Edwin Ronald

E. Frederick, W. Hahn,

BS

PhD2 #{149} Michael #{149} Gerald Entine,

R. Squillante, PhD

Accurate Automatic for Mammography:

terms: #{149} Physics

Breast

radiography,

#{149} Radiography,

much

1991;

is one of the techniques for

ic range.

safer.

Problems

still

must

subjectively

degree

technology,

exist,

how-

anticipate

of exposure

based

breast compression ly adjust a “darkness”

technology,

The need frequently

178:393-396

in the

and

x-ray

detectors

manualknob.

used,

is

software studied.

control Improvements

I

Hunt

From

Radiation

St,

Watertown,

Monitoring MA

Devices,

02172.

Received

Inc.

Current

c RSNA,

address:

1991

Eaton

Corp.

Beverly,

Mass.

existing

units

and

The

system

ride

al-

detectors.

design

mithms

are

uses

prototype to be

mammography

to be built

tector

into

four The

and

new

ones.

cadmium details

tellu-

of the

software

discussed

MATERIALS

44

April

26, 1990; revision requested June 20; revision received September 7; accepted September 1 1. Supported in part by National Cancer Institute contract N44-CM-67807. Address reprint requests to MRS. 2

has

added

onto

laboratory the potential

de-

algo-

the

diagnostic

0.1

mm

energy

dynam-

range

is less

telluride.

than

Cadmium

telluride

detectors

(Radiation

that

Monitoring

Mass)

were

were

Devices,

fabricated

Wa-

on crystal-

were

etched

polished

and chemically

in

to

provide smooth, clean electrodes were applied uum deposition. Because the detectors

surfaces. Metallic by means of vac-

small

have

and

breasts

are

may

variations

in density,

age breast

attenuation

relatively significant

estimating

the

requires

aver-

that multi-

ple detectors be used to monitor a representative area of the image. If an insufficient area is monitored, the system could calculate an incorrect exposure time for most mined

of the breast by examination

of radiographs

image.

that even

tissue

adequate

for

with

It was of a large

performance

highly

the

deternumber

textured

signals

from

three

colinear detectors. The signal from the detector receiving the middle-value flux level was used to determine the exposure.

controller.

wide

detector

1 cm wafers

breast

METHODS

and

x-ray

sensitivity to lowwide dynamic range, and a Photovoltaic cadmium telmeet these requirements.

line wafers of cadmium telluride diameter and 2 mm thick. The

ed

sensitivity

for the high

in cadmium

tertown,

Such

high

system.

tellunide is a semiconductor with a band gap of 1.47 eV, which makes it nearly ideal for use at room temperature. This cornbination of properties provides excellent sensitivity to x rays and little noise. The

Research into the new AEC system was divided into two areas: (a) detector design and (b) development of data analysis and control algorithms. Development of the detector required modifying the design of an existing x-ray detector that providboth

calithe

Design

The requirements were that it have

is obtained

below.

AND

empirical

to optimize

of the

Detector

used

had to be made both in the design of the x-ray detector and in software for data analysis and control. The result is a stand-alone system that

and

measurements

performance

in-

of fundamental

relationships

bration

cadmium

harden-

ing of the x-ray beam, and failure of the photographic reciprocity law (1,2). To address these problems, a new AEC detector system complete with the appropriate gomithms was

physical

of algorithms

combination

Cadmium telluride provides a high x-ray attenuation because it has an average atomic number of 50 and a density of 6.2 g/cm3 (0.62 kg/rn3). At 30 keV, for example, the mean free path of x rays within

the

for such subjective input attributed to limitations

the

energy x rays, fast response. luride detectors

on the

then control

#{149}

Development

volved

even, in obtaining adequate control of the exposure times to prevent the need for performing the procedure again. Many designs for automatic exposure controllers (AECs) have been tested, but most do not work well over the typical mange of breast size and density encountered. With most AECs, an experienced technolo-

00.11

Radiology

AMMOGRAPHY most reliable

MS

Controller Performance’

early diagnosis of breast cancer. Over the past few years, advances in the design of mammography machines and in film have significantly meduced the dose of x nays received by a patient and have made the procedure

gist Index 00.11

J Cirignano,

#{149} Leonard

Exposure Design and

M

An automatic exposure controller has been designed that controls the optical film density for film, screen, and radiographic techniques typically used in mammography to within 0.05 over a range of 1.3-6.7cm thickness of Lucite. This degree of accuracy is better than that reported for presently available controllers. The detector system consists of four cadmium telluride detectors and involves the use of a control algorithm to read the detectors and turn off the mammography unit at the correct time. This algorithm is implemented by a microprocessor, which also provides the means for a convenient calibration.

PhD

if one

a design

proved

of the

detectors

tenuating

smaller Abbreviation:

region,

to work

was

as could

well

even

outside

the

happen

for

at-

breasts. AEC

=

automatic

exposure

393

Figure

1.

detector lunide

Detector configuration

housing showing for four cadmium tel-

array

photovoltaic

detectors.

The

center-to-

center spacing of detectors A, B, and C is 2 cm. The spacing between C and D is 4 cm. These values were selected as appropriate based on the examination of 12 randomly chosen mammograms.

a.

b. 3.0

2.5 2.0

0

4

8

12

Lucite

16

Lucite

Figure

Signal current versus Lucite layer thickness for two photovoltaic cadmium telluride devices at increasing tube voltages. The detectors were similar in size and con2.

struction

and

varied

only

in

the

To ensure

satisfactory

the

in-

size, it is necessary

to

selected;

for a large

triplet A, C, and D is used. The is whether detector D is exposed direct x rays.

Algorithm

with

software

was developed

data

control

and

the

to analyze time.

attenuation of a monoenergetic beam is given by the well-known

where unit

N

area

absorbed N,

i

is the

per

time

nor scattered,

is the

linear

of

that N0

material, through

For polyenergetic

x-ray

394

emitted

Radiology

#{149}

and x is the the material.

beams,

by radiography

energy.

How-

radiation without bias), have a nearly

on the detector, photocurrent

follows

I

such machines,

as

=

an

the “2’

I0e,

,‘

where

I is the

is a function

photogenerated of both

detector

and

response,

current, the

and

do

a

cadmium

x is the

I

used:

Ortho-M

where

-a,

which

is seen

to be a

of the kilovolt peak. To develop an algorithm for controlling film exposure, it was also necessary to determine the film and screen response as a function of Lucite thickness.

for

different

opti-

Two mammographic

different

film

(Eastman

speeds

were

Kodak,

Roches-

ter, NY) and SP15O (Du Pont, Wilmington, Del). Each film type was used with a Min-R screen and cassette (Eastman Kodak). Exposures were made with a Mammomat-B mammographic unit (Siemens Medical Systems, Iselin, NJ). By means of a 15-step Lucite phantom, each exposure provided 16 data points. A plot of net optica! film density versus Lucite layers for both films are shown in Figure 3 for a family of exposures at two kilovolt-peak settings. Figure 4 was then generated with the values obtained from the data in Figure 3

of each

line is

of data

phantom. with

by ty.

thickness.

sets

films

Figure 2 shows the detector response as a function of attenuation by Lucite (Du Pont Diagnostic Imaging Division, Billerica, Mass). The slope

attenuator

this,

wedge

the cad-

function

coefficient

of the attenuating propagation length those

per

are neither is the incident

To

cal film densities were obtained by exposing the film and screen at various exposures between 5 and 320 mAs with a step

of x rays incident

The

(1)

attenuation

vary

the

x-ray equa-

photons

not

function

a range a matea beam

mium telluride equation

N0CCX,

number

unit

does

linear response to x-ray flux over of six orders of magnitude. When rial with a low Z value attenuates

telluride =

with

mode

tion N

varies

(ie, detecting externally applied

breast,

research,

exposure

flux

exponential

provide a good approximation (3). Cadmium telluride detectors, when optimized and operated in the photovoltaic

criterion to the

the detector

photon

as a simple

ever, for materials with low atomic numbers (Z) such as tissue, Equation (1) does

Development

In parallel

transmitted

exactly

of x because

the spacing between the three detectors, depending on whether the attenuation is by a large or small breast. This variance was achieved with four detectors spaced as shown in Figure 1 . For a small breast, the detector triplet A, B, and C is vary

automatically

d.

Figure 3. Optical film densities of two mammographic films for each region of the step wedge phantom obtained at varying milliampere-second settings: (a) Kodak film at 28 kVp, (b) KOdak film at 35 kVp, (c) Du Pont film at 28 kVp, (d) Du Pont film at 35 kVp. The numbers adjacent to each curve are the milliampere-second values used. These curves are similar to a “reversed” Hurter and Dniffield curve, plotted as a function of phantom thickness rather than as a function of x-ray intensity.

preparation

performance

of breast

Layers

Lucite

Layers

c

process of the semiconductor surface. Detector 3 (Det #3) was used in all of the subsequent measurements.

dependent

8

8

Layers

means From

of a constant Figure 4 we

It the

=

J is the x-ray exposure time.

characteristic calibration (3) can

combined

both

of the constants. be solved

optical film densiobtain the relation

Jot tube current is a constant

film,

and

and t is that is

to get Jt as a function

to are

Jo and

Equations for e’ and

(2)

and

then of I:

February

1991

lows 2BkVp

35kVp

A

on

KODAK 4

DUPONT

0’

. 0

16

Lucite

.

4

a.

.

8

Lucite

Layers

12

16

(3).

4.

Data

obtained from J is the x-ray voltage 28 kVp;

exposing film with a density of 1.0 through Lucite layers of tube current and t is the exposure time as given in Equation (b) tube voltage 35 kVp.

thickness.

(a) Tube

kilovolt-peak

mammography

selection

unit.

The

sor.

In both calibration modes, time is determined with the settings on the mammography I 750

U

1 500

1 500

any E

1 250

.

1 000

number

I 250

thickness

I 000

until

of times,

and

the

0 750

0 500 2

2

4

Lucite

Thickness

3

4

Lucite

[cmj

5

6

Thickness

7

[cm]

Figure 5. Film telluride-based points

are

for

density

results

AEC

to control

single

obtained

at 25 kVp

exposure

times

(a) and

30 kVp

for the Lorad

(b)

by using

mammography

the

Data

measurements.

(4)

gle constant

be combined

c, and t

into

a sin-

we have (5)

=

where K(J) = JoCJoto/J is a constant for a given current J. Thus the control circuit must determine the correct exposure time and turn off the mammography system

by means

of Equation

and c need calibration

(5). The constants

to be determined measurements.

by means

phy

machine.

uation.

Once

The

circuit

then

selection

calculation

data

of time

by an Intel 8052AH-BASIC

Design

sor that

basic

has a built-in

the

pulse accordof the selected

time t described

pro-

only

are saved and logged by the

once

exceptions

are

interpreter

values

off, calibrate A (thin Lucite), calibrate B (thick Lucite), and operate. Once calibrated, the AEC operates without the need for any interven-

two AEC electronic

compo-

nents, and (d) the radiography machine. The principle behind the design of the electronics is as follows. The signal current from each detector is the input to a current-to-voltage

en circuit. is used

converter

The

output

to select

the

and

voltage

amplifi-

of this

appropriate

stage

triplet at the beginning of the exposure based on the relative amplitude of the voltage for detector D (see Detector Design above). The three voltage outputs of the selected triplet are then sampled and digitized by means of a multiplex analogto-digital converter. Each digitized signal is then

Volume

used

178

to calculate

Number

#{149}

the

2

optimal

expo-

and

a real-

Operation

tion

and

a relatively

thus

can

be

one panel switch:

successfully

inexperienced

control.

used

by

radiography

technician.

Calibrate

detector

functions

The AEC has only This is a four-position

AEC

panel

the

are

operation

of the

requires that the two constants K and c in Equation (5) be determined for each detector and each kilovolt peak. To achieve this, a pair of flat sheets of Lucite for each kilovolt-peak setting are used as attenuation standards. The AEC detersystem

mines and stores The measurement

calibra-

possible

techthe noaging

retained

in the

in the

event

operate

on the

switch

that

would

cause

setting

also

on

position.

The

mammography

are set to a milliampere-second This

sys-

of a power

control

to the

controls

a slight serves

unit

setting overexposure.

as a fail-safe

timer

is mithe AEC

terminates the exposure. The optical film density of the mammogram at the location of the controlling cadmium telluride

detector mined

should be the density during the calibration

ticular

kilovolt

peak

as deterfor that par-

to within

the required is accomplished

constants. as fol-

AND

the

accura-

CONCLUSIONS

The AEC system LR-1 mammography

was

tested with machine

(Lomad Medical Systems, Danbury, Conn). With this unit, exposure times of 0.2-5 seconds and x-ray tube anode values of 22-30 kVp can be selected

from

the

R intensifying

control

panel.

screen

and

A Mm-

Ortho-M

film were used (Eastman Kodak). system was calibrated at 25 and

kVp mode-The

is set

RESULTS

tween

these

the

the

AEC

Lucite

the x-ray operator;

microproces-

BASIC

with time

math

if the

cy of the AEC.

t is done

components of the complete system are (a) the cassette that contains the x-ray film, the intensifying screen, and the four cadmium telluride photovoltaic detectors, (b) the AEC electronics (analog and digital), (c) the interface be-

intrinsic clock.

is achieved.

procedure

override for the AEC. An exposure tiated in the normal manner, and

selects

is made,

the turn-off

ing to the calculated detector. The above and

of this voltage the mammogra-

of the exposure time, the middle value atten-

the

AEC generates

cessing

System

A logic

the middle value which represents K

of

sure time t, according to Equation (5), for the attenuation at each detector location. Because the correct exposure time is a function of the x-ray tube voltage, an analog voltage representative is input to the AEC from

density

tion sheets niques are

calibration

Jt = can

fl/a

Lucite

setting

trial-and-error

tern memory even failure. Operate mode-The Finally,

the

and replacement of the x-ray tube. At most, this procedure would need to be carried out only two times per year. The

cadmium

system.

film

this

is required

table

b.

changing

milliampere-second

desired

However, 0 750

the exposure panel control unit itself,

not with the AEC. If the proper film exposure is not achieved during either calibration step, those steps can be repeated

3OkVp

Vp

255 1 750

correct

exposure time (or milliampere second) is selected for a thin Lucite calibration phantom. The AEC control knob is switched to calibrate A, and an exposure is made; the AEC measures current I from the detectors and the length of time t that the mammography unit is running an exposure. The procedure is repeated with the AEC switched to calibrate B and with the thick Lucite calibration phantom in the beam. From the values obtained, the unknowns K and c from Equation (5) are calculated and stored by the microproces-

Layers

b.

Figure varying

for a given the

with

target

film

densities

The 30

set at

1.2.

The

design

allowed

a simple

wire interface between the Lorad system. The sign required that one from the Lomad system

dicate in use,

which one

anode lead

kilovolt

to the

three-

the AEC and interface deelectrical lead be used to in-

peak

Lonad

Radiology

is

system

395

#{149}

transmits ground

The

the stop pulse, and a lead connects the two units.

AEC

from the tom itself

the

received Lonad directly

its start system sensing

x-ray beam. Tests to evaluate

ty of the

AEC

the

and

the

signal the detecthe onset

of

reproducibili-

performance

ned out. First, several were carried out with attenuator,

by

ment controlled exposure as a function of attenuator thickness, we made exposures over a wide range of Lucite thickness and at two x-ray tube

were

exposures a given signal

from

car-

a sin-

gle detector was used to terminate the exposure. Variations in exposures that were terminated by signals from each of the four detectors were also evaluated. The standard deviation in resulting film density from run to run (for a given detector) or from detectom to detector was less than ±0.03. To determine how well the instnu-

Radiology

#{149}

quality

the

need

and

1.

1.3-4.5

cm

at 25 kVp,

exposures

densities for which

within 0.05 the system

resulting

the

AEC

These results cleanly demonstrate that the cadmium tellumide detectors combined with a suitable algorithm developed for determining the connect exposure time overcome the dif-

without

manual

adjust-

U

References

in film

of the density was calibrated.

mammograms

for “expert”

ment.

Iklason

LT,

Barnes

GT,

Rubin

E.

raphy phototimer technique gy 1985; 157:539-540. 2.

LaFrance

R, Gelskey

cuit modification graphic phototimer gy 1988; 3.

Schleien

Health diological Nucleon

DE, that

Mammog-

chart. Barnes

RadioloGT.

improves performance.

A cir-

mammoRadiolo-

166:773-776. B, Terpilak

Physics.

MS.

In: The

eds.

health

Medical

physics

health

handbook.

Olney,

Lectern

Associates,

1984;

raMd:

chap

10.

ficulties encountered with other methods of exposure control. The resulting film density is close to the set density and is independent of attenuatom thickness over the range of in-

terest

396

congood-

voltages. The results are shown in Figure 5. The data indicate that, over the manges of 2.25-6.75 cm at 30 kVp produced

Lucite

tion in the clinical setting, this trollem should reliably provide

for mammography.

In opera-

February

1991