A new technique to characterize CT scanner bow-tie filter attenuation and applications in human cadaver dosimetry simulations

A method of rapid quantification of patient-specific organ doses for CT using deep-learning-based multi-organ segmentation and GPU-accelerated Monte Carlo dose computing

PURPOSE

One technical barrier to patient-specific computed tomography (CT) dosimetry has been the lack of computational tools for the automatic patient-specific multi-organ segmentation of CT images and rapid organ dose quantification. When previous CT images are available for the same body region of the patient, the ability to obtain patient-specific organ doses for CT - in a similar manner as radiation therapy treatment planning - will open the door to personalized and prospective CT scan protocols. This study aims to demonstrate the feasibility of combining deep-learning algorithms for automatic segmentation of multiple radiosensitive organs from CT images with the GPU-based Monte Carlo rapid organ dose calculation.

METHODS

A deep convolutional neural network (CNN) based on the U-Net for organ segmentation is developed and trained to automatically delineate multiple radiosensitive organs from CT images. Two databases are used: The lung CT segmentation challenge 2017 (LCTSC) dataset that contains 60 thoracic CT scan patients, each consisting of five segmented organs, and the Pancreas-CT (PCT) dataset, which contains 43 abdominal CT scan patients each consisting of eight segmented organs. A fivefold cross-validation method is performed on both sets of data. Dice similarity coefficients (DSCs) are used to evaluate the segmentation performance against the ground truth. A GPU-based Monte Carlo dose code, ARCHER, is used to calculate patient-specific CT organ doses. The proposed method is evaluated in terms of relative dose errors (RDEs). To demonstrate the potential improvement of the new method, organ dose results are compared against those obtained for population-average patient phantoms used in an off-line dose reporting software, VirtualDose, at Massachusetts General Hospital.

RESULTS

The median DSCs are found to be 0.97 (right lung), 0.96 (left lung), 0.92 (heart), 0.86 (spinal cord), 0.76 (esophagus) for the LCTSC dataset, along with 0.96 (spleen), 0.96 (liver), 0.95 (left kidney), 0.90 (stomach), 0.87 (gall bladder), 0.80 (pancreas), 0.75 (esophagus), and 0.61 (duodenum) for the PCT dataset. Comparing with organ dose results from population-averaged phantoms, the new patient-specific method achieved smaller absolute RDEs (mean ± standard deviation) for all organs: 1.8% ± 1.4% (vs 16.0% ± 11.8%) for the lung, 0.8% ± 0.7% (vs 34.0% ± 31.1%) for the heart, 1.6% ± 1.7% (vs 45.7% ± 29.3%) for the esophagus, 0.6% ± 1.2% (vs 15.8% ± 12.7%) for the spleen, 1.2% ± 1.0% (vs 18.1% ± 15.7%) for the pancreas, 0.9% ± 0.6% (vs 20.0% ± 15.2%) for the left kidney, 1.7% ± 3.1% (vs 19.1% ± 9.8%) for the gallbladder, 0.3% ± 0.3% (vs 24.2% ± 18.7%) for the liver, and 1.6% ± 1.7% (vs 19.3% ± 13.6%) for the stomach. The trained automatic segmentation tool takes <5 s per patient for all 103 patients in the dataset. The Monte Carlo radiation dose calculations performed in parallel to the segmentation process using the GPU-accelerated ARCHER code take <4 s per patient to achieve <0.5% statistical uncertainty in all organ doses for all 103 patients in the database.

CONCLUSION

This work shows the feasibility to perform combined automatic patient-specific multi-organ segmentation of CT images and rapid GPU-based Monte Carlo dose quantification with clinically acceptable accuracy and efficiency.

SHAPE ESTIMATION OF BOWTIE FILTERS BASED ON THE LUMINESCENCE FROM POLYETHYLENE TEREPHTHALATE RESIN BY X-RAY IRRADIATION

In this study, we devised a novel method estimating the bowtie filter shapes by imaging luminescence from a polyethylene terephthalate (PET) resin with X-ray irradiation in a computed tomography (CT) scanner. The luminescence distribution of the PET resin corresponding to the thickness of bowtie filter was imaged using a charge-coupled device camera. On the assumption that the material of bowtie filter is aluminium (Al), the shape of bowtie filters was estimated from the correlation between Al attenuation curves and the angular-dependent luminance attenuation profiles according to the thickness of bowtie filters. Dose simulations based on the estimated bowtie filter shapes were performed using head and body PMMA phantoms with 16 and 32 cm in diameter. The simulated values of head and body weighted CT dose index (CTDIw) based on bowtie filter shape by the luminescence imaging method agreed within ~9% with the measured values by a dosemeter.

Initial clinical experience with high-pitch dual-source CT as a rapid technique for thoraco-abdominal evaluation in awake infants and young children

AIM

To evaluate dual-source high-pitch computed tomography (HPCT) imaging of the chest and abdomen as a rapid scanning technique to obtain diagnostic-quality imaging evaluation of infants and young children without sedation.

MATERIALS AND METHODS

Fifty-three paediatric patients (age 24.1±2 months) who underwent chest or abdomen HPCT (≥1.5) and standard pitch CT (SPCT, <1.5) on a dual-source 128-row multidetector CT system were included in the study. Image quality assessment was performed by two paediatric radiologists for diagnostic confidence, image artefacts, and image noise. Objective image noise was measured.

RESULTS

Most of the CT examinations were performed in children who were >1 year old (n=15 and n=20) followed by ≤1 year old (n=8 and n=10) in SPCT and HPCT, respectively. The mean radiation dose (SSDE) from HPCT was 1.96±1 mGy compared to 2.2±1 mGy for SPCT (p=0.3). No major artefacts were reported and overall image quality of all HPCT examinations was acceptable diagnostically. In addition, objective image noise values were not significantly different between HPCT compared with SPCT (11±3 versus 11±5, p=0.7).

CONCLUSION

Ultra-fast, HPCT can be performed without the need for sedation as a potential alternative to anaesthetised magnetic resonance imaging in infants and young children.

A SIMPLE METHOD FOR CALIBRATING PIXEL VALUES OF THE CT LOCALIZER RADIOGRAPH FOR CALCULATING WATER-EQUIVALENT DIAMETER AND SIZE-SPECIFIC DOSE ESTIMATE

The purpose of this study is to establish the relationship between the pixel value (I) of the CT localizer radiograph and water-equivalent thickness (tw) in a straightforward procedure. We used a body CTDI phantom, which was scanned in the AP and LAT projections. After transformation from the pixel values of the images to tw, water-equivalent diameter (Dw) and size-specific dose estimate were calculated on an anthropomorphic phantom and 30 patients retrospectively. We found a linear correlation between I and tw, with R2 ≥ 0.980. The Dw values based on the CT localizer radiograph were comparable to those calculated using axial images. The Dw difference for the anthropomorphic phantom between AP projection and axial images was 5.4 ± 4.2%, and between LAT projection and axial images was 6.7 ± 5.3%. The Dw differences for the patients between CT localizer radiograph and axial images was 2.3 ± 3.2%.

Determination and verification of the x-ray spectrum of a CT scanner

The accuracy of Monte Carlo (MC) simulations in estimating the computed tomography radiation dose is highly dependent on the proprietary x-ray source information. To address this, this study develops a method to precisely estimate the x-ray spectrum and bowtie (BT) filter thickness of the x-ray source based on physical measurements and calculations. The static x-ray source of the CT localizer radiograph was assessed to measure the total filtration at the isocenter for the x-ray spectrum characterization and the BT profile (air-kerma values as a function of fan angle). With these values, the utilized BT filter in the localizer radiograph was assessed by integrating the measured air kerma in a full 360-deg cycle. The consistency observed between the integrated BT filter profiles and the directly measured profiles pointed to the similarity in the utilized BT filter in terms of thickness and material between the static and rotating x-ray geometries. Subsequently, the measured air kerma was used to calculate the BT filter thickness and was verified using MC simulations by comparing the calculated and measured air-kerma values, where a very good agreement was observed. This would allow a more accurate computed tomography simulation and facilitate the estimation of the dose delivered to the patients.

Generation and analysis of clinically relevant breast imaging x-ray spectra

PURPOSE

The purpose of this work was to develop and make available x-ray spectra for some of the most widely used digital mammography (DM), breast tomosynthesis (BT), and breast CT (bCT) systems in North America.

METHODS

The Monte Carlo code MCNP6 was used to simulate minimally filtered (only beryllium) x-ray spectra at 8 tube potentials from 20 to 49 kV for DM/BT, and 9 tube potentials from 35 to 70 kV for bCT. Vendor-specific anode compositions, effective anode angles, focal spot sizes, source-to-detector distances, and beryllium filtration were simulated. For each 0.5 keV energy bin in all simulated spectra, the fluence was interpolated using cubic splines across the range of simulated tube potentials to produce spectra in 1 kV increments from 20 to 49 kV for DM/BT and from 35 to 70 kV for bCT. The HVL of simulated spectra with conventional filtration (at 35 kV for DM/BT and 49 kV for bCT) was used to assess spectral differences resulting from variations in: (a) focal spot size (0.1 and 0.3 mm IEC), (b) solid angle at the detector (i.e., small and large FOV size), and (c) geometrical specifications for vendors that employ the same anode composition.

RESULTS

Averaged across all DM/BT vendors, variations in focal spot and FOV size resulted in HVL differences of 2.2% and 0.9%, respectively. Comparing anode compositions separately, the HVL differences for Mo (GE, Siemens) and W (Hologic, Philips, and Siemens) spectra were 0.3% and 0.6%, respectively. Both the commercial Koning and prototype "Doheny" (UC Davis) bCT systems utilize W anodes with a 0.3 mm focal spot. Averaged across both bCT systems, variations in FOV size resulted in a 2.2% difference in HVL. In addition, the Koning spectrum was slightly harder than Doheny with a 4.2% difference in HVL. Therefore to reduce redundancy, a generic DM/BT system and a generic bCT system were used to generate the new spectra reported herein. The spectral models for application to DM/BT were dubbed the Molybdenum, Rhodium, and Tungsten Anode Spectral Models using Interpolating Cubic Splines (MASMICS_M-T , RASMICS_M-T , and TASMICS_M-T ; subscript "M-T" indicating mammography and tomosynthesis). When compared against reference models (MASMIP_M , RASMIP_M , and TASMIP_M ; subscript "M" indicating mammography), the new spectral models were in close agreement with mean differences of 1.3%, -1.3%, and -3.3%, respectively, across tube potential comparisons of 20, 30, and 40 kV with conventional filtration. TASMICS_bCT -generated bCT spectra were also in close agreement with the reference TASMIP model with a mean difference of -0.8%, across tube potential comparisons of 35, 49, and 70 kV with 1.5 mm Al filtration.

CONCLUSIONS

The Mo, Rh, and W anode spectra for application in DM and BT (MASMICS_M-T , RASMICS_M-T , and TASMICS_M-T ) and the W anode spectra for bCT (TASMICS_bCT ) as described in this study should be useful for individuals interested in modeling the performance of modern breast x-ray imaging systems including dual-energy mammography which extends to 49 kV. These new spectra are tabulated in spreadsheet form and are made available to any interested party.

Direct and fast measurement of CT beam filter profiles with simultaneous geometrical calibration

PURPOSE

To accurately measure the beam filter profiles from a variety of CT scanner models and to provide reference data for Monte Carlo simulations of CT scanners.

METHODS

This study proposed a new method to measure CT beam filter profiles using a linear-array x-ray detector (X-Scan 0.8f3-512; Detection Technology Inc., Espoo, Finland) under gantry rotation mode. A robust geometrical calibration approach was developed to determine key geometrical parameters by considering the x-ray focal spot location relative to the linear-array detector and the gantry's angular increment at each acquisition point. CT beam intensity profiles were synthesized from continuously measured data during a 10° gantry rotation range with calibrated detector response and system geometry information. Relative transmission profiles of nineteen sets of beam filters were then derived for nine different CT scanner models from three different manufacturers. Equivalent aluminum thickness profiles of these beam filters were determined by analytical calculation using the Spektr Matlab software package to match the measured transmission profiles. Three experiments were performed to validate the accuracy of the geometrical calibration, detector response modeling, and the derived equivalent aluminum thickness profiles.

RESULTS

The beam intensity profiles measured from gantry rotation mode showed very good agreement with those measured with gantry stationary mode, with a maximal difference of 3%. The equivalent aluminum thickness determined by this proposed method agreed well with what was measured by an ion chamber, with a mean difference of 0.4%. The determined HVL profiles matched well with data from a previous study (max difference of 4.7%).

CONCLUSIONS

An accurate and robust method to directly measure profiles from a broad list of beam filters and CT scanner models was developed, implemented, and validated. Useful reference data was provided for future research on CT system modeling.

A measurement-based generalized source model for Monte Carlo dose simulations of CT scans

The goal of this study is to develop a generalized source model for accurate Monte Carlo dose simulations of CT scans based solely on the measurement data without a priori knowledge of scanner specifications. The proposed generalized source model consists of an extended circular source located at x-ray target level with its energy spectrum, source distribution and fluence distribution derived from a set of measurement data conveniently available in the clinic. Specifically, the central axis percent depth dose (PDD) curves measured in water and the cone output factors measured in air were used to derive the energy spectrum and the source distribution respectively with a Levenberg-Marquardt algorithm. The in-air film measurement of fan-beam dose profiles at fixed gantry was back-projected to generate the fluence distribution of the source model. A benchmarked Monte Carlo user code was used to simulate the dose distributions in water with the developed source model as beam input. The feasibility and accuracy of the proposed source model was tested on a GE LightSpeed and a Philips Brilliance Big Bore multi-detector CT (MDCT) scanners available in our clinic. In general, the Monte Carlo simulations of the PDDs in water and dose profiles along lateral and longitudinal directions agreed with the measurements within 4%/1 mm for both CT scanners. The absolute dose comparison using two CTDI phantoms (16 cm and 32 cm in diameters) indicated a better than 5% agreement between the Monte Carlo-simulated and the ion chamber-measured doses at a variety of locations for the two scanners. Overall, this study demonstrated that a generalized source model can be constructed based only on a set of measurement data and used for accurate Monte Carlo dose simulations of patients' CT scans, which would facilitate patient-specific CT organ dose estimation and cancer risk management in the diagnostic and therapeutic radiology.

Mathematical modelling of scanner-specific bowtie filters for Monte Carlo CT dosimetry

The purpose of bowtie filters in CT scanners is to homogenize the x-ray intensity measured by the detectors in order to improve the image quality and at the same time to reduce the dose to the patient because of the preferential filtering near the periphery of the fan beam. For CT dosimetry, especially for Monte Carlo calculations of organ and tissue absorbed doses to patients, it is important to take the effect of bowtie filters into account. However, material composition and dimensions of these filters are proprietary. Consequently, a method for bowtie filter simulation independent of access to proprietary data and/or to a specific scanner would be of interest to many researchers involved in CT dosimetry. This study presents such a method based on the weighted computer tomography dose index, CTDI_w, defined in two cylindrical PMMA phantoms of 16 cm and 32 cm diameter. With an EGSnrc-based Monte Carlo (MC) code, ratios CTDI_w/CTDI_100,a were calculated for a specific CT scanner using PMMA bowtie filter models based on sigmoid Boltzmann functions combined with a scanner filter factor (SFF) which is modified during calculations until the calculated MC CTDI_w/CTDI_100,a matches ratios CTDI_w/CTDI_100,a, determined by measurements or found in publications for that specific scanner. Once the scanner-specific value for an SFF has been found, the bowtie filter algorithm can be used in any MC code to perform CT dosimetry for that specific scanner. The bowtie filter model proposed here was validated for CTDI_w/CTDI_100,a considering 11 different CT scanners and for CTDI_100,c, CTDI_100,p and their ratio considering 4 different CT scanners. Additionally, comparisons were made for lateral dose profiles free in air and using computational anthropomorphic phantoms. CTDI_w/CTDI_100,a determined with this new method agreed on average within 0.89% (max. 3.4%) and 1.64% (max. 4.5%) with corresponding data published by CTDosimetry (www.impactscan.org) for the CTDI HEAD and BODY phantoms, respectively. Comparison with results calculated using proprietary data for the PHILIPS Brilliance 64 scanner showed agreement on average within 2.5% (max. 5.8%) and with data measured for that scanner within 2.1% (max. 3.7%). Ratios of CTDI_100,c/CTDI₁₀₀, _p for this study and corresponding data published by CTDosimetry (www.impactscan.org) agree on average within about 11% (max. 28.6%). Lateral dose profiles calculated with the proposed bowtie filter and with proprietary data agreed within 2% (max. 5.9%), and both calculated data agreed within 5.4% (max. 11.2%) with measured results. Application of the proposed bowtie filter and of the exactly modelled filter to human phantom Monte Carlo calculations show agreement on the average within less than 5% (max. 7.9%) for organ and tissue absorbed doses.