Graham M. Treece
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Graham M. Treece has a total of 133 co-authors. Here are the top co-authors by number of shared publications:
- Richard W. Prager
Department of Engineering, University of Cambridge, Cambridge CB2 1PZ, United Kingdom
13 shared publications - Andrew H. Gee
University of Cambridge, Department of Engineering, Trumpington Street, Cambridge CB2 1PZ, UK
9 shared publications - A. H. Gee
University of Cambridge Department of Engineering, Trumpington Street, Cambridge CB2 1PZ, UK
4 shared publications - Andrew H. Gee
University of Cambridge, Department of Engineering, Trumpington Street, Cambridge CB2 1PZ, UK
4 shared publications - Kenneth E. S. Poole
NIHR Biomedical Research Centre, Cambridge University Hospitals NHS Foundation Trust, UK NIHR Cambridge Biomedical Research Centre, Hills Rd, Cambridge CB2 0QQ, UK
4 shared publications Andrew H. Gee
Department of Engineering, University of Cambridge, Cambridge, UK
3 shared publications
Latest Publications
- 8 April 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Image Processing
- Vol. 31 (10577149) , 3151-3165
Real Image Denoising With a Locally-Adaptive Bitonic Filter
Image noise removal is a common problem with many proposed solutions. The current standard is set by learning-based approaches, however these are not appropriate in all scenarios, perhaps due to lack of training data or the need for predictability in novel circumstances. The bitonic filter is a non-learning-based filter for removing noise from signals, with a mathematical morphology (ranking) framework in which the signal is postulated to be locally bitonic (having only one minimum or maximum) over some domain of finite extent. A novel version of this filter is developed in this paper, with a domain that is locally-adaptive to the signal, and other adjustments to allow application to real image sensor noise. These lead to significant improvements in noise reduction performance at no cost to processing times. The new bitonic filter performs better than the block-matching 3D filter for high levels of additive white Gaussian noise. It also surpasses this and other more recent non-learning-based filters for two public data sets containing real image noise at various levels. This is despite an additional adjustment to the block-matching filter, which leads to significantly better performance than has previously been cited on these data sets. The new bitonic filter has a signal-to-noise ratio 2.4 dB lower than the best learning-based techniques when they are optimally trained. However, the performance gap is closed completely when these techniques are trained on data sets not directly related to the benchmark data. This demonstrates what can be achieved with a predictable, explainable, entirely local technique, which makes no assumptions of repeating patterns either within an image or across images, and hence creates residual images which are well behaved even in very high noise. Since the filter does not require training, it can still be used in situations where training is either difficult or inappropriate.
- 8 April 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Image Processing
- Vol. 31 (10577149) , 3151-3165
Real Image Denoising With a Locally-Adaptive Bitonic Filter
Image noise removal is a common problem with many proposed solutions. The current standard is set by learning-based approaches, however these are not appropriate in all scenarios, perhaps due to lack of training data or the need for predictability in novel circumstances. The bitonic filter is a non-learning-based filter for removing noise from signals, with a mathematical morphology (ranking) framework in which the signal is postulated to be locally bitonic (having only one minimum or maximum) over some domain of finite extent. A novel version of this filter is developed in this paper, with a domain that is locally-adaptive to the signal, and other adjustments to allow application to real image sensor noise. These lead to significant improvements in noise reduction performance at no cost to processing times. The new bitonic filter performs better than the block-matching 3D filter for high levels of additive white Gaussian noise. It also surpasses this and other more recent non-learning-based filters for two public data sets containing real image noise at various levels. This is despite an additional adjustment to the block-matching filter, which leads to significantly better performance than has previously been cited on these data sets. The new bitonic filter has a signal-to-noise ratio 2.4 dB lower than the best learning-based techniques when they are optimally trained. However, the performance gap is closed completely when these techniques are trained on data sets not directly related to the benchmark data. This demonstrates what can be achieved with a predictable, explainable, entirely local technique, which makes no assumptions of repeating patterns either within an image or across images, and hence creates residual images which are well behaved even in very high noise. Since the filter does not require training, it can still be used in situations where training is either difficult or inappropriate.
- 12 April 2021
- Published by Radiological Society of North America (RSNA) in Radiology
- Vol. 299 (3) , 649-659
Quantitative Three-dimensional Assessment of Knee Joint Space Width from Weight-bearing CT
Background Imaging of structural disease in osteoarthritis has traditionally relied on MRI and radiography. Joint space mapping (JSM) can be used to quantitatively map joint space width (JSW) in three dimensions from CT images. Purpose To demonstrate the reproducibility, repeatability, and feasibility of JSM of the knee using weight-bearing CT images. Materials and Methods Two convenience samples of weight-bearing CT images of left and right knees with radiographic Kellgren-Lawrence grades (KLGs) less than or equal to 2 were acquired from 2014 to 2018 and were analyzed retrospectively with JSM to deliver three-dimensional JSW maps. For reproducibility, images of three sets of knees were used for novice training, and then the JSM output was compared against an expert’s assessment. JSM was also performed on 2-week follow-up images in the second cohort, yielding three-dimensional JSW difference maps for repeatability. Statistical parametric mapping was performed on all knee imaging data (KLG, 0–4) to show the feasibility of a surface-based analysis in three dimensions. Results Reproducibility (in 20 individuals; mean age, 58 years ± 7 [standard deviation]; mean body mass index, 28 kg/m2 ± 6; 14 women) and repeatability (in nine individuals; mean age, 53 years ± 6; mean body mass index, 26 kg/m2 ± 4; seven women) reached their lowest performance at a smallest detectable difference less than ±0.1 mm in the central medial tibiofemoral joint space for individuals without radiographically demonstrated disease. The average root mean square coefficient of variation was less than 5% across all groups. Statistical parametric mapping (33 individuals; mean age, 57 years ± 7; mean body mass index, 27 kg/m2 ± 6; 23 women) showed that the central-to-posterior medial joint space was significantly narrower by 0.5 mm for each incremental increase in the KLG (threshold P < .05). One knee (KLG, 2) demonstrated a baseline versus 24-month change in its three-dimensional JSW distribution that was beyond the smallest detectable difference across the lateral joint space. Conclusion Joint space mapping of the knee using weight-bearing CT images is feasible, demonstrating a relationship between the three-dimensional joint space width distribution and structural joint disease. It is reliably learned by novice users, can be personalized for disease phenotypes, and can be used to achieve a smallest detectable difference that is at least 50% smaller than that reported to be achieved at the highest performance level in radiography. © RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Roemer in this issue. Download as PowerPoint
- 12 April 2021
- Published by Radiological Society of North America (RSNA) in Radiology
- Vol. 299 (3) , 649-659
Quantitative Three-dimensional Assessment of Knee Joint Space Width from Weight-bearing CT
Background Imaging of structural disease in osteoarthritis has traditionally relied on MRI and radiography. Joint space mapping (JSM) can be used to quantitatively map joint space width (JSW) in three dimensions from CT images. Purpose To demonstrate the reproducibility, repeatability, and feasibility of JSM of the knee using weight-bearing CT images. Materials and Methods Two convenience samples of weight-bearing CT images of left and right knees with radiographic Kellgren-Lawrence grades (KLGs) less than or equal to 2 were acquired from 2014 to 2018 and were analyzed retrospectively with JSM to deliver three-dimensional JSW maps. For reproducibility, images of three sets of knees were used for novice training, and then the JSM output was compared against an expert’s assessment. JSM was also performed on 2-week follow-up images in the second cohort, yielding three-dimensional JSW difference maps for repeatability. Statistical parametric mapping was performed on all knee imaging data (KLG, 0–4) to show the feasibility of a surface-based analysis in three dimensions. Results Reproducibility (in 20 individuals; mean age, 58 years ± 7 [standard deviation]; mean body mass index, 28 kg/m2 ± 6; 14 women) and repeatability (in nine individuals; mean age, 53 years ± 6; mean body mass index, 26 kg/m2 ± 4; seven women) reached their lowest performance at a smallest detectable difference less than ±0.1 mm in the central medial tibiofemoral joint space for individuals without radiographically demonstrated disease. The average root mean square coefficient of variation was less than 5% across all groups. Statistical parametric mapping (33 individuals; mean age, 57 years ± 7; mean body mass index, 27 kg/m2 ± 6; 23 women) showed that the central-to-posterior medial joint space was significantly narrower by 0.5 mm for each incremental increase in the KLG (threshold P < .05). One knee (KLG, 2) demonstrated a baseline versus 24-month change in its three-dimensional JSW distribution that was beyond the smallest detectable difference across the lateral joint space. Conclusion Joint space mapping of the knee using weight-bearing CT images is feasible, demonstrating a relationship between the three-dimensional joint space width distribution and structural joint disease. It is reliably learned by novice users, can be personalized for disease phenotypes, and can be used to achieve a smallest detectable difference that is at least 50% smaller than that reported to be achieved at the highest performance level in radiography. © RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Roemer in this issue. Download as PowerPoint
- Open Access
- 5 March 2020
- Published by Springer Nature in Scientific Reports
- Vol. 10 (1) , 1-11
Quantitative 3D imaging parameters improve prediction of hip osteoarthritis outcome
Osteoarthritis is an increasingly important health problem for which the main treatment remains joint replacement. Therapy developments have been hampered by a lack of biomarkers that can reliably predict disease, while 2D radiographs interpreted by human observers are still the gold standard for clinical trial imaging assessment. We propose a 3D approach using computed tomography—a fast, readily available clinical technique—that can be applied in the assessment of osteoarthritis using a new quantitative 3D analysis technique called joint space mapping (JSM). We demonstrate the application of JSM at the hip in 263 healthy older adults from the AGES-Reykjavík cohort, examining relationships between 3D joint space width, 3D joint shape, and future joint replacement. Using JSM, statistical shape modelling, and statistical parametric mapping, we show an 18% improvement in prediction of joint replacement using 3D metrics combined with radiographic Kellgren & Lawrence grade (AUC 0.86) over the existing 2D FDA-approved gold standard of minimum 2D joint space width (AUC 0.73). We also show that assessment of joint asymmetry can reveal significant differences between individuals destined for joint replacement versus controls at regions of the joint that are not captured by radiographs. This technique is immediately implementable with standard imaging technologies.
- 5 March 2020
- Published by Springer Nature in Scientific Reports
- Vol. 10 (1) , 1-11
Open AccessQuantitative 3D imaging parameters improve prediction of hip osteoarthritis outcome
Osteoarthritis is an increasingly important health problem for which the main treatment remains joint replacement. Therapy developments have been hampered by a lack of biomarkers that can reliably predict disease, while 2D radiographs interpreted by human observers are still the gold standard for clinical trial imaging assessment. We propose a 3D approach using computed tomography—a fast, readily available clinical technique—that can be applied in the assessment of osteoarthritis using a new quantitative 3D analysis technique called joint space mapping (JSM). We demonstrate the application of JSM at the hip in 263 healthy older adults from the AGES-Reykjavík cohort, examining relationships between 3D joint space width, 3D joint shape, and future joint replacement. Using JSM, statistical shape modelling, and statistical parametric mapping, we show an 18% improvement in prediction of joint replacement using 3D metrics combined with radiographic Kellgren & Lawrence grade (AUC 0.86) over the existing 2D FDA-approved gold standard of minimum 2D joint space width (AUC 0.73). We also show that assessment of joint asymmetry can reveal significant differences between individuals destined for joint replacement versus controls at regions of the joint that are not captured by radiographs. This technique is immediately implementable with standard imaging technologies.
- 24 January 2020
- Published by Springer Nature
- p. 327-338
Automatic Detection and Visualisation of Metastatic Bone Disease
- 24 January 2020
- Published by Springer Nature
- p. 327-338
Automatic Detection and Visualisation of Metastatic Bone Disease
- 7 August 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Image Processing
- Vol. 29 (10577149) , 336-350
Morphology-Based Noise Reduction: Structural Variation and Thresholding in the Bitonic Filter
The bitonic filter was recently developed to embody the novel concept of signal bitonicity (one local extremum within a set range) to differentiate from noise, by use of data ranking and linear operators. For processing images, the spatial extent was locally constrained to a fixed circular mask. Since structure in natural images varies, a novel structurally varying bitonic filter is presented, which locally adapts the mask, without following patterns in the noise. This new filter includes novel robust structurally varying morphological operations, with efficient implementations, and a novel formulation of non-iterative directional Gaussian filtering. Data thresholds are also integrated with the morphological operations, increasing noise reduction for low noise, and enabling a multi-resolution framework for high noise levels. The structurally varying bitonic filter is presented without presuming prior knowledge of morphological filtering, and compared to high-performance linear noise-reduction filters, to set this novel concept in context. These are tested over a wide range of noise levels, on a fairly broad set of images. The new filter is a considerable improvement on the fixed-mask bitonic, outperforms anisotropic diffusion and image-guided filtering in all but extremely low noise, non-local means at all noise levels, but not the block-matching 3D filter, though results are promising for very high noise. The structurally varying bitonic tends to have less characteristic residual noise in regions of smooth signal, and very good preservation of signal edges, though with some loss of small scale detail when compared to the block-matching 3D filter. The efficient implementation means that processing time, though slower than the fixed-mask bitonic filter, remains competitive.
- 7 August 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Image Processing
- Vol. 29 (10577149) , 336-350
Morphology-Based Noise Reduction: Structural Variation and Thresholding in the Bitonic Filter
The bitonic filter was recently developed to embody the novel concept of signal bitonicity (one local extremum within a set range) to differentiate from noise, by use of data ranking and linear operators. For processing images, the spatial extent was locally constrained to a fixed circular mask. Since structure in natural images varies, a novel structurally varying bitonic filter is presented, which locally adapts the mask, without following patterns in the noise. This new filter includes novel robust structurally varying morphological operations, with efficient implementations, and a novel formulation of non-iterative directional Gaussian filtering. Data thresholds are also integrated with the morphological operations, increasing noise reduction for low noise, and enabling a multi-resolution framework for high noise levels. The structurally varying bitonic filter is presented without presuming prior knowledge of morphological filtering, and compared to high-performance linear noise-reduction filters, to set this novel concept in context. These are tested over a wide range of noise levels, on a fairly broad set of images. The new filter is a considerable improvement on the fixed-mask bitonic, outperforms anisotropic diffusion and image-guided filtering in all but extremely low noise, non-local means at all noise levels, but not the block-matching 3D filter, though results are promising for very high noise. The structurally varying bitonic tends to have less characteristic residual noise in regions of smooth signal, and very good preservation of signal edges, though with some loss of small scale detail when compared to the block-matching 3D filter. The efficient implementation means that processing time, though slower than the fixed-mask bitonic filter, remains competitive.
- Open Access
- 28 August 2018
- Published by Springer Nature in Current Osteoporosis Reports
- Vol. 16 (5) , 617-625
Cortical Bone Mapping: Measurement and Statistical Analysis of Localised Skeletal Changes
Purpose of Review Cortical bone mapping (CBM) is a technique for measuring localised skeletal changes from computed tomography (CT) images. It can provide measurements with accuracy surpassing the underlying imaging resolution. CBM can detect changes in several properties of the cortex, with no prior assumptions about the likely location of said changes. This paper summarises the theory behind CBM, discusses its strengths and limitations, and reviews some studies in which it has been applied. Recent Findings CBM has revealed associations between fracture risk and cortical properties in specific regions of the proximal femur which present feasible therapeutic targets. Analyses of several pharmaceutical and exercise interventions quantify effects that are distinct both in location and in the nature of the micro-architectural changes. CBM has illuminated age-related changes in the proximal femur and has recently been applied to other bones, as well as to the assessment of cartilage. Summary The CBM processing pipeline is designed primarily for large cohort studies. Its main impact thus far has not been in the realm of clinical practice, but rather to improve our fundamental understanding of localised bone structure and changes.
- 28 August 2018
- Published by Springer Nature in Current Osteoporosis Reports
- Vol. 16 (5) , 617-625
Open AccessCortical Bone Mapping: Measurement and Statistical Analysis of Localised Skeletal Changes
Purpose of Review Cortical bone mapping (CBM) is a technique for measuring localised skeletal changes from computed tomography (CT) images. It can provide measurements with accuracy surpassing the underlying imaging resolution. CBM can detect changes in several properties of the cortex, with no prior assumptions about the likely location of said changes. This paper summarises the theory behind CBM, discusses its strengths and limitations, and reviews some studies in which it has been applied. Recent Findings CBM has revealed associations between fracture risk and cortical properties in specific regions of the proximal femur which present feasible therapeutic targets. Analyses of several pharmaceutical and exercise interventions quantify effects that are distinct both in location and in the nature of the micro-architectural changes. CBM has illuminated age-related changes in the proximal femur and has recently been applied to other bones, as well as to the assessment of cartilage. Summary The CBM processing pipeline is designed primarily for large cohort studies. Its main impact thus far has not been in the realm of clinical practice, but rather to improve our fundamental understanding of localised bone structure and changes.
- Open Access
- 1 April 2018
- Published by Elsevier in Medical Image Analysis
- Vol. 45, 55-67
How does the femoral cortex depend on bone shape? A methodology for the joint analysis of surface texture and shape
- 1 April 2018
- Published by Elsevier in Medical Image Analysis
- Vol. 45, 55-67
Open AccessHow does the femoral cortex depend on bone shape? A methodology for the joint analysis of surface texture and shape
- Open Access
- 1 February 2018
- Published by Elsevier in Medical Image Analysis
- Vol. 44, 28-40
Measurement of the bone endocortical region using clinical CT
- 1 February 2018
- Published by Elsevier in Medical Image Analysis
- Vol. 44, 28-40
Open AccessMeasurement of the bone endocortical region using clinical CT
- Open Access
- 30 August 2017
- Published by Springer Nature in Radiation Oncology
- Vol. 12 (1) , 1-4
Novel three-dimensional bone ‘mapping’ software can help assess progression of osseous metastases from routine CT
Imaging of bone metastasis response to therapy is a research priority. Stradwin is a new software-tool, with demonstrated sub-voxel accuracy in assessing cortical bone properties from routine CT. We applied this technology to the context of osseous metastases, with particular focus on disease progression using prostate cancer as a model. 3D–rendered ‘bone-maps’ were produced for 20 men with advanced prostate cancer, including a sub-cohort of 9 who had undergone serial scans. Correlation between baseline interpretation and assessments of progression between modalities was assessed. Bone-maps took significantly less time to interpret than CT bone windows (P < 0.001). Initial bone-mapping, without adjustment, demonstrated sensitivity and specificity for suspicious areas on CT of 70.7% and 73.1% respectively. Evaluating disease over time, concordance between bone-maps and current practice using RECIST outcomes was 100%. This study demonstrates the feasibility and potential use of this free post-processing software in the serial assessment of osseous metastases.
- 30 August 2017
- Published by Springer Nature in Radiation Oncology
- Vol. 12 (1) , 1-4
Open AccessNovel three-dimensional bone ‘mapping’ software can help assess progression of osseous metastases from routine CT
Imaging of bone metastasis response to therapy is a research priority. Stradwin is a new software-tool, with demonstrated sub-voxel accuracy in assessing cortical bone properties from routine CT. We applied this technology to the context of osseous metastases, with particular focus on disease progression using prostate cancer as a model. 3D–rendered ‘bone-maps’ were produced for 20 men with advanced prostate cancer, including a sub-cohort of 9 who had undergone serial scans. Correlation between baseline interpretation and assessments of progression between modalities was assessed. Bone-maps took significantly less time to interpret than CT bone windows (P < 0.001). Initial bone-mapping, without adjustment, demonstrated sensitivity and specificity for suspicious areas on CT of 70.7% and 73.1% respectively. Evaluating disease over time, concordance between bone-maps and current practice using RECIST outcomes was 100%. This study demonstrates the feasibility and potential use of this free post-processing software in the serial assessment of osseous metastases.
- 24 March 2017
- Published by Elsevier in European Urology Supplements
- Vol. 16 (3) , e104-e105
Novel three-dimensional bone ‘mapping’ software can help assess progression of osseous prostate cancer metastases from routine CT
- 24 March 2017
- Published by Elsevier in European Urology Supplements
- Vol. 16 (3) , e104-e105
Novel three-dimensional bone ‘mapping’ software can help assess progression of osseous prostate cancer metastases from routine CT