ORIGINAL RESEARCH • HEALTH POLICY AND PRACTICE
The Energy Consumption of Radiology: Energy- and  
Cost-saving Opportunities for CT and MRI Operation
Tobias Heye, MD • Roland Knoerl, MBA, B Eng • Thomas Wehrle, Dipl-Ing • Daniel Mangold • Alessandro Cerminara •  
Michael Loser, PhD • Martin Plumeyer, Dipl-Ing • Markus Degen, PhD • Rahel Lüthy, MSc •  
Dominique Brodbeck, PhD • Elmar Merkle, MD
From the Department of Radiology, University Hospital Basel, Petersgraben 4, Basel 4031, Switzerland (T.H., E.M.); Siemens Healthineers, Forchheim, Germany (R.K., 
M.L., M.P.); Building Management, University Hospital Basel, Basel, Switzerland (T.W., D.M., A.C.); and School of Life Sciences, University of Applied Sciences and Arts 
Northwestern (FHWN), Muttenz, Switzerland (M.D., R.L., D.B.). Received September 16, 2019; revision requested November 4; revision received January 20, 2020; 
accepted January 23. Address correspondence to T.H. (e-mail: tobias.heye@usb.ch).
Supported by Siemens Healthineers, Forchheim, Germany (individual project agreement #5).
Conflicts of interest are listed at the end of this article.
Radiology 2020; 295:593–605  • https://doi.org/10.1148/radiol.2020192084 • Content code: 
Background: Awareness of energy efficiency has been rising in the industrial and residential sectors but only recently in the health 
care sector.
Purpose: To measure the energy consumption of modern CT and MRI scanners in a university hospital radiology department and 
to estimate energy- and cost-saving potential during clinical operation.
Materials and Methods: Three CT scanners, four MRI scanners, and cooling systems were equipped with kilowatt-hour energy mea-
surement sensors (2-Hz sampling rate). Energy measurements, the scanners’ log files, and the radiology information system from 
the entire year 2015 were analyzed and segmented into scan modes, as follows: net scan (actual imaging), active (room time), idle, and 
system-on and system-off states (no standby mode was available). Per-examination and peak energy consumption were calculated.
Results: The aggregated energy consumption imaging 40 276 patients amounted to 614 825 kWh, dedicated cooling systems to 
492 624 kWh, representing 44.5% of the combined consumption of 1 107 450 kWh (at a cost of U.S. $199 341). This is equivalent 
to the usage in a town of 852 people and constituted 4.0% of the total yearly energy consumption at the authors' hospital. Mean 
consumption per CT examination over 1 year was 1.2 kWh, with a mean energy cost (6standard deviation) of $0.22 6 0.13. The 
total energy consumption of one CT scanner for 1 year was 26 226 kWh ($4721 in energy cost). The net consumption per CT 
examination over 1 year was 3580 kWh, which is comparable to the usage of a two-person household in Switzerland; however, idle 
state consumption was fourfold that of net consumption (14 289 kWh). Mean MRI consumption over 1 year was 19.9 kWh per 
examination, with a mean energy cost of $3.57 6 0.96. The mean consumption for a year in the system-on state was 82 174 kWh 
per MRI examination and 134 037 kWh for total consumption, for an energy cost of $24 127.
Conclusion: CT and MRI energy consumption is substantial. Considerable energy- and cost-saving potential is present during non-
productive idle and system-off modes, and this realization could decrease total cost of ownership while increasing energy efficiency.
© RSNA, 2020
Awareness of energy consumption of electrical or elec- Although energy-saving capabilities in the residential tronic personal devices and household or office equip- and office sector have been addressed and appear to have 
ment has increased during the past decades. Standby, sleep, been improved, the health care sector is largely devoid 
and power save modes were introduced and are now com- of comparable initiatives aimed at considerably optimiz-
monplace among various consumer devices, from personal ing energy consumption efficiency (2), albeit aside from 
computers, notebooks, and smart phones to washing ma- the self-regulatory initiative of the European Coordina-
chines and televisions. The European Union, for instance, tion Committee of the Radiological, Electromedical and 
has established a detailed directive for the setting of ecode- Healthcare IT Industry, which focused on participating in 
sign requirements of energy-using products (European ecodesign in the medical device industry (3). Radiology 
Commission Regulation no. 1275/2008). According to a departments are major energy consumers within a hospital 
study performed during the preparation of this directive, through operation of CT and MRI scanners, which require 
annual electricity consumption in the European commu- energy in the range of 0.5–30 kWh per examination, with 
nity related to standby functions and off-mode losses was peak consumption reaching beyond 100 kW for a short 
estimated to be 47 TWh in 2005, which equals 19 mega- time period.
tons of CO2 emissions, and was predicted to increase to 49 To gauge the impact of radiologic procedures on the 
TWh in 2020 (1). This illustrates that power consump- overall power consumption of a hospital and to evaluate 
tion of electrical and electronic devices during nonusage or the theoretical power-saving potential in the clinical opera-
idling is substantial on a macroscopic level, with economic tion of radiologic cross-sectional imaging modalities, the 
and ecologic impact. aim of this study was to perform a detailed assessment of 
This copy is for personal use only. To order printed copies, contact reprints@rsna.org
Energy- and Cost-saving Opportunities for CT and MRI Operation
had a gradient of 625A/2000 V, a radiofrequency amplifier 
Abbreviation peak root mean square power of 22.5 kW, and was installed 
RIS = radiology information system in 2004; MRI unit 2a (Magnetom Espree, software version 
Summary VB17A) had a gradient of 625A/2000 V, a radiofrequency 
CT and MRI scanners are energy intensive in their operation; con- amplifier peak root mean square power of 22.5 kW, and was 
siderable energy- and cost-saving potential is present during idle and installed in 2004; and MRI unit 2b (Magnetom Avanto FIT, 
system-off states. software version VB17A) had a gradient of 625A/2000 V, a 
Key Results radiofrequency amplifier peak root mean square power of 22.5 
The energy consumption of CT and MRI scanners was compa- kW, and was installed in 2015. There were two 3.0-T units: n 
rable to the energy requirements of a town of 852 people living in Magnetom Verio (software version VB17A), which had a gra-
four-person households, or 4% of our total yearly hospital energy dient of 900A/2250 V, a radiofrequency amplifier peak root 
consumption. mean square power of 37.5 kW, and was installed in 2008; 
n For CT, two-thirds of energy consumption took place during the and Magnetom Skyra (software version VD13A), which had 
nonproductive idle system state. a gradient of 750A/2250 V, a radiofrequency amplifier peak 
n For MRI, one-third of energy consumption was attributed to the 
system-off state owing to the need for constant helium cooling and root mean square power of 37.5 kW, and was installed in 2013. 
cooling head operation. One 1.5-T MRI scanner (MRI unit 2a) was replaced in 2015, 
resulting in 7 months of data acquisition with the old scanner 
and 3 months with the new scanner (MRI unit 2b), leaving a 
the energy consumption of CT and MRI scanners and their as- 2-month gap without data. In these instances, we extrapolated 
sociated cooling systems within a university hospital radiology the available data to 12 months for per-year energy consump-
department over a 1-year period. tion summary values. The routine MRI scanners were oper-
ated during weekdays from 7 am to 8 pm; on Saturdays, two 
Materials and Methods MRI scanners (1.5-T and 3.0-T units) were operated from 8 
This study received financial support from Siemens Health- am to 5 pm. On Sundays, MRI was performed for acute and 
ineers (Forchheim, Germany), which was used to equip modal- emergency indications. Data from only one of the 3.0-T MRI 
ities with sensors and to fund custom software development. scanners (3.0-T MRI Magnetom Verio) were included in this 
The authors who were not employees of Siemens Healthineers study. If data were needed for summary consumption analysis, 
had full control of any data and information that might present we duplicated the data from the available scanner. The routine 
a potential conflict of interest for authors who were employees. CT and MRI scanners were shut down outside of operating 
For this study, we combined several data sources—the local hours, representing a system-off state; during clinical operation 
building energy consumption measurement system, scanner log the scanners were in the system-on state, which consisted of net 
data, and the radiology information system (RIS)—to allow for scan, active, and idle system event states (Table 1). There was 
a detailed and stratified data analysis. The time period of the no standby mode for the scanners, which would have allowed 
entire year 2015 was chosen, and the corresponding data were for lower energy consumption.
retrieved from each respective source system. We set the study We equipped all MRI and CT scanners with energy con-
up to investigate the following aspects: (a) CT and MRI scanner sumption measurement sensors that provided a 2-Hz data sam-
energy consumption per type of examination, day, month, and pling rate for kilowatt-hour consumption for each scanner. The 
year; (b) energy consumption distribution for different scanner energy consumption sensors were connected with and accessible 
activity states; (c) peak energy consumption and accumulated by means of a central building information system (APROL 
peak energy consumption, if scanner activities coincided; and (d) 4.0; B&R, Frauenfeld, Switzerland), which enabled data export 
energy consumption of associated cooling systems required for on the basis of filter criteria. The measured data were stored 
CT and MRI operation. by means of a smart recording algorithm in which data entries 
were saved only when they differed from the previous entry. 
Energy Consumption Measurement of CT and MRI Scanners This optimized data storage usage. As a preparatory step for the 
There were three CT scanners, two in the radiology depart- analysis, the software reconstructed continuous data points with 
ment (CT1: dual-source Somatom Definition Flash [installed 500-msec intervals (2-Hz sampling rate). Data regarding energy 
in 2011] and CT2: single-source Somatom Definition Edge consumption of the associated cooling systems, which operated 
[installed in 2014], both from Siemens Healthineers) and one exclusively for MRI and CT scanners, were available by means of 
CT scanner in the emergency room (ER-CT: single-source the central building information system.
Somatom Definition AS+ [Siemens Healthineers], installed 
in 2011). All were 128-slice–detector scanners, and all used System Log Data for CT and MRI Scanners
software version VA48A. The radiology department scanners We retrieved the system log files of each scanner and processed 
were in operation from 7:30 am to 5 pm, and the emergency their data to identify events that reflected scanner activity states 
room scanner was in operation 24 hours a day. The radiology and examination-related information. There were abundant 
department was also equipped with four clinical MRI units data entries with millisecond temporal resolution time stamps 
(all from Siemens Healthineers). There were three 1.5-T units: in the scanner log files, representing scanner operation details 
MRI unit 1 (Magnetom Avanto, software version VB17A) including but not limited to gantry operation, patient table 
594 radiology.rsna.org  n  Radiology: Volume 295: Number 3—June 2020
Heye et al
Table 1: Definition of Scanner Activity System States
System State Definition Data Curation
Net scan The actual scan event during which energy consumption deviates  Primarily derived from energy consumption data by 
from the baseline; the productive phase in which images are  identifying peak consumption or deviation from 
acquired baseline consumption, respectively
Active The time period during which a scanner is used to examine  Derived from energy consumption data in  
a patient, which includes preparation for and planning of scan,  conjunction with scanner log file information
the actual net scan, and reconstruction of raw data into image  
data. Basically, patient room time
Idle The time interval between “active” time periods within system-on  Parameter calculated by subtracting “active” from 
time period; it is merely a defined time interval within system-on “system-on” data
state and not a scanner state that can be manually activated
System on A scanner system state derived from the log file; system is powered  Energy consumption data segmented according to 
on and immediate scanning is possible; net scan, active, and idle  scanner log file information
are system-state events occurring during system-on state
System off A scanner system state derived from the log file; in this state, the  Energy consumption data segmented according to 
system is powered down but may still consume energy owing  scanner log file information
to ancillary systems; immediate scanning is not possible, and a 
power-up sequence of several minutes is needed before scanning
Table 2: CT Examination Energy Consumption according to Body Region
Energy Consumption (kWh) Duration (min) Cost ($)*
Examination Region Median Mean SD Median Mean SD Median Mean
Head-neck 0.38 0.44 0.24 5 7 4 0.07 0.08
Head-spine 0.42 0.48 0.34 5 7 5 0.08 0.09
Head-chest 0.51 0.60 0.41 6 7 5 0.09 0.11
Head 0.59 0.68 0.42 9 11 7 0.11 0.12
Chest 0.69 0.77 0.44 8 9 6 0.12 0.14
Pelvis 0.72 0.92 0.53 10 11 7 0.13 0.17
Spine 0.73 0.97 0.50 12 12 7 0.13 0.18
Neck 0.79 0.91 0.48 11 12 7 0.14 0.16
Abdomen-pelvis 0.97 1.28 0.72 12 14 9 0.17 0.23
Extremities 1.01 1.23 0.29 13 16 4 0.18 0.22
Chest-abdomen-pelvis 1.03 1.14 0.58 116 12 7 0.19 0.20
Cardiac 1.26 1.39 0.49 13 17 5 0.23 0.25
Runoff 1.28 1.41 0.50 15 19 7 0.23 0.25
Neck-chest-abdomen-pelvis 1.41 1.44 0.58 14 16 7 0.25 0.26
Trauma 2.01 2.05 0.40 28 29 6 0.36 0.37
Intervention 2.19 3.45 0.71 41 51 14 0.39 0.62
Average of all examination regions 0.99 1.22 0.39 13 16 5 0.18 0.22
Note.—SD = standard deviation.
* Assumed cost per kilowatt-hour = U.S. $0.18.
movement, scanning sequence parameters, and system states that was irrelevant in this context. However, the examination 
such as system on and system off. A major challenge was that start and end times are often manually entered into the sys-
every scanner used a different format for the log entries, and tem by the operating staff and therefore are limited as a basis 
logged events were very diverse. Scanner-specific scripts were for the definition of examination duration. For the purpose of 
used to extract important events that were then combined into this study, RIS data were mainly used to identify the type of 
new time series for analysis. examination.
RIS Data Data Curation and Analysis
RIS data contain workflow information describing the type of In conjunction with the School of Life Sciences at the Univer-
examination (eg, CT of the chest, whole-spine MRI), the ex- sity of Applied Sciences and Arts Northwestern Switzerland, 
amination start and end time stamps, and other information we developed an in-house software application to merge the 
Radiology: Volume 295: Number 3—June 2020  n  radiology.rsna.org 595
Energy- and Cost-saving Opportunities for CT and MRI Operation
Table 3: MRI Examination Energy Consumption according to Body Region
Energy Consumption (kWh) Duration (min) Cost ($)*
Field Strength and Examination Region Median Mean SD Median Mean SD Median Mean
1.5 T
 Abdomen 14.3 15.5 4.8 50 52 17 2.58 2.78
 Abdomen-pelvis 11.8 14.1 3.2 37 48 11 2.12 2.53
 Breast 15.8 15.8 0.0 42 42 0 2.84 2.84
 Cardiac 15.6 17.4 8.1 57 61 29 2.80 3.13
 Chest 17.6 16.8 4.5 47 53 14 3.16 3.03
 Extremities 16.1 17.6 3.1 46 51 9 2.89 3.18
 Head 16.1 15.9 7.2 43 49 26 2.90 2.86
 Head-neck 20.4 21.4 3.7 56 59 11 3.68 3.85
 Hip 14.1 14.1 0.0 40 39 0 2.54 2.53
 Neck 16.2 15.0 3.7 42 41 9 2.91 2.70
 Pelvis 14.4 17.1 6.5 46 57 21 2.59 3.08
 Runoff 12.8 11.9 3.6 36 43 15 2.31 2.15
 Shoulder 16.2 17.4 3.6 40 43 11 2.91 3.13
 Spine 15.4 16.6 6.2 42 50 27 2.78 2.98
 Vascular 18.5 19.0 1.9 59 58 7 3.34 3.43
 Whole body 24.6 26.7 4.1 71 73 12 4.42 4.81
 All 16.2 17.0 4.0 47 51 14 2.92 3.06
3.0 T
 Abdomen 14.7 14.7 0.0 37 37 0 2.64 2.64
 Chest 34.2 34.2 0.0 77 77 0 6.15 6.15
 Extremities 22.8 23.8 3.1 43 47 7 4.10 4.28
 Head 21.4 23.9 9.6 46 50 23 3.85 4.30
 Head-neck 25.3 26.2 4.1 48 51 10 4.56 4.72
 Hip 21.6 21.1 1.8 39 42 4 3.88 3.79
 Neck 30.0 30.4 5.0 64 65 10 5.41 5.48
 Pelvis 20.2 21.4 5.5 44 48 15 3.63 3.86
 Runoff 17.4 17.5 2.8 43 45 8 3.13 3.16
 Shoulder 21.4 23.5 4.6 41 47 12 3.85 4.24
 Spine 23.2 22.7 4.4 46 47 10 4.18 4.08
 All 22.9 23.6 3.7 48 51 9 4.13 4.25
Note.—SD = standard deviation.
* Assumed cost per kilowatt-hour = U.S. $0.18.
data streams of continuous energy consumption measure- The algorithm of the software uses the defined event time 
ment, scanner log information, and RIS examination details stamps to identify segments of scanner energy consumption 
for the entire year 2015. The data streams were synchronized different from the baseline energy consumption signal. A scan 
to temporally align any differences in data entry time stamps. event is included in the per-examination summary statistics 
The software allowed for segmentation of the energy consump- (Tables 2, 3) only if a typical energy consumption profile of 
tion signal based on scanner system activity states as defined increased consumption is found (Figs 1, 2); otherwise, the scan 
by means of log file or RIS data (Table 1). For instance, a scan event is excluded (approximately 5% of scans were excluded). 
event on a CT scanner was defined as the time period between This was done to exclude data that were not representative of a 
the scanner log events “patient open” and “patient closed” (Fig regular patient examination, such as checkup scans without a 
1). This corresponds to loading patient data from the Digital patient present, research scans in which a phantom was used, 
Imaging and Communications in Medicine worklist onto the or instances of temporally mismatched data streams. With this 
CT control console (“patient open”) to plan and perform scan- approach, all RIS-defined examinations (eg, “chest CT”) per-
ning and to closing the same patient case (“patient closed”) formed in the year 2015 with a given scanner could be pro-
when scanning was finished and image reconstruction was per- jected onto the energy consumption signal to segment the data 
formed. This time period is interpreted as patient room time and extract summary statistics including number of examina-
(the time during which the patient is likely in the CT room), tions; maximum, sum, mean, standard deviation, and median 
which includes patient preparation, patient positioning, scan- energy consumption in kilowatts and kilowatt-hours; and du-
ning, and patient exiting. ration (Figs 1, 2).
596 radiology.rsna.org  n  Radiology: Volume 295: Number 3—June 2020
Heye et al
Figure 1: Diagram shows depiction and analysis of energy consumption by software application by means of 
synchronization of radiology information system (RIS), CT system log file events, and energy data streams. A chest 
CT examination is segmented (blue area) based on CT system log file events such as “patient open” and “patient 
closed” and identified by means of the RIS examination label matching the CT system log file–defined segment. 
Kilowatt-hour consumption is calculated by integrating the area under the energy signal within a defined segment. 
Duration of an event is given according to the length of the segmented area. Red dots on energy consumption 
signal exemplify consumption in kilowatts at the respective location on the graph. Darker blue area segmented in 
energy consumption signal demonstrates identification of instance of peak energy consumption, which is used to 
define net scan energy consumption and duration. Recon = reconstruction.
We used additional scanner system states defined in log file a detached building per year in Switzerland as a reference for 
information, such as system active, on, and off, to extract sum- comparing energy consumption (4). The study took place in a 
mary statistics regarding general scanner operation. Another level 1 trauma tertiary care center university hospital with 773 
algorithm of the analysis software enabled the analysis of peak beds, approximately 1 million outpatient visits, and 35 000 in-
energy consumption by segmenting the energy consumption sig- patients treated in 2015. The radiology department performed 
nal based on a desired threshold value for kilowatts and duration, approximately 130 000 imaging examinations in 2015.
effectively excluding the baseline energy consumption. Peak en-
ergy consumption coincidence detection—that is, detection of Statistical Analysis
two scanners scanning at the same time—was done by using a Data are given in absolute numbers and relatively in percent-
script that detects and flags overlapping examination duration ages. Graph creation was performed with commercially avail-
time stamps between two CT scanners. We performed overall able software (JMP, version 14.2.0; SAS Institute, Cary, NC).
theoretical peak power consumption estimation of the entire 
scanner fleet by means of summation of the daily peak power Results
consumption value per scanner per day in 2015. Therefore, the The aggregated energy consumption of three CT and four MRI 
theoretical maximum combined peak power consumption of all scanners amounted to 614 825 kWh in 2015. Adjunct cooling 
scanners, if coincidence occurred, could be determined to gauge systems required 492 624 kWh in 2015, which is 44.5% of 
the resulting load on the hospital power grid. the combined energy consumption. Therefore, the operation 
We based cost calculation for energy consumption on 0.18 of seven cross-sectional imaging units, including examination 
Swiss francs per kilowatt-hour, which corresponded to 0.18 U.S. of 40 276 patients and cooling, resulted in a total energy con-
dollars per kilowatt-hour. To facilitate international comparison sumption of 1 107 450 kWh and a cost of $199 341. This is 
of the provided data, herein costs are presented in U.S. dollars. equivalent to the usage in a town of 852 people living in four-
We used the average energy consumption of a two-person house- person households, each requiring 5200 kWh annually. The 
hold (3550 kWh) and a four-person household (5200 kWh) for aggregated energy consumption of scanners and cooling repre-
Radiology: Volume 295: Number 3—June 2020  n  radiology.rsna.org 597
Energy- and Cost-saving Opportunities for CT and MRI Operation
Figure 2: Diagram shows 1 day of synchronized radiology information system (RIS), MRI system log file events, and energy 
consumption data streams for a 1.5-T MRI scanner as depicted by the software application. The same principle as in Figure 1 ap-
plies, in which MRI system file events define segments (blue areas) on energy consumption signal and are labeled according to RIS 
examination information (eg, MRI of abdomen). On bottom image, an MRI examination is zoomed in to demonstrate distinct energy 
consumption footprint of various types of imaging sequences (eg, T2-weighted [T2w], steady state free precession [SSFP], and T1-
weighted [T1w] sequences).
sented 4.0% of the total yearly university hospital energy con- 0.7 (minimum, 0.4; maximum, 3.5 kWh) including single- 
sumption of 27 905 332 kWh in 2015. The theoretical com- or multiphase acquisitions, with a mean energy cost of $0.22 
bined peak power consumption of all scanners, analyzed per 6 0.13 per examination. The daily energy consumption for 
aggregated daily maxima, was 380 kW 6 118 (range, 25–570 a clinical scanner in operation for 12 hours was 92 kWh for 
kW). The assumed maximal peak load on the hospital electric- CT scanner 1 and 50 kWh for CT scanner 2 in comparison 
ity grid was 1029 kW, whereas the vendor`s sum maximal peak to 88 kWh for the emergency room CT scanner, which is in 
load was listed as 1390 kW. Hence, the theoretical but rarely operation for 24 hours. Figure 4 summarizes the energy con-
occurring maximum of the aggregated peak load on the elec- sumption and duration for each scanner system state (net scan, 
tricity grid if all seven systems were simultaneously operating at active [patient room time], idle, on, and off). Overall, the three 
peak consumption would be 41.0% of the accumulated vendor CT scanners consumed 10 741 kWh (mean, 3580 kWh per 
peak specifications and only 27.3% of the average aggregated scanner) for actual scanning, 23 797 kWh during the patient 
daily maximum. room time period, 42 867 kWh (mean, 14 289 kWh per scan-
ner) during idle time, 66 664 kWh during the system-on state, 
CT Scanners and 78 679 kWh (mean, 26 226 kWh; $4721 in energy cost 
The summary of CT energy consumption, examination du- per scanner) in total per year. The operation of one CT scanner 
ration, and examination energy cost per examination body during a year, on average imaging 7904 patients, was compa-
region is given in Table 2 and illustrated in Figure 3. The rable in terms of energy consumption to the usage of five four-
mean energy consumption per body region was 1.2 kWh 6 person households. On average over a year, a scanner spent 
598 radiology.rsna.org  n  Radiology: Volume 295: Number 3—June 2020
Heye et al
Figure 3: Bar graph shows mean energy consumption of CT examination and standard deviation in kilowatt-hours per 
body region averaged from 1 year of data. AP = abdomen and pelvis, CAP = chest, abdomen, and pelvis.
4.2% (mean, 1.36 weeks; range, 2.58 days to 2.33 weeks) of $24 127 in energy cost; 3.0-T MRI: 149 655 kWh, $26 938 
its system-on time scanning, 27.2% with the patient being in in energy cost). MRI scanners consumed 35 478–46 704 kWh 
the room, and 72.8% (23.8 weeks) idle. A coincidence of scans per year during the scanner-off system state, representing 
between two of the three CT scanners occurred in 4.8% of all 31.2%–38% of their total yearly energy consumption owing 
actual scan events in 2015, resulting in a combined mean peak to the continuous operation of the cold head cooling system. 
energy demand of 65 kW 6 38 (range, 0.8–266.7 kW) (Fig 5). In contrast to CT operation, MRI idle-time energy consump-
tion was lower, at 5.5%–13.4% (range, 8177–16 038 kWh). 
MRI Scanners The mean monthly energy consumption (corrected for number 
The summary of MRI energy consumption, examination dura- of patients scanned) did not change after replacement of 1.5-T 
tion, and examination energy cost per examination body region MRI scanner 2a with the newer 1.5-T MRI model scanner 2b.
is given in Table 3 and illustrated in Figure 6. The mean energy 
consumption per body region was 20 kWh 6 5 (range, 12–34 Discussion
kWh), with a mean energy cost per examination of $3.57 6 We investigated the energy consumption of CT and MRI op-
0.96. The daily energy consumption for a clinical scanner in eration in a university hospital setting on various levels, ranging 
operation for 13 hours was 363 kWh with 1.5-T units and from a microscopic perspective focused on individual exami-
530 kWh with 3.0-T units. Table 4 and Figure 7 summarize nations to a macroscopic view of scanner activity system-state 
the energy consumption and duration for each scanner system energy consumption aggregated over an entire year. The data 
state: net scan, active (patient room time), idle, on, and off. analysis revealed the following aspects of relevance and energy-
In comparison, 25.8 four-person households expend the same saving opportunities. The operation of cross-sectional imaging 
amount of energy per year as one MRI scanner that imaged systems is energy intensive, and in this study the energy us-
an average of 4140.5 patients (overall averages: 134 037 kWh, age was comparable to that of a town of 852 people living in 
Radiology: Volume 295: Number 3—June 2020  n  radiology.rsna.org 599
Energy- and Cost-saving Opportunities for CT and MRI Operation
Figure 4: Bar graphs show distribu-
tion of system states for each CT scanner 
for sum duration in minutes and for sum 
energy consumption in kilowatt-hours 
during 1 year of data. Owing to some 
inaccuracies in energy consumption 
curve segmentation, aggregated totals 
for each scanner do not result in duration 
of exactly 1 year. Net scan consumption 
for the emergency room (ER) CT scan-
ner is comparable to that of four-person 
household; however, idle state consump-
tion is fourfold that of net consumption 
and therefore contributes largely to total 
energy consumption.
four-person households. The 
operation of cross-sectional 
imaging systems such as MRI 
and CT scanners, including 
indispensable adjunct cool-
ing systems, comprised 4% of 
our total yearly hospital energy 
consumption. In general, the 
energy consumption of dedi-
cated cooling systems com-
prised almost half of the total 
energy needed for operation of 
cross-sectional imaging systems 
in a radiology department. For 
CT, the largest share of energy 
consumption—approximately 
two-thirds—took place during 
the nonproductive idle system 
state; therefore, the degree of 
utilization was low and energy 
inefficient. For MRI, approxi-
mately one-third of energy 
consumption was attributable 
to the system-off state, inher-
ently owing to the need for 
constant helium cooling and 
operation of the cooling head.
On the basis of our results, 
some energy-saving opportu-
nities can be identified and 
potentially pursued. Increas-
ing energy efficiency may be 
achieved either by decreasing 
scanner energy consumption 
during nonproductive idle and 
system-off states or by increas-
ing the degree of utilization per 
time period. The first aspect 
can be addressed only by ven-
dors through introduction of 
low–energy consumption idle 
and system-off states (5). As 
with consumer electronics, the 
600 radiology.rsna.org  n  Radiology: Volume 295: Number 3—June 2020
Heye et al
Figure 5: Graph shows frequency of simultaneous scanning events (◊) occurring between any two CT scanners in a three-scanner setting during 12 months and the 
resulting distribution of accumulated peak energy consumption (inset, top right). The y-axis demonstrates maximum energy consumption during a CT examination, and the 
x-axis shows time of day. Each point represents a CT scan event; the size of the symbol reflects scan duration.
introduction of power-down or standby modes may allow for informing ancillary systems of the current system state may help 
considerable energy savings in the operation of CT scanners be- regulate the operation of adjunct cooling systems, decreasing the 
cause their actual productive energy demand is confined to the output during idle or system-off states instead of operation at 
short time span within seconds of actual scanning, a fraction of a fixed value. In addition, in the case of radiology departments 
the system-on state. The latter point may be optimized by radi- with multiple scanners installed, a fleet concept may be used, 
ology departments through improved workflow and optimized with a modular setup of ancillary systems such as those for cool-
patient throughput, resulting in a larger proportion of the energy ing in general or, in the case of MRI systems, specific cold head 
spent during productive states of scanners compared with non- helium cooling. This would allow scaling up of ancillary systems 
productive idle and system-off states. instead of requiring that each newly installed scanner have its 
Furthermore, because energy demands for adjunct cooling own separate adjunct system.
systems are considerable, alternative methods of counteracting To date, the awareness of energy conservation in hospitals 
waste heat may be pursued. Waste heat recovery methods such as appears low, and the focus in planning and operating hospi-
heat transfer by means of heat pipes or heat-storing technologies tals is more bound to workflow, redundancy, and high quality 
such as thermal banks may be used to recycle heat-related energy standards (9). However, hospitals have a high energy use inten-
rather than spending additional energy to neutralize excess heat sity and demand use intensity owing to their 24 hours a day, 
(2,6–8). 7 days a week operation; large building footprint; and special 
In general, in planning a radiology department, building requirements for redundancy and quality standards (10,11). As 
floor plans may be optimized to facilitate synergetic use of cool- shown, radiology departments claim a large proportion of the to-
ing systems or scanner architecture. For instance, scanner signals tal hospital energy consumption. Consequently, the leverage or 
Radiology: Volume 295: Number 3—June 2020  n  radiology.rsna.org 601
Energy- and Cost-saving Opportunities for CT and MRI Operation
Figure 6: Bar graphs show mean energy consumption of MRI examinations and standard deviations in kilowatt-hours 
per body region stratified according to field strength and averaged from 1 year of data. AP = abdomen and pelvis.
potential for energy savings appears sufficiently large to warrant achieved in MRI systems owing to technical improvements 
exploring these opportunities from an ecologic as well as an eco- and in CT scanners owing to low-power mode or shutdown 
nomic perspective. Energy conservation technologies developed, during off hours, the European Coordination Committee 
implemented, and established in other residential or commercial of the Radiological, Electromedical and Healthcare IT In-
sectors may be transplanted to the health care sector without dustry report acknowledges that for these modalities, saving 
considerable investment. In addition, apart from the urgently re- energy through reduction of system idle state consumption 
quired reduction in energy consumption and its associated CO2 is complex and some technical developments may further 
emission footprint, the return on investment in terms of reduc- increase energy consumption. On the other hand, refurbish-
ing energy costs is similarly beneficial and likely more convinc- ment of medical devices is increasing and able to reduce 
ing for executives. Nevertheless, with carbon emission reduction the energy needed to manufacture a medical device, with 
goals from the Paris Agreement (12), it is mandatory that we approximately 5% of MRI and CT scanners sold as refur-
look into all possible carbon emission reduction opportunities bished units in 2016 (3).
to mitigate climate change. Aside from energy conservation implications, our study 
There have been some activities to improve the energy provides sufficient data to model and extrapolate the energy 
efficiency of medical devices, such as the self-regulatory ini- consumption of any given radiology department based on 
tiative of the European Coordination Committee of the Ra- certain parameters, such as scanner fleet details, profile of ser-
diological, Electromedical and Healthcare IT Industry. For vice and type, and number of examinations performed per 
instance, there was a reduction of 20% in average annual year. These models may be used for planning new installa-
energy consumption for U.S. devices in 2012 compared tions or radiology department expansions and may also help 
with 2005 (3). However, although energy savings were identify the size and scope of required ancillary systems—for 
602 radiology.rsna.org  n  Radiology: Volume 295: Number 3—June 2020
Heye et al
Table 4: Distribution of MRI Energy Consumption and Duration during 12-month Period Stratified according to Scanning Events
Energy Consumption (kWh) Duration (min)
Scanner and System State Sum Total (%)* Mean SD Median Sum Total (%)† Mean SD Median
1.5-T MRI scanner 1
 Active 59 634 63.3 15 13 13 207 245 99.8 54 66 44
 Idle NA‡ NA‡ NA§ NA§ NA§ 302 0.2 NA§ NA§ NA§
 On 58 672 62.3 177 86 213 207 547 44.8 625 296 754
 Off 35 478 37.7 107 51 90 255 629 55.2 772 373 659
 Total 94 150 100.0 NA§ NA§ NA§ 463 176 100.0 NA§ NA§ NA§
1.5-T MRI scanner 2a||
 Active 58 857 49.0 15 15 13 154 426 69.9 38 80 32
 Idle 16 038 13.4 NA§ NA§ NA§ 66 528 30.1 NA§ NA§ NA§
 On 74 895 62.3 243 61 255 220 954 43.7 718 147 755
 Off 45 238 37.7 147 116 106 284 256 56.3 923 709 694
 Total 120 133 100.0 NA§ NA§ NA§ 505 210 100.0 NA§ NA§ NA§
1.5-T MRI scanner 2b||
 Active 52 069 50.4 15 13 14 156 730 76.4 45 54 37
 Idle 12 053 11.7 NA§ NA§ NA§ 48 384 23.6 NA§ NA§ NA§
 On 64 122 62.0 213 94 229 205 113 43.3 680 316 721
 Off 39 288 38.0 130 88 99 268 704 56.7 710 512 716
 Total 103 410 100.0 NA§ NA§ NA§ 473 817 100.0 NA§ NA§ NA§
3.0-T MRI scanner
 Active 94 774 63.3 22 18 20 197 770 86.9 46 53 39
 Idle 8177 5.5 NA§ NA§ NA§ 29 837 13.1 NA§ NA§ NA§
 On 102 951 68.8 321 112 344 227 606 45.5 709 265 750
 Off 46 704 31.2 146 100 113 272 462 54.5 851 560 691
 Total 149 655 100.00 NA§ NA§ NA§ 500 068 100.0 NA§ NA§ NA§
Note.—Mean, standard deviation (SD), and median are calculated based on the number of events per system-state category. For each 
system-state category, this translates to a different meaning (eg, basic statistics for “active” reflect per-individual scan data). For “on” or 
“off” states, basic statistics represent per-day data because an MRI scanner is usually powered down in the evening. However, there is some 
variance in usage (eg, number of “on” and “off” system states, respectively) owing to power up and down for emergency scans during night 
times or weekend shifts. NA = not applicable.
* Percentage of the total.
† Active + idle = 100% of on duration. On + off duration = 100% of total.
‡ Negative value due to some overlap in energy signal segmentation.
§ There was not a meaningful denominator to calculate basic statistics. There was no data event that is identified as idle or total; these data 
are calculated from active, on, off events.
|| One MRI system (1.5-T MRI scanner 2a) was replaced during 2015 with a new MRI scanner (1.5-T MRI scanner 2b). Numbers are 
extrapolated 12-month data based on measurement during 3 months (scanner 2a) and 7 months (scanner 2b).
instance, the required energy capacity of the local hospital CT scanners. The scanner fleet composition measured in this 
electricity grid. study may not be representative of other departments and instal-
Peak energy consumption is a cost factor because it deter- lations, but the detailed measurements should allow for a reason-
mines the theoretical capacity of the hospital electricity grid and able estimation and extrapolation for a differing infrastructure 
its magnitude is priced in by the electricity provider. Coinci- and usage profile.
dence of peak power consumption across a scanner fleet may The energy cost calculations in this study are based on 
be compensated for either by ancillary energy buffer systems to the kilowatt-hour price specific to our hospital and may not 
reduce the load on the electricity grid or by traffic control of scan be comparable to that of other countries—for instance, the 
events to avoid simultaneity of peak consumption. United States—or a region such as Asia, because energy costs 
This study had some limitations. The measured energy con- differ greatly. However, with simple multiplication, the pro-
sumption was based on information from one vendor. Although vided kilowatt-hour values can be converted to the respec-
values from other vendors might differ, the variation should re- tive local kilowatt-hour price, allowing for comparable cost 
side within an acceptable range, possibly within 1 standard devi- calculations.
ation per examination measured in this study, because of techni- In conclusion, CT and MRI scanners are energy intensive in 
cal and physical requirements necessary for operating MRI and their operation and constitute a large proportion of total hospital 
Radiology: Volume 295: Number 3—June 2020  n  radiology.rsna.org 603
Energy- and Cost-saving Opportunities for CT and MRI Operation
Figure 7: Bar graphs show distribution of system states for each MRI scanner for sum duration in minutes and for sum energy consumption in kilowatt-hours during 1 year 
of data. * = minimal idle time for one MRI scanner, which is possibly a false estimate resulting from summation errors and variance in energy signal segmentation. One MRI 
system (1.5-T MRI scanner 2a) was replaced during 2015 with a new MRI scanner (1.5-T MRI scanner 2b). Presented data are extrapolated for a 1-year time period.
energy consumption. However, considerable energy- and cost- present article: disclosed no relevant relationships. Other relationships: disclosed no 
saving potential is present during idle and system-off states, relevant relationships. R.K. disclosed no relevant relationships. T.W. disclosed no relevant relationships. D.M. disclosed no relevant relationships. A.C. disclosed no 
which can be converted to more energy-efficient operating relevant relationships. M.L. disclosed no relevant relationships. M.P. disclosed no rel-
modes. Our data may allow for detailed modeling of the energy evant relationships. M.D. Activities related to the present article: institution received a 
consumption of a given radiology department infrastructure for grant from University Hospital Basel, Switzerland. Activities not related to the present article: disclosed no relevant relationships. Other relationships: disclosed no relevant 
planning and optimization purposes. relationships. R.L. disclosed no relevant relationships. D.B. Activities related to the 
present article: institution received a grant from University Hospital Basel, Switzer-
Author contributions: Guarantors of integrity of entire study, T.H., T.W., D.M.; land. Activities not related to the present article: disclosed no relevant relationships. 
study concepts/study design or data acquisition or data analysis/interpretation, all au- Other relationships: disclosed no relevant relationships. E.M. Activities related to 
thors; manuscript drafting or manuscript revision for important intellectual content, the present article: institution received a grant from Siemens Healthineers. Activities 
all authors; approval of final version of submitted manuscript, all authors; agrees to not related to the present article: is on the board at Siemens Healthineers; institution 
ensure any questions related to the work are appropriately resolved, all authors; litera- received grants/grants pending from Siemens Healthineers; institution received pay-
ture research, T.H., D.M., R.L., E.M.; experimental studies, R.K., D.M., M.L., M.D., ment for lectures including service on speakers bureaus from Siemens Healthineers. 
D.B.; statistical analysis, T.H., R.K., A.C., M.L., M.D., R.L., D.B.; and manuscript Other relationships: disclosed no relevant relationships.
editing, T.H., R.K., T.W., D.M., M.L., M.P., M.D., R.L., D.B., E.M.
References
Disclosures of Conflicts of Interest: T.H. Activities related to the present article:  1. Commission Regulation (EC) No 1275/2008. https://eur-lex.europa.eu/legal-content/ 
institution received a grant from Siemens Healthineers. Activities not related to the EN/TXT/HTML/?uri=CELEX:32008R1275&from=EN. Accessed July 4, 2018.
604 radiology.rsna.org  n  Radiology: Volume 295: Number 3—June 2020
Heye et al
 2. Kolokotsa D, Tsoutsos T, Papantoniou S. Energy conservation techniques for hospi-  7. Bujak JW. Production of waste energy and heat in hospital facilities. Energy 
tal buildings. Adv Build Energy Res 2012;6(1):159–172. https://doi.org/10.1080/1 2015;91:350–362.
7512549.2012.672007.  8. Noie-Baghban SH, Majideian GR. Waste heat recovery using heat pipe heat ex-
 3. COCIR Self-Regulatory Initiative for the Ecodesign of Medical Imaging Equipment changer (HPHE) for surgery rooms in hospitals. Appl Therm Eng 2000;20(14): 
Status Report. 2018. COCIR; 2018. https://www.cocir.org/fileadmin/6_Initiatives_ 1271–1282.
SRI/SRI_Status_Report/COCIR_SRI_Status_Report_2018_-_June_2019.pdf. Ac-  9. Rohde T, Martinez R. Equipment and Energy Usage in a Large Teaching Hospital in 
cessed July 20, 2019. Norway. J Healthc Eng 2015;6(3):419–433.
 4. Nipkow J. SAFE study—typical household energy consumption. 13-09-2019. Zürich:  10. Chung JW, Meltzer DO. Estimate of the carbon footprint of the US health care sec-
Schweizerische Agentur für Energieeffizienz S.A.F.E., 2013; 2. https://www. tor. JAMA 2009;302(18):1970–1972.
energieeffizienz.ch/dam/studien/2013_typischer_haushalt_stromverbrauch_d/pdf_  11. Hu SC, Chen JD, Chuah YK. Energy Cost and Consumption in a Large Acute 
de/SAFE_typischer_Haushaltstromverbrauch_12_2013/Der typische Haushalt-SV- Hospital. Int J Archit Sci 2004;5(1):11–19. https://www.bse.polyu.edu.hk/research-
SAFE-Dez-2013.pdf. Accessed March 14, 2018. centre/fire_engineering/summary_of_output/journal/IJAS/V5/p.11-19.pdf.
 5. Herrmann C, Rock A. Magnetic Resonance Equipment (MRI)—study on the  12. United Nations. Paris Agreement COP21. 21 Convention on Climate Change. 
potential for environmental improvement by the aspect of energy efficiency. doi:FCCC/CP/2015/L.9. https://unfccc.int/process-and-meetings/the-paris-agree-
2012;1–30. https://www.eceee.org/static/media/uploads/site-2/ecodesign/products/ ment/the-paris-agreement. Published 2015. Accessed June 7, 2019.
medical-imaging-equipment/improvement-potentials-for-mri-2012-03-09-draft-
for-discussion-with-cf.pdf. Accessed April 8, 2018.
 6. Poerner M, Rimpel A. Waste heat recovery. In: Brun K, Friedman P, Dennis R, 
eds. Fundamentals and Applications of Supercritical Carbon Dioxide (sCO₂) Based 
Power Cycles. Duxford, England: Woodhead/Elsevier, 2017; 255–267.
Radiology: Volume 295: Number 3—June 2020  n  radiology.rsna.org 605