Artificial intelligence and the medical physics profession - A Swedish perspective.

BACKGROUND: There is a continuous and dynamic discussion on artificial intelligence (AI) in present-day society. AI is expected to impact on healthcare processes and could contribute to a more sustainable use of resources allocated to healthcare in the future. The aim for this work was to establish a foundation for a Swedish perspective on the potential effect of AI on the medical physics profession. MATERIALS AND METHODS: We designed a survey to gauge viewpoints regarding AI in the Swedish medical physics community. Based on the survey results and present-day situation in Sweden, a SWOT analysis was performed on the implications of AI for the medical physics profession. RESULTS: Out of 411 survey recipients, 163 responded (40%). The Swedish medical physicists with a professional license believed (90%) that AI would change the practice of medical physics but did not foresee (81%) that AI would pose a risk to their practice and career. The respondents were largely positive to the inclusion of AI in educational programmes. According to self-assessment, the respondents' knowledge of and workplace preparedness for AI was generally low. CONCLUSIONS: From the survey and SWOT analysis we conclude that AI will change the medical physics profession and that there are opportunities for the profession associated with the adoption of AI in healthcare. To overcome the weakness of limited AI knowledge, potentially threatening the role of medical physicists, and build upon the strong position in Swedish healthcare, medical physics education and training should include learning objectives on AI.

Age-specific and gender-specific radiation risks in paediatric angiography and interventional cardiology: conversion coefficients and risk reference values.

OBJECTIVES: To estimate risk for exposure-induced cancer death (REID), organ-specific risks of exposure-induced cancer death (REIDHT) and associated conversion coefficients (CCREID:KAP=REID/kerma-area product (KAP), CCREIDHT:KAP=REIDHT/KAP) in paediatric cardiac catheterizations using data from radiation dose structured reports (RDSR). A novel risk surveillance tool consisting of age-specific and gender-specific risk reference values (RRVs) related to population cancer risk is suggested. METHODS: The PCXMC v.2.0 code is used together with exposure-related information from RDSR from a cohort of 238 children to assess cancer risks and related conversion coefficients. The KAP corresponding to 1 in 1000 of increased REID is used to define age-specific and gender-specific KAP values to monitor risk in such patient cohorts, here denoted as RRVs. RESULTS: The REID estimates ranged from below 1 up to 300 in 100,000, and the RRVs for the different age groups and gender ranged from 0.77 Gycm2 and 2.1 Gycm2 for neonates (female, male) to 11 Gycm2 and 25 Gycm2 for 15-year-olds (female, male). The CCREID:KAP and CCREIDHT:KAP decreased biexponentially with increased age, being notably higher for female patients. CONCLUSIONS: Prominent risk contributing organs were the lungs and the (female) breast. The concept of age-specific and gender-specific RRVs related to population cancer risk is introduced and is intended to be used as a supporting tool for physicians performing such interventions. ADVANCES IN KNOWLEDGE: Age-related and gender-related conversion coefficients for radiation risk, CCREID:KAP and CCREIDHT:KAP, are introduced and a novel risk surveillance concept, the RRV, is suggested for paediatric cardiac catheterizations.

Calculating organ and effective doses in paediatric interventional cardiac radiology based on DICOM structured reports - Is detailed examination data critical to dose estimates?

PURPOSE: To estimate effective dose (E), equivalent organ doses (HT) and associated conversion coefficients (CCE:KAP = E/KAP, CCHT:KAP = HT/KAP; KAP = Kerma-area product) in paediatric cardiac interventions, using detailed exposure data from radiation dose structured reports (RDSR). These "RDSR dose estimations" have been compared with estimations performed using the approach currently implemented in the clinic that is based on a simplified assumptions method (SAM). METHODS: The Monte Carlo system PCXMC, incorporated into a previously developed framework, was used to calculate E and HT for 202 children. The calculations were performed with input values from RDSR, and also using simplified assumptions, including fixed nominal values for the focus-skin distance, collimated beam size, irradiation geometry and patient size (age, weight and height). RESULTS: Mean HT to critical organs were: 5-25 mSv (lungs), 5-8 mSv (breasts) and 5-22 mSv (heart), with the lower and upper end of the doses associated with the neonatal and 15 years group, respectively. The associated mean CCHT:KAP for the different age groups were: 9.4-1.6 mSv/Gycm2 (lungs), 8.9-0.54 mSv/Gycm2 (breasts) and 9.3-1.4 mSv/Gycm2 (heart). CONCLUSIONS: The extension of the concept of a conversion coefficient for HT is introduced and CCHT:KAP values for paediatric cardiac interventions divided in age groups are presented. This method of linking the KAP to HT is intended for use in epidemiological/cohort studies or in clinics that do not have access to RDSR. Further, the population-averaged conversion coefficients for the critical organs estimated from RDSR, displayed no statistically significant difference compared with the SAM approach.

Assessment of the occupational eye lens dose for clinical staff in interventional radiology, cardiology and neuroradiology.

In accordance with recommendations by the International Commission on Radiological Protection, the current European Basic Safety Standards has adopted a reduced occupational eye lens dose limit of 20 mSv yr-1. The radiation safety implications of this dose limit is of concern for clinical staff that work with relatively high dose x-ray angiography and interventional radiology. Presented in this work is a thorough assessment of the occupational eye lens dose based on clinical measurements with active personal dosimeters worn by staff during various types of procedures in interventional radiology, cardiology and neuroradiology. Results are presented in terms of the estimated equivalent eye lens dose for various medical professions. In order to compare the risk of exceeding the regulatory annual eye lens dose limit for the widely different clinical situations investigated in this work, the different medical professions were separated into categories based on their distinct work pattern: staff that work (a) regularly beside the patient, (b) in proximity to the patient and (c) typically at a distance from the patient. The results demonstrate that the risk of exceeding the annual eye lens dose limit is of concern for staff category (a), i.e. mainly the primary radiologist/cardiologist. However, the results also demonstrate that the risk can be greatly mitigated if radiation protection shields are used in the clinical routine. The results presented in this work cover a wide range of clinical situations, and can be used as a first indication of the risk of exceeding the annual eye lens dose limit for staff at other medical centres.

Practical approaches to approximating MTF and NPS in CT with an example application to task-based observer studies.

PURPOSE: To investigate two methods of approximating the Modulation Transfer Function (MTF) and Noise Power Spectrum (NPS) in computed tomography (CT) for a range of scan parameters, from limited image acquisitions. METHODS: The two methods consist of 1) using a linear systems approach to approximate the NPS for different filtered backprojection (FBP) kernels with a filter function derived from the kernel ratio of determined MTFs and 2) using an empirical fitted model to approximate the MTF and NPS. In both cases a scaling function accounts for variations in mAs and kV. The two methods of approximating the MTF/NPS are further investigated by comparing image quality figure of merits (FOM) d' and AUC calculated using approximations of the MTF/NPS and MTF/NPS that have been determined for different mAs/kV levels and reconstruction kernels. RESULTS: The greatest RMSE for NPS approximated for a range of mAs/kVp/convolution kernels using both methods and compared to determined NPS was 0.05 of the peak value. The RMSE for FOM with the kernel ratio method were at most 0.1 for d' and 0.01 for the AUC. Using the empirical model method, the RMSE for FOM were at most 0.02 for d' and 0.001 for the AUC. CONCLUSIONS: The two methods proposed in this paper can provide a convenient way of approximating the MTF and NPS for use in, among other things, mathematical observer studies. Both methods require a relatively small number of direct determinations of NPS from scan acquisitions to model the NPS/MTF for arbitrary mAs and kV.

A framework for organ dose estimation in x-ray angiography and interventional radiology based on dose-related data in DICOM structured reports.

Although interventional x-ray angiography (XA) procedures involve relatively high radiation doses that can lead to deterministic tissue reactions in addition to stochastic effects, convenient and accurate estimation of absorbed organ doses has traditionally been out of reach. This has mainly been due to the absence of practical means to access dose-related data that describe the physical context of the numerous exposures during an XA procedure. The present work provides a comprehensive and general framework for the determination of absorbed organ dose, based on non-proprietary access to dose-related data by utilizing widely available DICOM radiation dose structured reports. The framework comprises a straightforward calculation workflow to determine the incident kerma and reconstruction of the geometrical relation between the projected x-ray beam and the patient's anatomy. The latter is difficult in practice, as the position of the patient on the table top is unknown. A novel patient-specific approach for reconstruction of the patient position on the table is presented. The proposed approach was evaluated for 150 patients by comparing the estimated position of the primary irradiated organs (the target organs) with their position in clinical DICOM images. The approach is shown to locate the target organ position with a mean (max) deviation of 1.3 (4.3), 1.8 (3.6) and 1.4 (2.9) cm for neurovascular, adult and paediatric cardiovascular procedures, respectively. To illustrate the utility of the framework for systematic and automated organ dose estimation in routine clinical practice, a prototype implementation of the framework with Monte Carlo simulations is included.

Quality control of CT systems by automated monitoring of key performance indicators: a two-year study.

The purpose of this study was to develop a method of performing routine periodical quality controls (QC) of CT systems by automatically analyzing key performance indicators (KPIs), obtainable from images of manufacturers' quality assurance (QA) phantoms. A KPI pertains to a measurable or determinable QC parameter that is influenced by other underlying fundamental QC parameters. The established KPIs are based on relationships between existing QC parameters used in the annual testing program of CT scanners at the Karolinska University Hospital in Stockholm, Sweden. The KPIs include positioning, image noise, uniformity, homogeneity, the CT number of water, and the CT number of air. An application (MonitorCT) was developed to automatically evaluate phantom images in terms of the established KPIs. The developed methodology has been used for two years in clinical routine, where CT technologists perform daily scans of the manufacturer's QA phantom and automatically send the images to MonitorCT for KPI evaluation. In the cases where results were out of tolerance, actions could be initiated in less than 10 min. 900 QC scans from two CT scanners have been collected and analyzed over the two-year period that MonitorCT has been active. Two types of errors have been registered in this period: a ring artifact was discovered with the image noise test, and a calibration error was detected multiple times with the CT number test. In both cases, results were outside the tolerances defined for MonitorCT, as well as by the vendor. Automated monitoring of KPIs is a powerful tool that can be used to supplement established QC methodologies. Medical physicists and other professionals concerned with the performance of a CT system will, using such methods, have access to comprehensive data on the current and historical (trend) status of the system such that swift actions can be taken in order to ensure the quality of the CT examinations, patient safety, and minimal disruption of service.

On the feasibility of utilizing active personal dosimeters worn on the chest to estimate occupational eye lens dose in x-ray angiography.

The International Commission on Radiological Protection (ICRP) has recommended that the occupational dose limit to the eye lens be substantially reduced. To ensure compliance with these recommendations, monitoring of the occupational eye lens dose is essential in certain hospital work environments. For assessment of the eye lens dose it is recommended to use a supplementary dosimeter placed at a position adjacent to the eye(s). Wearing a dosimeter at eye level can, however, be impractical and distributing and managing additional dosimeters over long periods of time is cumbersome and costly for large clinical sites. An attractive alternative is to utilize active personal dosimeters (APDs), which are routinely used by clinical staff for real-time monitoring of the personal dose equivalent rate (H(p)(10)). In this work, a formalism for the determination of eye lens dose from the response of such APD's worn on the chest is proposed and evaluated. The evaluation is based on both phantom and clinical measurements performed in an x-ray angiography suite for interventional cardiology. The main results show that the eye lens dose to the primary operator and to the assisting clinical staff can be conservatively estimated from the APD response as D(eye)(conductor) = 2.0 APD chest and D(eye)(assisting) = 1.0 APD chest, respectively. However, care should be exercised for particularly short assisting staff and if radiation protection shields are misused. These concerns can be greatly mitigated if the clinical staff are provided with adequate radiation protection training.

Influence of phantom thickness and material on the backscatter factors for diagnostic x-ray beam dosimetry.

Most of the existing backscatter factors for the dosimetry of clinical diagnostic x-ray beams have been calculated for 15 cm thick phantoms; these data are used for skin dose determinations which in general ignore the influence of phantom material and thickness. The former should strictly be required whenever dosimetry measurements are made on phantom materials different from those used for the backscatter factor calculations. The phantom or patient thickness is of special importance when skin dose determinations are made for infants or paediatric patients. In this work, the recently published formalism for reference dosimetry and comprehensive database of backscatter factors for clinical beams and water phantoms have been extended using two correction factors which account for phantom material and thickness. These were determined with simulations using the PENELOPE Monte Carlo system, for PMMA to analyse the influence of the phantom material relative to water, and for a broad range of thicknesses of water and PMMA to investigate the role of this parameter in patient dose estimates. The material correction factor was found to be in the range 3-10%, depending on the field size and the HVL. The thickness correction factor was in the range 2-12% for a 5 cm thick phantom and square field sizes between 5 and 35 cm, reaching a plateau of about ±1% for thicknesses beyond 13 cm. Expressions in the form of surface fits over the calculated data are provided which streamline the determination of backscatter factors for arbitrary thicknesses and phantom materials, as well as field sizes. Results demonstrate the inadequacy of using conventional backscatter factors (calculated for 15 cm thick phantoms) without correction factors that take into account the phantom material and its thickness.

Backscatter factors and mass energy-absorption coefficient ratios for diagnostic radiology dosimetry.

Backscatter factors, B, and mass energy-absorption coefficient ratios, (µ(en)/ρ)(w, air), for the determination of the surface dose in diagnostic radiology were calculated using Monte Carlo simulations. The main purpose was to extend the range of available data to qualities used in modern x-ray techniques, particularly for interventional radiology. A comprehensive database for mono-energetic photons between 4 and 150 keV and different field sizes was created for a 15 cm thick water phantom. Backscattered spectra were calculated with the PENELOPE Monte Carlo system, scoring track-length fluence differential in energy with negligible statistical uncertainty; using the Monte Carlo computed spectra, B factors and (µ(en)/ρ)(w, air) were then calculated numerically for each energy. Weighted averaging procedures were subsequently used to convolve incident clinical spectra with mono-energetic data. The method was benchmarked against full Monte Carlo calculations of incident clinical spectra obtaining differences within 0.3-0.6%. The technique used enables the calculation of B and (µ(en)/ρ)(w, air) for any incident spectrum without further time-consuming Monte Carlo simulations. The adequacy of the extended dosimetry data to a broader range of clinical qualities than those currently available, while keeping consistency with existing data, was confirmed through detailed comparisons. Mono-energetic and spectra-averaged values were compared with published data, including those in ICRU Report 74 and IAEA TRS-457, finding average differences of 0.6%. Results are provided in comprehensive tables appropriated for clinical use. Additional qualities can easily be calculated using a designed GUI interface in conjunction with software to generate incident photon spectra.

Polyp detection with MDCT: a phantom-based evaluation of the impact of dose and spatial resolution.

OBJECTIVE: The objective of our study was to evaluate the impact of dose and spatial resolution on the detection of colonic polyps using a 4-MDCT scanner. MATERIALS AND METHODS: Twenty-four latex phantoms that simulate the large bowel and contain artificial polyps of different sizes and shapes were constructed. The polyps were divided into three size groups (diameter, 0-2, 2-5, and 5-10 mm) and were classified into four shape groups: pedunculated; broad-based; ulcerated or depressed; and sessile or flat. The colon phantoms were submerged in a water tank and scanned on a 4-MDCT scanner using 12 protocols with various settings of slice thickness, pitch, and tube current. The images were independently evaluated by three radiologists using axial 2D multiplanar reconstruction images and a 3D surface-rendering technique (fly-through). RESULTS: At a constant dose (i.e., dose-length product [DLP]), the polyp detection rate increased with increasing axial spatial resolution. For the standard protocol (2.50-mm slice thickness, 1.5 pitch), the detection rate for all polyp sizes decreased from approximately 70% at 100 mA to 55% at 40 mA. Between a 60- and 100-mA tube current, the detection rate for the largest polyps (> 5 mm) was almost constant, close to 90%. CONCLUSION: The detection of polyps in the large bowel using a standard protocol can be improved without dose penalty by increasing the axial spatial resolution of the image acquisition and adjusting the tube current setting. If the analysis can be restricted to polyps larger than 5 mm, the dose can be substantially reduced without compromising the detection rate.

Effects of geometric distortion in 0.2T MRI on radiotherapy treatment planning of prostate cancer.

BACKGROUND AND PURPOSE: To evaluate the impact of two different methods of geometric distortion correction of MR images from a Siemens Magnetom Open Viva 0.2T resistive MR unit on the process of external beam radiotherapy treatment planning for prostate cancer. PATIENTS AND METHODS: A method for correction of system related and object induced distortions and one for correction of purely system related distortions have been evaluated. The latter used information extracted from MR images of a 3D phantom specifically designed for geometric distortion evaluation. An active shim procedure was performed prior to all phantom and patient scans. For each of five patients five standard treatment plans were compared using uncorrected and corrected MR images alone (density=water) and CT images alone. Finally internal anatomical landmarks were used for image registration between MR images (corrected and uncorrected) and CT images to evaluate the impact of distortion correction on the image registration process. RESULTS: Maximum distortions of 28 mm (mean 2.2 mm) were found within the FOV in frequency encode direction. Maximum distortions could be reduced by a factor of two (mean factor four) by our phantom measurement based technique. Distortion patterns were found to be stable and reproducible over several weeks with this MR unit. For 4/5 patients, relative doses at the normalization point as calculated on the distortion corrected MR images only (all tissues taken water equivalent) were all within 1% of the corresponding value from the standard CT-based plan (actual Hounsfield units). The largest differences in isocentric dose found in one case were 3.1% MR uncorrected vs. CT and 2.6% MR corrected vs. CT. Typical sites of internal anatomical landmarks chosen for image registration show distortions up to 3 mm. CONCLUSIONS: Object induced distortions are negligible at such low field strengths compared to system related distortions. Treatment plans for prostate cancer do not seem to differ significantly from "standard" plans calculated on CT images when calculated on distortion corrected MR images, even if all tissues are assigned the electron density of water. Distortion correction of MR images can theoretically improve the starting point for image registration of MR and CT images.

Interobserver comparison of CT and MRI-based prostate apex definition. Clinical relevance for conformal radiotherapy treatment planning.

BACKGROUND: CT is widely used for conformal radiotherapy treatment planning of prostate carcinoma. Its limitations are especially at the prostatic apex which cannot be separated from the urogenital diaphragm. The aim of this study was to compare the localization of the prostatic apex in CT and axial MRI to the sagittal MRI in an interobserver analysis. PATIENTS AND METHODS: 22 patients with pathologically proven prostatic carcinoma were included in the analysis. In all patients sagittal and axial T2-weighted MRI and conventional CT were performed. The position of the MRI and CT apices were localized independently by three observers in relation to the intertrochanteric line. Additional subjective judgment of the ability to define the apical border of the prostatic gland was performed by a five-scaled score. RESULTS: The apex of the prostate could be discriminated statistically significant (p < 0.001) better in the MRI as compared to CT with best judgement for the sagittal MRI. The interobserver variation for the definition of the prostatic apex was statistically significant (p = 0.009) smaller for the sagittal MRI compared to axial MRI and CT. On the average the apex as determined by sagittal MRI, axial MRI and CT was located 29 mm, 27 mm and 24 mm above the intertrochanteric line. The apex defined by CT would have led to an additional treatment of 6-13 mm in 10/22 patients compared to the sagittal MRI, defined by axial MRI only in five patients. CONCLUSION: Additional MRI provides a superior anatomic information especially in the apical portion of the prostate. It should be recommended for every single patient in the treatment planning process. It helps to avoid an unnecessary irradiation of healthy tissue and could lead to a decrease of anal side effects and radiation-induced impotency due to a reduction of the extent of irradiated penile structures.