A novel wafer-scale CMOS APS X-ray detector for breast

6
Journal of Instrumentation A novel wafer-scale CMOS APS X-ray detector for breast cancer diagnosis using X-ray diffraction studies To cite this article: A Konstantinidis et al 2012 JINST 7 C12005 View the article online for updates and enhancements. You may also like Relative gravimeter prototype based on micro electro mechanical system A S A Rozy, H A Nugroho and M Yusuf - ADX: a high field, high power density, advanced divertor and RF tokamak B. LaBombard, E. Marmar, J. Irby et al. - Detector control system for the CBM-TOF S. Dong, G.M. Huang, J. Frühauf et al. - This content was downloaded from IP address 186.105.112.119 on 26/02/2022 at 13:53

Transcript of A novel wafer-scale CMOS APS X-ray detector for breast

Journal of Instrumentation

A novel wafer-scale CMOS APS X-ray detector forbreast cancer diagnosis using X-ray diffractionstudiesTo cite this article A Konstantinidis et al 2012 JINST 7 C12005

View the article online for updates and enhancements

You may also likeRelative gravimeter prototype based onmicro electro mechanical systemA S A Rozy H A Nugroho and M Yusuf

-

ADX a high field high power densityadvanced divertor and RF tokamakB LaBombard E Marmar J Irby et al

-

Detector control system for the CBM-TOFS Dong GM Huang J Fruumlhauf et al

-

This content was downloaded from IP address 186105112119 on 26022022 at 1353

2012 JINST 7 C12005

PUBLISHED BY IOP PUBLISHING FOR SISSA MEDIALAB

RECEIVED September 13 2012ACCEPTED November 7 2012

PUBLISHED December 7 2012

14th INTERNATIONAL WORKSHOP ON RADIATION IMAGING DETECTORS1ndash5 JULY 2012FIGUEIRA DA FOZ PORTUGAL

A novel wafer-scale CMOS APS X-ray detector forbreast cancer diagnosis using X-ray diffractionstudies

A Konstantinidisa1 Y Zhenga D Philipb S Vinnicombec and R Spellera

aDepartment of Medical Physics and Bioengineering University College LondonLondon WC1E 6BT UK

bSchool of Engineering and Materials Science Queen Mary University of LondonLondon E1 4NS UK

cRadiology Department University of DundeeDundee DD1 4HN UK

E-mail anastasioskonstantinidisuclacuk

ABSTRACT The current study uses a novel large area (128 cm times131 cm) complementary metal-oxide-semiconductor (CMOS) active pixel sensor (APS) X-ray detector named Dynamic rangeAdjustable for Medical Imaging Technology (DynAMITe) for breast cancer diagnosis The de-tector consists of two geometrically superimposed grids a) 2560times 2624 fine-pitch grid of pixels(50 microm pitch) named Sub-Pixels (SP camera) for low intrinsic noise and high spatial resolu-tion and b) 1280times 1312 large-pitch grid of pixels (100 microm pitch) named Pixels (P camera) forhigh dynamic range X-ray performance characterization measurements show that the detectivequantum efficiency (DQE) of the SP camera is in the range 07ndash075 at low spatial frequenciesusing a tungsten (W) anode X-ray source at 28 kV Hence the detector is suitable for mammogra-phy Furthermore we used the SP camera to combine mammograms with angle dispersive X-raydiffraction (ADXRD) measurements in order to apply the X-ray biopsy concept in one examina-tion The results show that ADXRD technique indicates the presence of cancer in suspicious areason the mammogram Hence it could be used to determine the region affected by cancer and assistin planning surgery This study is the proof of concept that mammography and ADXRD can becombined in one examination

KEYWORDS X-ray radiography and digital radiography (DR) X-ray diffraction detectors

1Corresponding author

ccopy 2012 IOP Publishing Ltd and Sissa Medialab srl doi1010881748-0221712C12005

2012 JINST 7 C12005

Contents

1 Introduction 1

2 Materials and methods 121 Evaluation of the X-ray performance of the detector 122 Angle dispersive X-ray diffraction (ADXRD) measurements 2

3 Results and discussion 331 Detective quantum efficiency results 332 X-ray biopsy results 3

4 Conclusions and future work 4

1 IntroductionDigital X-ray detectors based on complementary metal-oxide-semiconductor (CMOS) active pixelsensor (APS) technology have been recently introduced in mammography [1] However conven-tional mammography does not always show the extent of tumour infiltration This study uses anovel large area CMOS APS X-ray detector to combine mammographic imaging with angle dis-persive X-ray diffraction (ADXRD) measurements to solve this problem

The Multidimensional Integrated Intelligent Imaging (MI-3) Plus consortium has developeda wafer scale (128 cm times131 cm) 3-T CMOS APS X-ray detector named DynAMITe (Dynamicrange Adjustable for Medical Imaging Technology) DynAMITe consists of two geometrically su-perimposed grids with two pixel pitches a) the Sub-Pixels (SP) camera (at 50 microm) for low noiseand high spatial resolution and b) the Pixels (P) camera (at 100 microm) for high dynamic range Thediodes of the two arrays are reset at different voltages to control the different depletion depthsof each diode Hence the imager can be used either as a 50 microm resolution or a 100 microm resolu-tion sensor or a combination of resolutions with extended dynamic range within the same pixelarray This study evaluates the X-ray performance of SP camera in mammography In particularthe presampled modulation transfer function (pMTF) and detective quantum efficiency (DQE) ofDynAMITersquos SP camera are compared to those of other mammographic detectors

Previous studies have demonstrated that X-ray diffraction can be used to distinguish normaland neoplastic breast tissues [2] This study looks at using the novel features of DynAMITe todemonstrate the concept of the lsquoX-ray biopsyrsquo mdash an in-vivo analysis of suspicious regions ofthe mammogram

2 Materials and methods21 Evaluation of the X-ray performance of the detector

The detective quantum efficiency (DQE) expresses the ability of an X-ray detector to transfer thesignal-to-noise ratio (SNR) from its input to output It is calculated as [3]

DQE( f ) =SNR2outSNR2in

=pMT F2( f )

Φ

KamiddotKa middotNNPS( f )

(21)

ndash 1 ndash

2012 JINST 7 C12005

Figure 1 a) Schematic of the ADXRD setup (not to scale) b) Photograph of the setup

where pMTF is the presampled modulation transfer function (ie the signal transfer) NNPS is thenormalized noise power spectrum Φ is the X-ray photon fluence (in X-raysmm2) and Ka is theAir Kerma at detector surface (in microGy) The X-ray performance evaluation of DynAMITe wasmade using TungstenAluminum (WAl) anodefiltration combination at 28 kV based on the IECstandard [3] The CMOS APS sensor was coupled to 200 microm Thallium-activated structured CesiumIodide (CsITl) scintillator

22 Angle dispersive X-ray diffraction (ADXRD) measurements

Using an excised breast tissue sample we made ADXRD measurements Briefly we recordedthe X-ray diffraction patterns of suspicious areas in the mammogram by measuring the angulardispersion of a pseudo-monochromatic beam [4]ndash[6] The diffraction pattern changes as a functionof the momentum transfer (x) which depends on the energy and scattering angle (θ )

x =1

2d=

sin(θ

2) (22)

where d is the spacing of atomic planes in the material and λ is the wavelength of X-rays Accord-ing to a previous study the diffraction pattern gives information about the molecular structure ofmaterials with healthy adipose tissue giving a single peak at around 12 nmminus1 while that of cancer-ous tissue exhibits a peak at 16 nmminus1 [2] The procedure was to capture mammograms that werethen annotated by a radiologist to indicate regions of suspected infiltrating disease These regionswere then subjected to ADXRD analysis This required the imaging system to be re-configuredinto a pencil beam of 2 mm diameter and DynAMITe to be put into a diffraction mode ie angularintegration around the incident beam and beam stop To get the required pseudo-monochromaticX-ray beam we applied balanced filtration We used an W anode X-ray source at 70 kV and wecaptured an average (over 5 frames) ADXRD image with 01 mm Samarium (Sm) filtration andanother one with 02 mm Tin (Sn) filtration [4]ndash[6] The difference image corresponds to 385 keVmean energy and 161 spectral width [6] Figure 1 shows the experimental setup used to scanthe breast biopsy samples and apply X-ray diffraction at specific points Custom-built softwarewritten in LabVIEW was used to control the filter and sample stepper motors by selecting specificpoints on the captured mammograms

ndash 2 ndash

2012 JINST 7 C12005

Figure 2 a) Presampled MTF b) average DQE of SP camera c) physical parameters of various mammo-graphic systems (WAl at 28 kV) [7]

3 Results and discussion

31 Detective quantum efficiency results

Figure 2a shows the horizontal vertical and average pMTF of DynAMITersquos SP camera The aver-age pMTF reaches the values 05 and 01 at around 3 and 8 lpmm respectively Figure 2b showsthe average DQE values of SP camera in the Ka range 436ndash1176 microGy

The table in figure 2c compares SP camerarsquos average pMTF at 50 level and maximum DQEat 05 lpmm (DQE(05)) to the performance of other digital X-ray detectors (using WAl at 28kV) [7] The spatial resolution of DynAMITe is not very high compared to the others Howeverthe MTF depends on the X-ray detection technology (Anrad is a direct conversion detector) thetype and thickness of the scintillator and the pixel pitch (ranging from 225 microm for Remote RadEyeHR to 85 microm for Anrad detector [7]) The physical performance is better described by the DQEbecause it takes into account the SNR contrast and noise of the detector In this case DynAMITeshows the highest low frequency DQE which corresponds to high X-ray detectability

32 X-ray biopsy results

Figure 3a shows a mammogram using DynAMITersquos SP camera Dark areas correspond to noncancerous tissue while white areas indicate possible presence of cancer (the image is inverted) Itmay be observed that the resolution of the image is relatively poor This is due to size of the focalspot of the X-ray source used (around 3 mm) and the geometric magnification of the measurementset-up (around 143) The use of micro focus Molybdenum (Mo) anode X-ray source will result inbetter image quality and lower dose because there is no need for balanced filtration As an examplewe applied ADXRD on five points in the mammogram which were selected by a radiologist fromprevious images of the sample taken on a Mo anode source with 70 microm focal spot This was made inorder to identify the extent of the infiltration of the tumour Figure 3b shows the respective ADXRDresults Combining the information from both mammographic and ADXRD results we can say thatpoint 1 is clearly within the tumour area and the scatter peak has broad shape (because tumourrsquoscomposition is less crystalline) and is approximately 15 nmminus1 However the mammogram cannot give sufficient information for the other four points From the ADXRD results we can see thatthe scatter peaks of points 3 to 5 are approximately 12 nmminus1 corresponding to fat tissue Point 2may correspond to the combination of fat and tumour tissue because the shape of the scatter peak

ndash 3 ndash

2012 JINST 7 C12005

Figure 3 a) Mammogram b) X-ray diffraction patterns of suspicious areas

is relatively broad Further analysis needs to be made in order to accurately identify the percentageof tumour in suspicious areas

4 Conclusions and future work

Preliminary results demonstrate that the SP camera is capable of capturing a large area imageand diffraction data in one examination Future versions of the detector will allow us to combinedata from both cameras to increase the dynamic range It should also be noted that different an-odefiltration combinations (such as micro focus MoMo) may result in decreased dose (no needfor balanced filtration) and better image quality We will use multivariate analysis based on varioussamples to predict the percentage of tumour in unknown areas

Acknowledgments

The authors wish to gratefully acknowledge the support of EPSRC for funding this work whichwas part of the Translation Grant MI-3 Plus (EPG0376712)

References

[1] T Graeve and GP Weckler High-resolution CMOS imaging detector Proc SPIE 4320 (2001) 68

[2] G Kidane et al X-ray scatter signatures for normal and neoplastic breast tissues Phys Med Biol 44(1999) 1791

[3] International Electrotechnical Commission Medical electrical equipment-characteristics of digitalX-ray imaging devices Part 1ndash2 Determination of the detective quantum efficiency-detectors used inmammography IEC 62220-1-2 Geneva Switzerland (2007)

[4] SE Bohndiek et al A CMOS active pixel sensor system for laboratory-based X-ray diffraction studiesof biological tissue Phys Med Biol 53 (2008) 655

[5] SE Bohndiek et al An Active Pixel sensor X-Ray Diffraction (APXRD) system for breast cancerdiagnosis Phys Med Biol 54 (2009) 3513

[6] AC Konstantinidis et al DynAMITe A prototype large area CMOS APS for breast cancer diagnosisusing X-ray diffraction measurements Proc SPIE 8313 (2012) 83135H-1

[7] AC Konstantinidis Evaluation of digital X-ray detectors for medical imaging applications PhDthesis University College London London UK (2011)

ndash 4 ndash

  • Introduction
  • Materials and methods
    • Evaluation of the X-ray performance of the detector
    • Angle dispersive X-ray diffraction (ADXRD) measurements
      • Results and discussion
        • Detective quantum efficiency results
        • X-ray biopsy results
          • Conclusions and future work

2012 JINST 7 C12005

PUBLISHED BY IOP PUBLISHING FOR SISSA MEDIALAB

RECEIVED September 13 2012ACCEPTED November 7 2012

PUBLISHED December 7 2012

14th INTERNATIONAL WORKSHOP ON RADIATION IMAGING DETECTORS1ndash5 JULY 2012FIGUEIRA DA FOZ PORTUGAL

A novel wafer-scale CMOS APS X-ray detector forbreast cancer diagnosis using X-ray diffractionstudies

A Konstantinidisa1 Y Zhenga D Philipb S Vinnicombec and R Spellera

aDepartment of Medical Physics and Bioengineering University College LondonLondon WC1E 6BT UK

bSchool of Engineering and Materials Science Queen Mary University of LondonLondon E1 4NS UK

cRadiology Department University of DundeeDundee DD1 4HN UK

E-mail anastasioskonstantinidisuclacuk

ABSTRACT The current study uses a novel large area (128 cm times131 cm) complementary metal-oxide-semiconductor (CMOS) active pixel sensor (APS) X-ray detector named Dynamic rangeAdjustable for Medical Imaging Technology (DynAMITe) for breast cancer diagnosis The de-tector consists of two geometrically superimposed grids a) 2560times 2624 fine-pitch grid of pixels(50 microm pitch) named Sub-Pixels (SP camera) for low intrinsic noise and high spatial resolu-tion and b) 1280times 1312 large-pitch grid of pixels (100 microm pitch) named Pixels (P camera) forhigh dynamic range X-ray performance characterization measurements show that the detectivequantum efficiency (DQE) of the SP camera is in the range 07ndash075 at low spatial frequenciesusing a tungsten (W) anode X-ray source at 28 kV Hence the detector is suitable for mammogra-phy Furthermore we used the SP camera to combine mammograms with angle dispersive X-raydiffraction (ADXRD) measurements in order to apply the X-ray biopsy concept in one examina-tion The results show that ADXRD technique indicates the presence of cancer in suspicious areason the mammogram Hence it could be used to determine the region affected by cancer and assistin planning surgery This study is the proof of concept that mammography and ADXRD can becombined in one examination

KEYWORDS X-ray radiography and digital radiography (DR) X-ray diffraction detectors

1Corresponding author

ccopy 2012 IOP Publishing Ltd and Sissa Medialab srl doi1010881748-0221712C12005

2012 JINST 7 C12005

Contents

1 Introduction 1

2 Materials and methods 121 Evaluation of the X-ray performance of the detector 122 Angle dispersive X-ray diffraction (ADXRD) measurements 2

3 Results and discussion 331 Detective quantum efficiency results 332 X-ray biopsy results 3

4 Conclusions and future work 4

1 IntroductionDigital X-ray detectors based on complementary metal-oxide-semiconductor (CMOS) active pixelsensor (APS) technology have been recently introduced in mammography [1] However conven-tional mammography does not always show the extent of tumour infiltration This study uses anovel large area CMOS APS X-ray detector to combine mammographic imaging with angle dis-persive X-ray diffraction (ADXRD) measurements to solve this problem

The Multidimensional Integrated Intelligent Imaging (MI-3) Plus consortium has developeda wafer scale (128 cm times131 cm) 3-T CMOS APS X-ray detector named DynAMITe (Dynamicrange Adjustable for Medical Imaging Technology) DynAMITe consists of two geometrically su-perimposed grids with two pixel pitches a) the Sub-Pixels (SP) camera (at 50 microm) for low noiseand high spatial resolution and b) the Pixels (P) camera (at 100 microm) for high dynamic range Thediodes of the two arrays are reset at different voltages to control the different depletion depthsof each diode Hence the imager can be used either as a 50 microm resolution or a 100 microm resolu-tion sensor or a combination of resolutions with extended dynamic range within the same pixelarray This study evaluates the X-ray performance of SP camera in mammography In particularthe presampled modulation transfer function (pMTF) and detective quantum efficiency (DQE) ofDynAMITersquos SP camera are compared to those of other mammographic detectors

Previous studies have demonstrated that X-ray diffraction can be used to distinguish normaland neoplastic breast tissues [2] This study looks at using the novel features of DynAMITe todemonstrate the concept of the lsquoX-ray biopsyrsquo mdash an in-vivo analysis of suspicious regions ofthe mammogram

2 Materials and methods21 Evaluation of the X-ray performance of the detector

The detective quantum efficiency (DQE) expresses the ability of an X-ray detector to transfer thesignal-to-noise ratio (SNR) from its input to output It is calculated as [3]

DQE( f ) =SNR2outSNR2in

=pMT F2( f )

Φ

KamiddotKa middotNNPS( f )

(21)

ndash 1 ndash

2012 JINST 7 C12005

Figure 1 a) Schematic of the ADXRD setup (not to scale) b) Photograph of the setup

where pMTF is the presampled modulation transfer function (ie the signal transfer) NNPS is thenormalized noise power spectrum Φ is the X-ray photon fluence (in X-raysmm2) and Ka is theAir Kerma at detector surface (in microGy) The X-ray performance evaluation of DynAMITe wasmade using TungstenAluminum (WAl) anodefiltration combination at 28 kV based on the IECstandard [3] The CMOS APS sensor was coupled to 200 microm Thallium-activated structured CesiumIodide (CsITl) scintillator

22 Angle dispersive X-ray diffraction (ADXRD) measurements

Using an excised breast tissue sample we made ADXRD measurements Briefly we recordedthe X-ray diffraction patterns of suspicious areas in the mammogram by measuring the angulardispersion of a pseudo-monochromatic beam [4]ndash[6] The diffraction pattern changes as a functionof the momentum transfer (x) which depends on the energy and scattering angle (θ )

x =1

2d=

sin(θ

2) (22)

where d is the spacing of atomic planes in the material and λ is the wavelength of X-rays Accord-ing to a previous study the diffraction pattern gives information about the molecular structure ofmaterials with healthy adipose tissue giving a single peak at around 12 nmminus1 while that of cancer-ous tissue exhibits a peak at 16 nmminus1 [2] The procedure was to capture mammograms that werethen annotated by a radiologist to indicate regions of suspected infiltrating disease These regionswere then subjected to ADXRD analysis This required the imaging system to be re-configuredinto a pencil beam of 2 mm diameter and DynAMITe to be put into a diffraction mode ie angularintegration around the incident beam and beam stop To get the required pseudo-monochromaticX-ray beam we applied balanced filtration We used an W anode X-ray source at 70 kV and wecaptured an average (over 5 frames) ADXRD image with 01 mm Samarium (Sm) filtration andanother one with 02 mm Tin (Sn) filtration [4]ndash[6] The difference image corresponds to 385 keVmean energy and 161 spectral width [6] Figure 1 shows the experimental setup used to scanthe breast biopsy samples and apply X-ray diffraction at specific points Custom-built softwarewritten in LabVIEW was used to control the filter and sample stepper motors by selecting specificpoints on the captured mammograms

ndash 2 ndash

2012 JINST 7 C12005

Figure 2 a) Presampled MTF b) average DQE of SP camera c) physical parameters of various mammo-graphic systems (WAl at 28 kV) [7]

3 Results and discussion

31 Detective quantum efficiency results

Figure 2a shows the horizontal vertical and average pMTF of DynAMITersquos SP camera The aver-age pMTF reaches the values 05 and 01 at around 3 and 8 lpmm respectively Figure 2b showsthe average DQE values of SP camera in the Ka range 436ndash1176 microGy

The table in figure 2c compares SP camerarsquos average pMTF at 50 level and maximum DQEat 05 lpmm (DQE(05)) to the performance of other digital X-ray detectors (using WAl at 28kV) [7] The spatial resolution of DynAMITe is not very high compared to the others Howeverthe MTF depends on the X-ray detection technology (Anrad is a direct conversion detector) thetype and thickness of the scintillator and the pixel pitch (ranging from 225 microm for Remote RadEyeHR to 85 microm for Anrad detector [7]) The physical performance is better described by the DQEbecause it takes into account the SNR contrast and noise of the detector In this case DynAMITeshows the highest low frequency DQE which corresponds to high X-ray detectability

32 X-ray biopsy results

Figure 3a shows a mammogram using DynAMITersquos SP camera Dark areas correspond to noncancerous tissue while white areas indicate possible presence of cancer (the image is inverted) Itmay be observed that the resolution of the image is relatively poor This is due to size of the focalspot of the X-ray source used (around 3 mm) and the geometric magnification of the measurementset-up (around 143) The use of micro focus Molybdenum (Mo) anode X-ray source will result inbetter image quality and lower dose because there is no need for balanced filtration As an examplewe applied ADXRD on five points in the mammogram which were selected by a radiologist fromprevious images of the sample taken on a Mo anode source with 70 microm focal spot This was made inorder to identify the extent of the infiltration of the tumour Figure 3b shows the respective ADXRDresults Combining the information from both mammographic and ADXRD results we can say thatpoint 1 is clearly within the tumour area and the scatter peak has broad shape (because tumourrsquoscomposition is less crystalline) and is approximately 15 nmminus1 However the mammogram cannot give sufficient information for the other four points From the ADXRD results we can see thatthe scatter peaks of points 3 to 5 are approximately 12 nmminus1 corresponding to fat tissue Point 2may correspond to the combination of fat and tumour tissue because the shape of the scatter peak

ndash 3 ndash

2012 JINST 7 C12005

Figure 3 a) Mammogram b) X-ray diffraction patterns of suspicious areas

is relatively broad Further analysis needs to be made in order to accurately identify the percentageof tumour in suspicious areas

4 Conclusions and future work

Preliminary results demonstrate that the SP camera is capable of capturing a large area imageand diffraction data in one examination Future versions of the detector will allow us to combinedata from both cameras to increase the dynamic range It should also be noted that different an-odefiltration combinations (such as micro focus MoMo) may result in decreased dose (no needfor balanced filtration) and better image quality We will use multivariate analysis based on varioussamples to predict the percentage of tumour in unknown areas

Acknowledgments

The authors wish to gratefully acknowledge the support of EPSRC for funding this work whichwas part of the Translation Grant MI-3 Plus (EPG0376712)

References

[1] T Graeve and GP Weckler High-resolution CMOS imaging detector Proc SPIE 4320 (2001) 68

[2] G Kidane et al X-ray scatter signatures for normal and neoplastic breast tissues Phys Med Biol 44(1999) 1791

[3] International Electrotechnical Commission Medical electrical equipment-characteristics of digitalX-ray imaging devices Part 1ndash2 Determination of the detective quantum efficiency-detectors used inmammography IEC 62220-1-2 Geneva Switzerland (2007)

[4] SE Bohndiek et al A CMOS active pixel sensor system for laboratory-based X-ray diffraction studiesof biological tissue Phys Med Biol 53 (2008) 655

[5] SE Bohndiek et al An Active Pixel sensor X-Ray Diffraction (APXRD) system for breast cancerdiagnosis Phys Med Biol 54 (2009) 3513

[6] AC Konstantinidis et al DynAMITe A prototype large area CMOS APS for breast cancer diagnosisusing X-ray diffraction measurements Proc SPIE 8313 (2012) 83135H-1

[7] AC Konstantinidis Evaluation of digital X-ray detectors for medical imaging applications PhDthesis University College London London UK (2011)

ndash 4 ndash

  • Introduction
  • Materials and methods
    • Evaluation of the X-ray performance of the detector
    • Angle dispersive X-ray diffraction (ADXRD) measurements
      • Results and discussion
        • Detective quantum efficiency results
        • X-ray biopsy results
          • Conclusions and future work

2012 JINST 7 C12005

Contents

1 Introduction 1

2 Materials and methods 121 Evaluation of the X-ray performance of the detector 122 Angle dispersive X-ray diffraction (ADXRD) measurements 2

3 Results and discussion 331 Detective quantum efficiency results 332 X-ray biopsy results 3

4 Conclusions and future work 4

1 IntroductionDigital X-ray detectors based on complementary metal-oxide-semiconductor (CMOS) active pixelsensor (APS) technology have been recently introduced in mammography [1] However conven-tional mammography does not always show the extent of tumour infiltration This study uses anovel large area CMOS APS X-ray detector to combine mammographic imaging with angle dis-persive X-ray diffraction (ADXRD) measurements to solve this problem

The Multidimensional Integrated Intelligent Imaging (MI-3) Plus consortium has developeda wafer scale (128 cm times131 cm) 3-T CMOS APS X-ray detector named DynAMITe (Dynamicrange Adjustable for Medical Imaging Technology) DynAMITe consists of two geometrically su-perimposed grids with two pixel pitches a) the Sub-Pixels (SP) camera (at 50 microm) for low noiseand high spatial resolution and b) the Pixels (P) camera (at 100 microm) for high dynamic range Thediodes of the two arrays are reset at different voltages to control the different depletion depthsof each diode Hence the imager can be used either as a 50 microm resolution or a 100 microm resolu-tion sensor or a combination of resolutions with extended dynamic range within the same pixelarray This study evaluates the X-ray performance of SP camera in mammography In particularthe presampled modulation transfer function (pMTF) and detective quantum efficiency (DQE) ofDynAMITersquos SP camera are compared to those of other mammographic detectors

Previous studies have demonstrated that X-ray diffraction can be used to distinguish normaland neoplastic breast tissues [2] This study looks at using the novel features of DynAMITe todemonstrate the concept of the lsquoX-ray biopsyrsquo mdash an in-vivo analysis of suspicious regions ofthe mammogram

2 Materials and methods21 Evaluation of the X-ray performance of the detector

The detective quantum efficiency (DQE) expresses the ability of an X-ray detector to transfer thesignal-to-noise ratio (SNR) from its input to output It is calculated as [3]

DQE( f ) =SNR2outSNR2in

=pMT F2( f )

Φ

KamiddotKa middotNNPS( f )

(21)

ndash 1 ndash

2012 JINST 7 C12005

Figure 1 a) Schematic of the ADXRD setup (not to scale) b) Photograph of the setup

where pMTF is the presampled modulation transfer function (ie the signal transfer) NNPS is thenormalized noise power spectrum Φ is the X-ray photon fluence (in X-raysmm2) and Ka is theAir Kerma at detector surface (in microGy) The X-ray performance evaluation of DynAMITe wasmade using TungstenAluminum (WAl) anodefiltration combination at 28 kV based on the IECstandard [3] The CMOS APS sensor was coupled to 200 microm Thallium-activated structured CesiumIodide (CsITl) scintillator

22 Angle dispersive X-ray diffraction (ADXRD) measurements

Using an excised breast tissue sample we made ADXRD measurements Briefly we recordedthe X-ray diffraction patterns of suspicious areas in the mammogram by measuring the angulardispersion of a pseudo-monochromatic beam [4]ndash[6] The diffraction pattern changes as a functionof the momentum transfer (x) which depends on the energy and scattering angle (θ )

x =1

2d=

sin(θ

2) (22)

where d is the spacing of atomic planes in the material and λ is the wavelength of X-rays Accord-ing to a previous study the diffraction pattern gives information about the molecular structure ofmaterials with healthy adipose tissue giving a single peak at around 12 nmminus1 while that of cancer-ous tissue exhibits a peak at 16 nmminus1 [2] The procedure was to capture mammograms that werethen annotated by a radiologist to indicate regions of suspected infiltrating disease These regionswere then subjected to ADXRD analysis This required the imaging system to be re-configuredinto a pencil beam of 2 mm diameter and DynAMITe to be put into a diffraction mode ie angularintegration around the incident beam and beam stop To get the required pseudo-monochromaticX-ray beam we applied balanced filtration We used an W anode X-ray source at 70 kV and wecaptured an average (over 5 frames) ADXRD image with 01 mm Samarium (Sm) filtration andanother one with 02 mm Tin (Sn) filtration [4]ndash[6] The difference image corresponds to 385 keVmean energy and 161 spectral width [6] Figure 1 shows the experimental setup used to scanthe breast biopsy samples and apply X-ray diffraction at specific points Custom-built softwarewritten in LabVIEW was used to control the filter and sample stepper motors by selecting specificpoints on the captured mammograms

ndash 2 ndash

2012 JINST 7 C12005

Figure 2 a) Presampled MTF b) average DQE of SP camera c) physical parameters of various mammo-graphic systems (WAl at 28 kV) [7]

3 Results and discussion

31 Detective quantum efficiency results

Figure 2a shows the horizontal vertical and average pMTF of DynAMITersquos SP camera The aver-age pMTF reaches the values 05 and 01 at around 3 and 8 lpmm respectively Figure 2b showsthe average DQE values of SP camera in the Ka range 436ndash1176 microGy

The table in figure 2c compares SP camerarsquos average pMTF at 50 level and maximum DQEat 05 lpmm (DQE(05)) to the performance of other digital X-ray detectors (using WAl at 28kV) [7] The spatial resolution of DynAMITe is not very high compared to the others Howeverthe MTF depends on the X-ray detection technology (Anrad is a direct conversion detector) thetype and thickness of the scintillator and the pixel pitch (ranging from 225 microm for Remote RadEyeHR to 85 microm for Anrad detector [7]) The physical performance is better described by the DQEbecause it takes into account the SNR contrast and noise of the detector In this case DynAMITeshows the highest low frequency DQE which corresponds to high X-ray detectability

32 X-ray biopsy results

Figure 3a shows a mammogram using DynAMITersquos SP camera Dark areas correspond to noncancerous tissue while white areas indicate possible presence of cancer (the image is inverted) Itmay be observed that the resolution of the image is relatively poor This is due to size of the focalspot of the X-ray source used (around 3 mm) and the geometric magnification of the measurementset-up (around 143) The use of micro focus Molybdenum (Mo) anode X-ray source will result inbetter image quality and lower dose because there is no need for balanced filtration As an examplewe applied ADXRD on five points in the mammogram which were selected by a radiologist fromprevious images of the sample taken on a Mo anode source with 70 microm focal spot This was made inorder to identify the extent of the infiltration of the tumour Figure 3b shows the respective ADXRDresults Combining the information from both mammographic and ADXRD results we can say thatpoint 1 is clearly within the tumour area and the scatter peak has broad shape (because tumourrsquoscomposition is less crystalline) and is approximately 15 nmminus1 However the mammogram cannot give sufficient information for the other four points From the ADXRD results we can see thatthe scatter peaks of points 3 to 5 are approximately 12 nmminus1 corresponding to fat tissue Point 2may correspond to the combination of fat and tumour tissue because the shape of the scatter peak

ndash 3 ndash

2012 JINST 7 C12005

Figure 3 a) Mammogram b) X-ray diffraction patterns of suspicious areas

is relatively broad Further analysis needs to be made in order to accurately identify the percentageof tumour in suspicious areas

4 Conclusions and future work

Preliminary results demonstrate that the SP camera is capable of capturing a large area imageand diffraction data in one examination Future versions of the detector will allow us to combinedata from both cameras to increase the dynamic range It should also be noted that different an-odefiltration combinations (such as micro focus MoMo) may result in decreased dose (no needfor balanced filtration) and better image quality We will use multivariate analysis based on varioussamples to predict the percentage of tumour in unknown areas

Acknowledgments

The authors wish to gratefully acknowledge the support of EPSRC for funding this work whichwas part of the Translation Grant MI-3 Plus (EPG0376712)

References

[1] T Graeve and GP Weckler High-resolution CMOS imaging detector Proc SPIE 4320 (2001) 68

[2] G Kidane et al X-ray scatter signatures for normal and neoplastic breast tissues Phys Med Biol 44(1999) 1791

[3] International Electrotechnical Commission Medical electrical equipment-characteristics of digitalX-ray imaging devices Part 1ndash2 Determination of the detective quantum efficiency-detectors used inmammography IEC 62220-1-2 Geneva Switzerland (2007)

[4] SE Bohndiek et al A CMOS active pixel sensor system for laboratory-based X-ray diffraction studiesof biological tissue Phys Med Biol 53 (2008) 655

[5] SE Bohndiek et al An Active Pixel sensor X-Ray Diffraction (APXRD) system for breast cancerdiagnosis Phys Med Biol 54 (2009) 3513

[6] AC Konstantinidis et al DynAMITe A prototype large area CMOS APS for breast cancer diagnosisusing X-ray diffraction measurements Proc SPIE 8313 (2012) 83135H-1

[7] AC Konstantinidis Evaluation of digital X-ray detectors for medical imaging applications PhDthesis University College London London UK (2011)

ndash 4 ndash

  • Introduction
  • Materials and methods
    • Evaluation of the X-ray performance of the detector
    • Angle dispersive X-ray diffraction (ADXRD) measurements
      • Results and discussion
        • Detective quantum efficiency results
        • X-ray biopsy results
          • Conclusions and future work

2012 JINST 7 C12005

Figure 1 a) Schematic of the ADXRD setup (not to scale) b) Photograph of the setup

where pMTF is the presampled modulation transfer function (ie the signal transfer) NNPS is thenormalized noise power spectrum Φ is the X-ray photon fluence (in X-raysmm2) and Ka is theAir Kerma at detector surface (in microGy) The X-ray performance evaluation of DynAMITe wasmade using TungstenAluminum (WAl) anodefiltration combination at 28 kV based on the IECstandard [3] The CMOS APS sensor was coupled to 200 microm Thallium-activated structured CesiumIodide (CsITl) scintillator

22 Angle dispersive X-ray diffraction (ADXRD) measurements

Using an excised breast tissue sample we made ADXRD measurements Briefly we recordedthe X-ray diffraction patterns of suspicious areas in the mammogram by measuring the angulardispersion of a pseudo-monochromatic beam [4]ndash[6] The diffraction pattern changes as a functionof the momentum transfer (x) which depends on the energy and scattering angle (θ )

x =1

2d=

sin(θ

2) (22)

where d is the spacing of atomic planes in the material and λ is the wavelength of X-rays Accord-ing to a previous study the diffraction pattern gives information about the molecular structure ofmaterials with healthy adipose tissue giving a single peak at around 12 nmminus1 while that of cancer-ous tissue exhibits a peak at 16 nmminus1 [2] The procedure was to capture mammograms that werethen annotated by a radiologist to indicate regions of suspected infiltrating disease These regionswere then subjected to ADXRD analysis This required the imaging system to be re-configuredinto a pencil beam of 2 mm diameter and DynAMITe to be put into a diffraction mode ie angularintegration around the incident beam and beam stop To get the required pseudo-monochromaticX-ray beam we applied balanced filtration We used an W anode X-ray source at 70 kV and wecaptured an average (over 5 frames) ADXRD image with 01 mm Samarium (Sm) filtration andanother one with 02 mm Tin (Sn) filtration [4]ndash[6] The difference image corresponds to 385 keVmean energy and 161 spectral width [6] Figure 1 shows the experimental setup used to scanthe breast biopsy samples and apply X-ray diffraction at specific points Custom-built softwarewritten in LabVIEW was used to control the filter and sample stepper motors by selecting specificpoints on the captured mammograms

ndash 2 ndash

2012 JINST 7 C12005

Figure 2 a) Presampled MTF b) average DQE of SP camera c) physical parameters of various mammo-graphic systems (WAl at 28 kV) [7]

3 Results and discussion

31 Detective quantum efficiency results

Figure 2a shows the horizontal vertical and average pMTF of DynAMITersquos SP camera The aver-age pMTF reaches the values 05 and 01 at around 3 and 8 lpmm respectively Figure 2b showsthe average DQE values of SP camera in the Ka range 436ndash1176 microGy

The table in figure 2c compares SP camerarsquos average pMTF at 50 level and maximum DQEat 05 lpmm (DQE(05)) to the performance of other digital X-ray detectors (using WAl at 28kV) [7] The spatial resolution of DynAMITe is not very high compared to the others Howeverthe MTF depends on the X-ray detection technology (Anrad is a direct conversion detector) thetype and thickness of the scintillator and the pixel pitch (ranging from 225 microm for Remote RadEyeHR to 85 microm for Anrad detector [7]) The physical performance is better described by the DQEbecause it takes into account the SNR contrast and noise of the detector In this case DynAMITeshows the highest low frequency DQE which corresponds to high X-ray detectability

32 X-ray biopsy results

Figure 3a shows a mammogram using DynAMITersquos SP camera Dark areas correspond to noncancerous tissue while white areas indicate possible presence of cancer (the image is inverted) Itmay be observed that the resolution of the image is relatively poor This is due to size of the focalspot of the X-ray source used (around 3 mm) and the geometric magnification of the measurementset-up (around 143) The use of micro focus Molybdenum (Mo) anode X-ray source will result inbetter image quality and lower dose because there is no need for balanced filtration As an examplewe applied ADXRD on five points in the mammogram which were selected by a radiologist fromprevious images of the sample taken on a Mo anode source with 70 microm focal spot This was made inorder to identify the extent of the infiltration of the tumour Figure 3b shows the respective ADXRDresults Combining the information from both mammographic and ADXRD results we can say thatpoint 1 is clearly within the tumour area and the scatter peak has broad shape (because tumourrsquoscomposition is less crystalline) and is approximately 15 nmminus1 However the mammogram cannot give sufficient information for the other four points From the ADXRD results we can see thatthe scatter peaks of points 3 to 5 are approximately 12 nmminus1 corresponding to fat tissue Point 2may correspond to the combination of fat and tumour tissue because the shape of the scatter peak

ndash 3 ndash

2012 JINST 7 C12005

Figure 3 a) Mammogram b) X-ray diffraction patterns of suspicious areas

is relatively broad Further analysis needs to be made in order to accurately identify the percentageof tumour in suspicious areas

4 Conclusions and future work

Preliminary results demonstrate that the SP camera is capable of capturing a large area imageand diffraction data in one examination Future versions of the detector will allow us to combinedata from both cameras to increase the dynamic range It should also be noted that different an-odefiltration combinations (such as micro focus MoMo) may result in decreased dose (no needfor balanced filtration) and better image quality We will use multivariate analysis based on varioussamples to predict the percentage of tumour in unknown areas

Acknowledgments

The authors wish to gratefully acknowledge the support of EPSRC for funding this work whichwas part of the Translation Grant MI-3 Plus (EPG0376712)

References

[1] T Graeve and GP Weckler High-resolution CMOS imaging detector Proc SPIE 4320 (2001) 68

[2] G Kidane et al X-ray scatter signatures for normal and neoplastic breast tissues Phys Med Biol 44(1999) 1791

[3] International Electrotechnical Commission Medical electrical equipment-characteristics of digitalX-ray imaging devices Part 1ndash2 Determination of the detective quantum efficiency-detectors used inmammography IEC 62220-1-2 Geneva Switzerland (2007)

[4] SE Bohndiek et al A CMOS active pixel sensor system for laboratory-based X-ray diffraction studiesof biological tissue Phys Med Biol 53 (2008) 655

[5] SE Bohndiek et al An Active Pixel sensor X-Ray Diffraction (APXRD) system for breast cancerdiagnosis Phys Med Biol 54 (2009) 3513

[6] AC Konstantinidis et al DynAMITe A prototype large area CMOS APS for breast cancer diagnosisusing X-ray diffraction measurements Proc SPIE 8313 (2012) 83135H-1

[7] AC Konstantinidis Evaluation of digital X-ray detectors for medical imaging applications PhDthesis University College London London UK (2011)

ndash 4 ndash

  • Introduction
  • Materials and methods
    • Evaluation of the X-ray performance of the detector
    • Angle dispersive X-ray diffraction (ADXRD) measurements
      • Results and discussion
        • Detective quantum efficiency results
        • X-ray biopsy results
          • Conclusions and future work

2012 JINST 7 C12005

Figure 2 a) Presampled MTF b) average DQE of SP camera c) physical parameters of various mammo-graphic systems (WAl at 28 kV) [7]

3 Results and discussion

31 Detective quantum efficiency results

Figure 2a shows the horizontal vertical and average pMTF of DynAMITersquos SP camera The aver-age pMTF reaches the values 05 and 01 at around 3 and 8 lpmm respectively Figure 2b showsthe average DQE values of SP camera in the Ka range 436ndash1176 microGy

The table in figure 2c compares SP camerarsquos average pMTF at 50 level and maximum DQEat 05 lpmm (DQE(05)) to the performance of other digital X-ray detectors (using WAl at 28kV) [7] The spatial resolution of DynAMITe is not very high compared to the others Howeverthe MTF depends on the X-ray detection technology (Anrad is a direct conversion detector) thetype and thickness of the scintillator and the pixel pitch (ranging from 225 microm for Remote RadEyeHR to 85 microm for Anrad detector [7]) The physical performance is better described by the DQEbecause it takes into account the SNR contrast and noise of the detector In this case DynAMITeshows the highest low frequency DQE which corresponds to high X-ray detectability

32 X-ray biopsy results

Figure 3a shows a mammogram using DynAMITersquos SP camera Dark areas correspond to noncancerous tissue while white areas indicate possible presence of cancer (the image is inverted) Itmay be observed that the resolution of the image is relatively poor This is due to size of the focalspot of the X-ray source used (around 3 mm) and the geometric magnification of the measurementset-up (around 143) The use of micro focus Molybdenum (Mo) anode X-ray source will result inbetter image quality and lower dose because there is no need for balanced filtration As an examplewe applied ADXRD on five points in the mammogram which were selected by a radiologist fromprevious images of the sample taken on a Mo anode source with 70 microm focal spot This was made inorder to identify the extent of the infiltration of the tumour Figure 3b shows the respective ADXRDresults Combining the information from both mammographic and ADXRD results we can say thatpoint 1 is clearly within the tumour area and the scatter peak has broad shape (because tumourrsquoscomposition is less crystalline) and is approximately 15 nmminus1 However the mammogram cannot give sufficient information for the other four points From the ADXRD results we can see thatthe scatter peaks of points 3 to 5 are approximately 12 nmminus1 corresponding to fat tissue Point 2may correspond to the combination of fat and tumour tissue because the shape of the scatter peak

ndash 3 ndash

2012 JINST 7 C12005

Figure 3 a) Mammogram b) X-ray diffraction patterns of suspicious areas

is relatively broad Further analysis needs to be made in order to accurately identify the percentageof tumour in suspicious areas

4 Conclusions and future work

Preliminary results demonstrate that the SP camera is capable of capturing a large area imageand diffraction data in one examination Future versions of the detector will allow us to combinedata from both cameras to increase the dynamic range It should also be noted that different an-odefiltration combinations (such as micro focus MoMo) may result in decreased dose (no needfor balanced filtration) and better image quality We will use multivariate analysis based on varioussamples to predict the percentage of tumour in unknown areas

Acknowledgments

The authors wish to gratefully acknowledge the support of EPSRC for funding this work whichwas part of the Translation Grant MI-3 Plus (EPG0376712)

References

[1] T Graeve and GP Weckler High-resolution CMOS imaging detector Proc SPIE 4320 (2001) 68

[2] G Kidane et al X-ray scatter signatures for normal and neoplastic breast tissues Phys Med Biol 44(1999) 1791

[3] International Electrotechnical Commission Medical electrical equipment-characteristics of digitalX-ray imaging devices Part 1ndash2 Determination of the detective quantum efficiency-detectors used inmammography IEC 62220-1-2 Geneva Switzerland (2007)

[4] SE Bohndiek et al A CMOS active pixel sensor system for laboratory-based X-ray diffraction studiesof biological tissue Phys Med Biol 53 (2008) 655

[5] SE Bohndiek et al An Active Pixel sensor X-Ray Diffraction (APXRD) system for breast cancerdiagnosis Phys Med Biol 54 (2009) 3513

[6] AC Konstantinidis et al DynAMITe A prototype large area CMOS APS for breast cancer diagnosisusing X-ray diffraction measurements Proc SPIE 8313 (2012) 83135H-1

[7] AC Konstantinidis Evaluation of digital X-ray detectors for medical imaging applications PhDthesis University College London London UK (2011)

ndash 4 ndash

  • Introduction
  • Materials and methods
    • Evaluation of the X-ray performance of the detector
    • Angle dispersive X-ray diffraction (ADXRD) measurements
      • Results and discussion
        • Detective quantum efficiency results
        • X-ray biopsy results
          • Conclusions and future work

2012 JINST 7 C12005

Figure 3 a) Mammogram b) X-ray diffraction patterns of suspicious areas

is relatively broad Further analysis needs to be made in order to accurately identify the percentageof tumour in suspicious areas

4 Conclusions and future work

Preliminary results demonstrate that the SP camera is capable of capturing a large area imageand diffraction data in one examination Future versions of the detector will allow us to combinedata from both cameras to increase the dynamic range It should also be noted that different an-odefiltration combinations (such as micro focus MoMo) may result in decreased dose (no needfor balanced filtration) and better image quality We will use multivariate analysis based on varioussamples to predict the percentage of tumour in unknown areas

Acknowledgments

The authors wish to gratefully acknowledge the support of EPSRC for funding this work whichwas part of the Translation Grant MI-3 Plus (EPG0376712)

References

[1] T Graeve and GP Weckler High-resolution CMOS imaging detector Proc SPIE 4320 (2001) 68

[2] G Kidane et al X-ray scatter signatures for normal and neoplastic breast tissues Phys Med Biol 44(1999) 1791

[3] International Electrotechnical Commission Medical electrical equipment-characteristics of digitalX-ray imaging devices Part 1ndash2 Determination of the detective quantum efficiency-detectors used inmammography IEC 62220-1-2 Geneva Switzerland (2007)

[4] SE Bohndiek et al A CMOS active pixel sensor system for laboratory-based X-ray diffraction studiesof biological tissue Phys Med Biol 53 (2008) 655

[5] SE Bohndiek et al An Active Pixel sensor X-Ray Diffraction (APXRD) system for breast cancerdiagnosis Phys Med Biol 54 (2009) 3513

[6] AC Konstantinidis et al DynAMITe A prototype large area CMOS APS for breast cancer diagnosisusing X-ray diffraction measurements Proc SPIE 8313 (2012) 83135H-1

[7] AC Konstantinidis Evaluation of digital X-ray detectors for medical imaging applications PhDthesis University College London London UK (2011)

ndash 4 ndash

  • Introduction
  • Materials and methods
    • Evaluation of the X-ray performance of the detector
    • Angle dispersive X-ray diffraction (ADXRD) measurements
      • Results and discussion
        • Detective quantum efficiency results
        • X-ray biopsy results
          • Conclusions and future work