Automated size-specific dose estimates using deep learning image processing

Polish Journal of Medical Physics and Engineering

Hidayanto

et al. 2021

Purpose: The aim of this work was to establish the relationships of patient size in terms of effective diameter (Deff) and water-equivalent diameter (Dw) with lateral (LAT) and anterior-posterior (AP) dimensions in order to predict the specific patient dose for thoracic, abdominal, and pelvic computed tomography (CT) examinations. Methods: A total of 47 thoracic images, 79 abdominal images, and 50 pelvic images were analyzed in this study. The patient’s images were retrospectively collected from Dr. Kariadi and Kensaras Hospitals, Semarang, Indonesia. The slices measured were taken from the middle of the scan range. The calculations of patient sizes (LAT, AP, Deff, and Dw) were automatically performed by IndoseCT 20b software. Deff and Dw were plotted as functions of LAT, AP, and AP+LAT. In addition, Dw was plotted as a function of Deff. Results: Strong correlations of Deff and Dw with LAT, AP, and AP+LAT were found. Stronger correlations were found in the Deff curves (R2 > 0.9) than in the Dw curves (R2 > 0.8). It was found that the average Deff was higher than the average Dw in the thoracic region, the average values were similar in the abdominal and pelvic regions. Conclusion: The current study extended the study of the relationships between Deff and Dw and the basic geometric diameter LAT, AP, and AP+LAT beyond those previously reported by AAPM. We evaluated the relationships for three regions, i.e. thoracic, abdominal, and pelvic regions. Based on our findings, it was possible to estimate Deff and Dw from only the LAT or AP dimension.

Section: Discussionmentioning

confidence: 99%

The effective and water-equivalent diameters as geometrical size functions for estimating CT dose in the thoracic, abdominal, and pelvic regions

Dewi

Polish Journal of Medical Physics and Engineering

Hidayanto

et al. 2021

“…Juszczyk et al. ( 21 ) proposed an automated segmentation approach to calculate D W using a convolutional neural network (CNN) and reported that the proposed method produces accurate results. However, they are retrospective techniques and do not allow for real-time assessment of D W and SSDE.…”

Section: Introductionmentioning

confidence: 99%

“…Gharbi et al (20) also successfully proposed an automated approach to measure D W based on the Fuzzy C-means classification and edge detection algorithms. Juszczyk et al (21) proposed an automated segmentation approach to calculate D W using a convolutional neural network (CNN) and reported that the proposed method produces accurate results. However, they are retrospective techniques and do not allow for real-time assessment of D W and SSDE.…”

Section: Introductionmentioning

confidence: 99%

Real-time fully automated dosimetric computation for CT images in the clinical workflow: A feasibility study

Porzio

2022

Front. Oncol.

BackgroundCurrently, the volume computed tomography dose index (CTDIvol), the most-used quantity to express the output dose of a computed tomography (CT) patient’s dose, is not related to the real size and attenuation properties of each patient. The size-specific dose estimates (SSDE), based on the water-equivalent diameter (DW) overcome those issues. The proposed methods found in the literature do not allow real-time computation of DW and SSDE.PurposeThis study aims to develop a software to compute DW and SSDE in a real-time clinical workflow.MethodIn total, 430 CT studies and scans of a water-filled funnel phantom were used to compute accuracy and evaluate the times required to compute the DW and SSDE. Two one-sided tests (TOST) equivalence test, Bland–Altman analysis, and bootstrap-based confidence interval estimations were used to evaluate the differences between actual diameter and DW computed automatically and between DW computed automatically and manually.ResultsThe mean difference between the DW computed automatically and the actual water diameter for each slice is −0.027% with a TOST confidence interval equal to [−0.087%, 0.033%]. Bland–Altman bias is −0.009% [−0.016%, −0.001%] with lower limits of agreement (LoA) equal to −0.0010 [−0.094%, −0.068%] and upper LoA equal to 0.064% [0.051%, 0.077%]. The mean difference between DW computed automatically and manually is −0.014% with a TOST confidence interval equal to [−0.056%, 0.028%] on phantom and 0.41% with a TOST confidence interval equal to [0.358%, 0.462%] on real patients. The mean time to process a single image is 13.99 ms [13.69 ms, 14.30 ms], and the mean time to process an entire study is 11.5 s [10.62 s, 12.63 s].ConclusionThe system shows that it is possible to have highly accurate DW and SSDE in almost real-time without affecting the clinical workflow of CT examinations.

“…20 An accurate automated approach to finding D w , in order to obtain accurate patient dose, is in great demand. [24][25][26][27][28][29][30] Anam et al 25 proposed a fully automated method to calculate D w in a phantom and in human anatomic regions using a region of interest (ROI) automatically fitted to the patient border. The automated calculation produced an excellent correlation to the manual calculation (R 2 = 0.999).…”

Section: Introductionmentioning

confidence: 99%

“…Gharbi et al 27 also successfully proposed an automated method to measure D w and SSDE based on the Fuzzy C‐means classification algorithm and edge detection. Recently, Juszczyk et al 24 proposed an automated segmentation of patient images to calculate D w and SSDE using a convolutional neural network (CNN) and reported accurate results. The algorithms were effectively designed to calculate D w based on segmentation of the largest object 28 .…”

Section: Introductionmentioning

confidence: 99%

An improved method for automated calculation of the water‐equivalent diameter for estimating size‐specific dose in CT

J Applied Clin Med Phys

Mahdani

Dewi

et al. 2021

The aim of this study is to propose an algorithm for the automated calculation of water-equivalent diameter (D w ) and size-specific dose estimation (SSDE) from clinical computed tomography (CT) images containing one or more substantial body part.Methods: All CT datasets were retrospectively acquired by the Toshiba Aquilion 128 CT scanner. The proposed algorithm consisted of a contouring stage for the D w calculation, carried out by taking the six largest objects in the cross-sectional image of the patient's body, followed by the removal of the CT table depending on the center position (y-axis) of each object. Validation of the proposed algorithm used images of patients who had undergone chest examination with both arms raised up, one arm placed down and both arms placed down, images of the pelvic region consisting of one substantial object, and images of the lower extremities consisting of two separated areas. Results:The proposed algorithm gave the same results for D w and SSDE as the previous algorithm when images consisted of one substantial body part. However, when images consisted of more than one substantial body part, the new algorithm was able to detect all parts of the patient within the image. The D w values from the proposed algorithm were 9.5%, 15.4%, and 39.6% greater than the previous algorithm for the chest region with one arm placed down, both arms placed down, and images with two legs, respectively. The SSDE values from the proposed algorithm were 8.2%, 11.2%, and 20.6% lower than the previous algorithm for the same images, respectively. Conclusions:We have presented an improved algorithm for automated calculation of D w and SSDE. The proposed algorithm is more general and gives accurate results for both D w and SSDE whether the CT images contain one or more than one substantial body part.How to cite this article: Anam C, Mahdani FR, Dewi WK, et al. An improved method for automated calculation of the water-equivalent diameter for estimating size-specific dose in CT.