber couchtop. The couchtop also induced extra imagenoise (16.5% for the h&n phantom and 4.2% for the body phantom). However, CBCTimaging with the couchtop did not affect the spatial resolution. After applying the projection image correction, there was a significant improvement in CNR (by 19.5% and 25.8%), image uniformity (by 9.2% and 13.1%), and imagenoise (by 10.2% and 3.9%), with respect to CBCTimages obtained with the couchtop. Conclusions: The presence of the carbon fiber couchtop and the patient positioning devices can significantly impair CBCTimage quality in terms of the CNR, the image uniformity, and the imagenoise. By removing the influence of the couchtop and the patient-positioning devices from CB projection images, the correction method improves CBCTimage quality and thus image guidance in radiotherapy.

Segmental intensity-modulated radiation therapy (IMRT) delivers a sequence of segments to obtain a desired intensity distribution. Many leaf sequencing algorithms for segmental IMRT have been developed with the aim to reduce delivered monitor units (MUs) and (or) number of segments, and consequently to reduce the total treatment delivery time. With the development of real-time detection technology, it is useful to develop leaf sequencing algorithms that consider the detecting probability of markers implanted into or near target volume. In this study, we defined the concept of marker visibility to denote the marker’s detecting probability, and proposed a new leaf sequencing algorithm based on the algorithm of Kamath et al (Phys. Med. Biol. 48: 307-324, 2003). The new algorithm firstly uses Kamath algorithm to generate an initial leaf sequence and then performs a series of column transformations to obtain a new leaf sequence that is optimal in terms of MU efficiency and marker visibility. We evaluated the performance of the new algorithm with six test fields that had randomly-generated intensity matrices. Results show that, compared with Kamath algorithm, the new algorithm does not increase MUs, but improves marker visibility. When the radiation field contains multiple (3) markers, the marker visibility increased from 66.67% to 91.67% for small size (5×5) radiation field; from 39.29% to 42.86% for medium size (10×10 ) field; and from 31.48% to 37.04% for large size ( 20× 20 ) field. We also rigorously proved the optimality of the new algorithm. In conclusion, a new leaf sequencing algorithm was developed for optimal MU efficiency and marker visibility.

This study introduces a new concept, the super-omni wedge, and proposes an algorithm for optimizing beam weights, wedge angles, and wedge orientations on the basis of this new concept. The super-omni wedge is a generalization of the omni wedge. Instead of combining one open beam and two orthogonal wedged beams, it uses two orthogonal pairs of nominal wedged beams to generate a wedged dose distribution with an arbitrary wedge angle and an arbitrary wedge orientation. The orientations of a pair of nominal wedges are opposite each other. In this way, the effective wedge orientation can vary from 0 degrees to 360 degrees rather than being restricted to one quadrant. When the concept of the super-omni wedge is used, the optimization of beam weights, wedge angles, and wedge orientations for J beams is transformed into the optimization of beam weights for 4J beams. A quadratic dose-based objective function is defined, and the method of sequential quadratic programming is used to find the 4J beam weights that minimize it. After the weights of the nominal wedged beams have been determined, the beams can be delivered in one of four methods: Directly, by using the omni wedge technique, by using the universal wedge technique, and by using the virtual wedge technique. When tested with two clinical cases, the algorithm achieved homogeneous dose distributions in target volumes while meeting the constraints to the organs at risk. A prominent feature of the algorithm is that there is no need to manually preselect the orientations of nominal wedges.

The purpose of this work is to investigate algorithms for the delivery of intensity-modulated fields using independent collimators (IC). Two heuristic algorithms are proposed to calculate jaw-setting sequences for arbitrary 2D intensity distributions. The first algorithm is based on searching the whole intensity matrix to find the largest nonzero rectangular area as a segment while the second algorithm is to find a nonzero rectangular area as a segment which makes the complexity of the remaining intensity matrix minimum. After a sequence is obtained, the delivery order of all its segments is optimized with the technique of simulated annealing to minimize the total jaw-moving time. To evaluate these two algorithms, randomly generated intensity matrices and three clinical cases of different complexity have been tested, and the results have been compared with one algorithm proposed for MLC technique. It is shown that the efficiency of IC technique becomes increasingly lower than that of MLC technique, and the relative efficiency of two algorithms proposed here is related to machine dose rate and jaw speed. Assuming the prescribed dose is 200 cGY per fraction, machine dose rate is 250 MU/min, and jaw speed is 1.5 cm/s, the treatment can be delivered within about 20 min for all three cases with the first algorithm. The second algorithm requires longer delivery time under such assumptions. The delivery time can be further reduced through increasing machine dose rate and jaw speed, and developing more efficient algorithms. The use of IC for intensity-modulation radiotherapy has some potential advantages over other techniques.

Volumetric-modulated arc therapy (VMAT) is delivered through synchronized variation of gantry angle, dose rate, and multileaf collimator (MLC) leaf positions. The delivery dynamic nature challenges the parameter setting accuracy of linac control system. The purpose of this study was to develop a novel method for routine quality assurance (QA) of VMAT linacs.ArcCheck is a detector array with diodes distributing in spiral pattern on cylindrical surface. Utilizing its features, a QA plan was designed to strictly test all varying parameters during VMAT delivery on an Elekta Synergy linac. In this plan, there are 24 control points. The gantry rotates clockwise from 181° to 179°. The dose rate, gantry speed, and MLC positions cover their ranges commonly used in clinic. The two borders of MLC-shaped field seat over two columns of diodes of ArcCheck when the gantry rotates to the angle specified by each control point. The ratio of dose rate between each of these diodes and the diode closest to the field center is a certain value and sensitive to the MLC positioning error of the leaf crossing the diode. Consequently, the positioning error can be determined by the ratio with the help of a relationship curve. The time when the gantry reaches the angle specified by each control point can be acquired from the virtual inclinometer that is a feature of ArcCheck. The gantry speed between two consecutive control points is then calculated. The aforementioned dose rate is calculated from an acm file that is generated during ArcCheck measurements. This file stores the data measured by each detector in 50 ms updates with each update in a separate row. A computer program was written in MATLAB language to process the data. The program output included MLC positioning errors and the dose rate at each control point as well as the gantry speed between control points. To evaluate this method, this plan was delivered for four consecutive weeks. The actual dose rate and gantry speed were compared with the QA plan specified. Additionally, leaf positioning errors were intentionally introduced to investigate the sensitivity of this method.The relationship curves were established for detecting MLC positioning errors during VMAT delivery. For four consecutive weeks measured, 98.4%, 94.9%, 89.2%, and 91.0% of the leaf positioning errors were within ± 0.5 mm, respectively. For the intentionally introduced leaf positioning systematic errors of -0.5 and +1 mm, the detected leaf positioning errors of 20 Y1 leaf were -0.48 ± 0.14 and 1.02 ± 0.26 mm, respectively. The actual gantry speed and dose rate closely followed the values specified in the VMAT QA plan.This method can assess the accuracy of MLC positions and the dose rate at each control point as well as the gantry speed between control points at the same time. It is efficient and suitable for routine quality assurance of VMAT.

To evaluate the influence of carbon fiber couchtop and patient positioning devices on cone beam CT (CBCT) image quality and develop an effective method to correct the influence.A standard CT phantom (Catphan 500) was used to evaluate the influence of iBeam evo carbon fiber couchtop on the quality of CBCT image obtained from an Elekta synergy machine. The evaluation indices were contrast-to-noise ratio (CNR), spatial resolution, image uniformity, and image noise. With using the Beer-Lambert law and the energy-response function of the flat-panel imager, a method was applied to deduct the image signal of the couchtop (and the positioning devices) from each projection image of a phantom/patient, and then used all corrected projection images to reconstruct a CBCT image. The performance of the correction method was evaluated using the CBCT images of a Catphan 500 phantom, a head-and-neck cancer patient, and a prostate cancer patient.In two phantom studies (the phantom to simulate a human head and neck and the one to simulate a human body), the CNR of the CBCT images obtained with the couchtop reduced 18.1% and 29.8%, respectively with respect to those obtained without the couchtop; meanwhile, the image uniformity reduced 16.4% and 24.1% due to the use of the carbon fiber couchtop. The couchtop also induced extra image noise (16.5% for the hn phantom and 4.2% for the body phantom). However, CBCT imaging with the couchtop did not affect the spatial resolution. After applying the projection image correction, there was a significant improvement in CNR (by 19.5% and 25.8%), image uniformity (by 9.2% and 13.1%), and image noise (by 10.2% and 3.9%), with respect to CBCT images obtained with the couchtop.The presence of the carbon fiber couchtop and the patient positioning devices can significantly impair CBCT image quality in terms of the CNR, the image uniformity, and the image noise. By removing the influence of the couchtop and the patient-positioning devices from CB projection images, the correction method improves CBCT image quality and thus image guidance in radiotherapy.

This paper proposes a sequencing algorithm for intensity-modulated radiation therapy with a multileaf collimator in the static mode. The algorithm aims to minimize the number of segments in a delivery sequence. For a machine with a long verification and recording overhead time (e.g., 15 s per segment), minimizing the number of segments is equivalent to minimizing the delivery time. The proposed new algorithm is based on checking numerous candidates for a segment and selecting the candidate that results in a residual intensity matrix with the least complexity. When there is more than one candidate resulting in the same complexity, the candidate with the largest size is selected. The complexity of an intensity matrix is measured in the new algorithm in terms of the number of segments in the delivery sequence obtained by using a published algorithm. The beam delivery efficiency of the proposed algorithm and the influence of different published algorithms used to calculate the complexity of an intensity matrix were tested with clinical intensity-modulated beams. The results show that no matter which published algorithm is used to calculate the complexity of an intensity matrix, the sequence generated by the algorithm proposed here is always more efficient than that generated by the published algorithm itself. The results also show that the algorithm used to calculate the complexity of an intensity matrix affects the efficiency of beam delivery. The delivery sequences are frequently most efficient when the algorithm of Bortfeld et al. is used to calculate the complexity of an intensity matrix. Because no single variation is most efficient for all beams tested, we suggest implementing multiple variations of our algorithm.

This study proposes an algorithm for selecting beam weight, wedge angle, and wedge orientation for three-dimensional radiation therapy treatment planning. According to dose gradient analysis, the necessary and sufficient condition for achieving a homogeneous dose over the target volume is that the total vector sum of the dose gradients of all beams be zero everywhere in the target volume. This study presents equations for calculating the beam weight, wedge angle, and collimator angle (because the collimator angle determines wedge orientation when beam direction is known) for treatment plans using two angled beams or three coplanar or noncoplanar beams. It also provides suggestions for calculations of treatment plans using more than three beams, for which many feasible solutions will be available. When tested using two clinical cases, this algorithm achieved homogeneous dose distributions over target volumes. With this algorithm, repeated manual adjustments are reduced, and the quality and efficiency of treatment planning are improved.

Purpose: Volumetric-modulated arc therapy (VMAT) is delivered through synchronized variation of gantry angle, dose rate, and multileaf collimator (MLC) leaf positions. The delivery dynamic nature challenges the parameter setting accuracy of linac control system. The purpose of this study was to develop a novel method for routine quality assurance (QA) of VMAT linacs. Methods: ArcCheck is a detector array with diodes distributing in spiral pattern on cylindrical surface. Utilizing its features, a QA plan was designed to strictly test all varying parameters during VMAT delivery on an Elekta Synergy linac. In this plan, there are 24 control points. The gantry rotates clockwise from 181° to 179°. The dose rate, gantry speed, and MLC positions cover their ranges commonly used in clinic. The two borders of MLC-shaped field seat over two columns of diodes of ArcCheck when the gantry rotates to the angle specified by each control point. The ratio of dose rate between each of these diodes and the diode closest to the field center is a certain value and sensitive to the MLC positioning error of the leaf crossing the diode. Consequently, the positioning error can be determined by the ratio with the help of a relationship curve. The time when the gantry reaches the angle specified by each control point can be acquired from the virtual inclinometer that is a feature of ArcCheck. The gantry speed between two consecutive control points is then calculated. The aforementioned dose rate is calculated from an acm file that is generated during ArcCheck measurements. This file stores the data measured by each detector in 50 ms updates with each update in a separate row. A computer program was written in MATLAB language to process the data. The program output included MLC positioning errors and the dose rate at each control point as well as the gantry speed between control points. To evaluate this method, this plan was delivered for four consecutive weeks. The actual dose rate and gantry speed were compared with the QA plan specified. Additionally, leaf positioning errors were intentionally introduced to investigate the sensitivity of this method. Results: The relationship curves were established for detecting MLC positioning errors during VMAT delivery. For four consecutive weeks measured, 98.4%, 94.9%, 89.2%, and 91.0% of the leaf positioning errors were within ±0.5 mm, respectively. For the intentionally introduced leaf positioning systematic errors of −0.5 and +1 mm, the detected leaf positioning errors of 20 Y1 leaf were −0.48 ± 0.14 and 1.02 ± 0.26 mm, respectively. The actual gantry speed and dose rate closely followed the values specified in the VMAT QA plan. Conclusions: This method can assess the accuracy of MLC positions and the dose rate at each control point as well as the gantry speed between control points at the same time. It is efficient and suitable for routine quality assurance of VMAT.

Academy of Medical Sciences, Peking Union Medical College, Beijing 100021, China Purpose: To develop a backprojection dosimetric algorithm for dosimetric verification. The algorithm reconstructs a three dimensional dose distribution from the treatment beams' portal images and the CTimages for a patient/phantom. Method and Materials: The reconstruction process covers four steps: (a) To acquire a portal image with an electronic portal imaging device(EPID) and convert it into a transmission dose distribution on EPID plane. (b) To reconstruct the incident primary dose distribution from the transmission dose distribution. (c) To calculate the primary dose distribution in the phantom using the phantom's CTimage set. (d) To calculate the scatter dose distribution by superposing the scatter kernels in the patient/phantom; then to make the summation of the primary and the scatter dose distribution to get the total dose distribution in the patient/phantom. The dosimetric algorithm was implemented in a C program and applied to five phantoms, which were homogeneous, inhomogeneous, regular or irregular, irradiated by a regular‐shaped, irregular‐shaped or intensity‐modulated beam. The calculated dose distributions were compared with the measured ones. Results: For all the experiments, the agreement between the calculated and measured dose distribution was within 5% in the field areas with low dose gradients. Large deviation happened to the field edge in the lung, which had a low density. Conclusion: The accuracy of the developed backprojection dosimetric algorithm can meet the requirement of clinical dosimetric verification. But the algorithm should be improved furthering in order to calculate the dose in the region of electronic disequilibrium accurately.