Journal: Medical Physics

Article Title: Deep learning‐based upsampling of 2D detector array measurements for patient plan verification in radiotherapy

doi: 10.1002/mp.70358

Figure Lengend Snippet: Bilinear interpolation on a rectangular grid.

Article Snippet: The Varian data showed a passing rate of (58.8 ± 9.6)% for the bilinear interpolation and (81.0 ± 7.3)% for the neural network prediction.

Techniques:

Journal: Medical Physics

Article Title: Deep learning‐based upsampling of 2D detector array measurements for patient plan verification in radiotherapy

doi: 10.1002/mp.70358

Figure Lengend Snippet: Comparisons between the OD 1500 measurements, high‐resolution measurement obtained using bilinear interpolation, high‐resolution measurements predicted by trained neural network and high‐resolution measurement obtained using OD 1600 SRS for three IMRT segments (segment 1: subfigure (a–d), segment 2: subfigure (e–h), segment 3: subfigure (i–l)).

Article Snippet: The Varian data showed a passing rate of (58.8 ± 9.6)% for the bilinear interpolation and (81.0 ± 7.3)% for the neural network prediction.

Techniques:

Journal: Medical Physics

Article Title: Deep learning‐based upsampling of 2D detector array measurements for patient plan verification in radiotherapy

doi: 10.1002/mp.70358

Figure Lengend Snippet: The results of gamma index analysis for the three IMRT segments presented in Figure between the high‐resolution measurements obtained using bilinear interpolation (subfigures a, d and g) and neural network (subfigures b, e and h) by using the OD 1600 SRS measurements as reference. The failed points in red and blue color are marked in the figure. Subfigures c, f and I show a histogram for each segment truncated at Gamma = 4 to illustrate the magnitude in Gamma value of the failed points.

Article Snippet: The Varian data showed a passing rate of (58.8 ± 9.6)% for the bilinear interpolation and (81.0 ± 7.3)% for the neural network prediction.

Techniques:

Journal: Medical Physics

Article Title: Deep learning‐based upsampling of 2D detector array measurements for patient plan verification in radiotherapy

doi: 10.1002/mp.70358

Figure Lengend Snippet: Histogram of the gamma passing rates of bilinear interpolation and the neural network predictions for all investigated IMRT segments, (a) for the Elekta Synergy plans, (b) for the Varian trueBeam plans.

Article Snippet: The Varian data showed a passing rate of (58.8 ± 9.6)% for the bilinear interpolation and (81.0 ± 7.3)% for the neural network prediction.

Techniques:

Journal: Medical Physics

Article Title: Deep learning‐based upsampling of 2D detector array measurements for patient plan verification in radiotherapy

doi: 10.1002/mp.70358

Figure Lengend Snippet: 1D profiles of the field with 1 cm leaf gaps measured by the OD 1500, upsampled with bilinear interpolation (BL), predicted by neural network (NN), measured with OD 1600 SRS and EBT3 films. Additionally, a plot for the difference between film measurement and BL as well as NN, ∆D in Gy, has been included.

Article Snippet: The Varian data showed a passing rate of (58.8 ± 9.6)% for the bilinear interpolation and (81.0 ± 7.3)% for the neural network prediction.

Techniques:

Journal: Medical Physics

Article Title: Deep learning‐based upsampling of 2D detector array measurements for patient plan verification in radiotherapy

doi: 10.1002/mp.70358

Figure Lengend Snippet: Example of a cross section through a cylindric VMAT dose volume. Left: Failed points map of the standard bilinear interpolation when comparing to the TPS calculated dose distribution. Right: Comparison of the neural network output and the TPS calculated dose distribution.

Article Snippet: The Varian data showed a passing rate of (58.8 ± 9.6)% for the bilinear interpolation and (81.0 ± 7.3)% for the neural network prediction.

Techniques: Comparison

Journal: Medical Physics

Article Title: Deep learning‐based upsampling of 2D detector array measurements for patient plan verification in radiotherapy

doi: 10.1002/mp.70358

Figure Lengend Snippet: Line profiles through the dose plane shown in Figure . The green line in Figure indicates the position of the line plot. The dose profile shows the standard bilinear interpolation (blue) and neural network interpolation (red) compared to the TPS calculated dose profile (black).

Article Snippet: The Varian data showed a passing rate of (58.8 ± 9.6)% for the bilinear interpolation and (81.0 ± 7.3)% for the neural network prediction.

Techniques: