NMNicholas William Marshall

Nicholas William Marshall

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Nicholas William Marshall has a total of 253 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Technical performance evaluation of mobile x-ray systems for chest imaging

    by N.W. Marshall, J. Binst, H. Verhoeven, T. Stinglhamber, W. Michiels, H. Bosmans

  • Comparison of Artificial Intelligence Detection and Short-Term Breast Cancer Risk Models within and beyond Their Intended Use in Mammography Screening

    by Yao-Kuan Wang, Lesley Cockmartin, Tobias Wagner, Zan Klanecek, Nicholas Marshall, Valerie Celis, Chantal Van Ongeval, Andrej Studen, Robert Jeraj, Hilde Bosmans

    Artificial intelligence–based mammographic detection and short-term breast cancer risk prediction models performed best in their intended applications, with lower performance when applied outside these settings.

  • flEXPOSE: a flexible exposure parameter optimisation engine for x-ray projection imaging

    by Rodrigo T Massera, Nicholas W Marshall, Hilde Bosmans

    Objective. To develop and illustrate the potential of a new, flexible, open-source software engine for task-based optimisation of exposure parameter settings in x-ray projection imaging. Approach. The engine was built from several Python-scripted modules to automate exposure parameter optimization. Input is taken from a set of pre-calculated tables containing image quality and dose data, and system parameters defined by the user. Modular code is employed, with classes responsible for image quality and dose calculations. For this study, the image quality (IQ) module incorporated the standard signal-to-noise ratio ( SNR ) and a version of SNR ( SNR w ) that is weighted for the influence of the x-ray focus size and finite x-ray pulse width on the task. Optimal x-ray factors for a specified task are established by the Optimizer class that maximizes a figure-of-merit defined as SNR 2 or SNR w 2 divided by dose. A set of six experiments with different degrees of complexity was performed to illustrate the engine and the influence of x-ray factor selection for a cardiac imaging task. Main results . A full parameter search covering 2400 different combinations of tube potential, additional copper filtration and focal spot size for 11 distinct patient thicknesses took approximately 30 min. The six experiments demonstrated that it is essential to consider x-ray tube power limitations and, when applicable, object motion and dose limits when determining the optimal exposure parameters. Significance . The proposed engine automates optimal exposure parameter selection for user-defined image quality metrics and dose estimates. The influence of x-ray system parameters on system performance can be explored systematically. The engine is provided as an open-source resource with a modular structure that can be extended to include different figures of merit, and different image quality and dose metrics. The repository containing the engine is available at https://gitlab.kuleuven.be/medphysqa/deploy/flexpose/flexpose .


  • Using Explainable AI to Characterize Features in the Mirai Mammographic Breast Cancer Risk Prediction Model

    by Yao-Kuan Wang, Zan Klanecek, Tobias Wagner, Lesley Cockmartin, Nicholas Marshall, Andrej Studen, Robert Jeraj, Hilde Bosmans

    A feature-centric explainable artificial intelligence pipeline showed that the Mirai model uses lesion features on mammograms for lesion detection and breast cancer risk prediction. A simplified model using only calcification features achieved comparable risk prediction performance to the original Mirai model in a screen-negative population.


  • Open Access

    Impact of synthetic data on training a deep learning model for lesion detection and classification in contrast-enhanced mammography

    by Astrid Van Camp, Henry C. Woodruff, Lesley Cockmartin, Marc Lobbes, Michael Majer, Corinne Balleyguier, Nicholas W. Marshall, Hilde Bosmans, Philippe Lambin

    PurposePredictive models for contrast-enhanced mammography often perform better at detecting and classifying enhancing masses than (non-enhancing) microcalcification clusters. We aim to investigate whether incorporating synthetic data with simulated microcalcification clusters during training can enhance model performance.ApproachMicrocalcification clusters were simulated in low-energy images of lesion-free breasts from 782 patients, considering local texture features. Enhancement was simulated in the corresponding recombined images. A deep learning (DL) model for lesion detection and classification was trained with varying ratios of synthetic and real (850 patients) data. In addition, a handcrafted radiomics classifier was trained using delineations and class labels from real data, and predictions from both models were ensembled. Validation was performed on internal (212 patients) and external (279 patients) real datasets.ResultsThe DL model trained exclusively with synthetic data detected over 60% of malignant lesions. Adding synthetic data to smaller real training sets improved detection sensitivity for malignant lesions but decreased precision. Performance plateaued at a detection sensitivity of 0.80. The ensembled DL and radiomics models performed worse than the standalone DL model, decreasing the area under this receiver operating characteristic curve from 0.75 to 0.60 on the external validation set, likely due to falsely detected suspicious regions of interest.ConclusionsSynthetic data can enhance DL model performance, provided model setup and data distribution are optimized. The possibility to detect malignant lesions without real data present in the training set confirms the utility of synthetic data. It can serve as a helpful tool, especially when real data are scarce, and it is most effective when complementing real data.


  • Open Access

    Sensitivity of a deep-learning-based breast cancer risk prediction model

    by Zan Klanecek, Yao-Kuan Wang, Tobias Wagner, Lesley Cockmartin, Nicholas Marshall, Brayden Schott, Ali Deatsch, Andrej Studen, Katja Jarm, Mateja Krajc, et al.

    Objective. When it comes to the implementation of deep-learning based breast cancer risk (BCR) prediction models in clinical settings, it is important to be aware that these models could be sensitive to various factors, especially those arising from the acquisition process. In this work, we investigated how sensitive the state-of-the-art BCR prediction model is to realistic image alterations that can occur as a result of different positioning during the acquisition process. Approach. 5076 mammograms (1269 exams, 650 participants) from the Slovenian and Belgium (University Hospital Leuven) Breast Cancer Screening Programs were collected. The Original MIRAI model was used for 1–5 year BCR estimation. First, BCR was predicted for the original mammograms, which were not changed. Then, a series of different image alteration techniques was performed, such as swapping left and right breasts, removing tissue below the inframammary fold, translations, cropping, rotations, registration and pectoral muscle removal. In addition, a subset of 81 exams, where at least one of the mammograms had to be retaken due to inadequate image quality, served as an approximation of a test–retest experiment. Bland–Altman plots were used to determine prediction bias and 95% limits of agreement (LOA). Additionally, the mean absolute difference in BCR (Mean AD) was calculated. The impact on the overall discrimination performance was evaluated with the AUC. Results. Swapping left and right breasts had no impact on the predicted BCR. The removal of skin tissue below the inframammary fold had minimal impact on the predicted BCR (1–5 year LOA: [−0.02, 0.01]). The model was sensitive to translation, rotation, registration, and cropping, where LOAs of up to ±0.1 were observed. Partial pectoral muscle removal did not have a major impact on predicted BCR, while complete removal of pectoral muscle introduced substantial prediction bias and LOAs (1 year LOA: [−0.07, 0.04], 5 year LOA: [−0.06, 0.03]). The approximation of a real test–retest experiment resulted in LOAs similar to those of simulated image alterations. None of the alterations impacted the overall BCR discrimination performance; the initial 1 year AUC (0.90 [0.88, 0.92]) and 5 year AUC (0.77 [0.75, 0.80]) remained unchanged. Significance. While tested image alterations do not impact overall BCR discrimination performance, substantial changes in predicted 1–5 year BCR can occur on an individual basis.


  • Open Access

    Impact of pectoral muscle removal on deep-learning-based breast cancer risk prediction

    by Zan Klanecek, Yao-Kuan Wang, Tobias Wagner, Lesley Cockmartin, Nicholas Marshall, Brayden Schott, Ali Deatsch, Andrej Studen, Katja Jarm, Mateja Krajc, et al.

    Objective. State-of-the-art breast cancer risk (BCR) prediction models have been originally trained on mammograms with pectoral muscle (PM) included. This study investigated whether excluding PM during training/fine-tuning improves the model’s BCR discrimination performance, calibration, and robustness. Approach. First, the Original deep learning model (MIRAI), trained on the US (Massachusetts General Hospital) data, was validated, and the relative contribution of PM to BCR predictions was evaluated using saliency maps. Additionally, 23 792 mammograms from the Slovenian screening program were collected and two datasets were created, with and without screening positive exams. The original MIRAI was then fine-tuned on the training/fine-tuning set of Slovenian mammograms with and without PM, creating Fine-tuned MIRAI models. In total, four models (Original MIRAI with PM, Original MIRAI without PM, Fine-tuned MIRAI with PM, Fine-tuned MIRAI without PM) were compared on a test set in terms of discrimination performance for 1–5 Year BCR (evaluating area under the curve), calibration performance (measured with expected calibration error—ECE) and robustness to incremental PM removals/additions, and to incremental breast tissue removals. Results. The relative contribution of PM to the BCR prediction on the Original MIRAI model was low (∼5%); however, there were significant outliers where the relative contribution was more than 50%. The removal of PM did not impact the 1–5 Year BCR discrimination performance of the Original MIRAI (with screening positive exams: 0.77–0.91, without screening positive exams: 0.64–0.67). Fine-tuned MIRAI on mammograms with PM removed achieved significantly higher 1-5 Year BCR discrimination performance (with screening positive exams: 0.82–0.93, without screening positive exams: 0.71–0.79). After recalibration, all models had similar ECE (with screening positive exams: 0.04–0.05, without screening positive exams: 0.02–0.03). Significance. Improved BCR discrimination performance can be achieved when the model is trained/fine-tuned on mammograms with PM removed.


  • Open Access

    Longitudinal interpretability of deep learning based breast cancer risk prediction

    by Zan Klanecek, Yao-Kuan Wang, Tobias Wagner, Lesley Cockmartin, Nicholas Marshall, Brayden Schott, Ali Deatsch, Andrej Studen, Katja Jarm, Mateja Krajc, et al.

    : Objective. Deep-learning-based models have achieved state-of-the-art breast cancer risk (BCR) prediction performance. However, these models are highly complex, and the underlying mechanisms of BCR prediction are not fully understood. Key questions include whether these models can detect breast morphologic changes that lead to cancer. These findings would boost confidence in utilizing BCR models in practice and provide clinicians with new perspectives. In this work, we aimed to determine when oncogenic processes in the breast provide sufficient signal for the models to detect these changes. Approach. In total, 1210 screening mammograms were collected for patients screened at different times before the cancer was screen-detected and 2400 mammograms for patients with at least ten years of follow-up. MIRAI, a BCR risk prediction model, was used to estimate the BCR. Attribution heterogeneity was defined as the relative difference between the attributions obtained from the right and left breasts using one of the eight interpretability techniques. Model reliance on the side of the breast with cancer was quantified with AUC. The Mann–Whitney U test was used to check for significant differences in median absolute Attribution Heterogeneity between cancer patients and healthy individuals. Results. All tested attribution methods showed a similar longitudinal trend, where the model reliance on the side of the breast with cancer was the highest for the 0–1 years-to-cancer interval (AUC = 0.85–0.95), dropped for the 1–3 years-to-cancer interval (AUC = 0.64–0.71), and remained above the threshold for random performance for the 3–5 years-to-cancer interval (AUC = 0.51–0.58). For all eight attribution methods, the median values of absolute attribution heterogeneity were significantly larger for patients diagnosed with cancer at one point (p < 0.01). Significance. Interpretability of BCR prediction has revealed that long-term predictions (beyond three years) are most likely based on typical breast characteristics, such as breast density; for mid-term predictions (one to three years), the model appears to detect early signs of tumor development, while for short-term predictions (up to a year), the BCR model essentially functions as a breast cancer detection model.


  • Open Access

    Evaluation of the flying focal spot technology in a wide-angle digital breast tomosynthesis system

    by Katrien Houbrechts, Nicholas Marshall, Lesley Cockmartin, Hilde Bosmans

    Even though digital breast tomosynthesis (DBT) is a well-established imaging modality, efforts to optimize system accuracy and efficiency are still ongoing. Currently, different modes of gantry and/or detector movement exist. The two main acquisition modes for acquiring DBT projection data are a continuous gantry motion mode and a step-and-shoot mode. In the continuous mode, the x-ray tube moves continuously while x-rays are generated in short pulses at specific angles around the breast. This has the potential for faster acquisition times compared with the step-and-shoot operation but incurs source motion blurring along the direction of the source motion. Source motion blurring has been shown to reduce technical image quality, as quantified by a number of metrics.1 – 5 In a step-and-shoot mode, the x-ray tube stops at each angular location to acquire a projection image. This eliminates source motion blur provided that the tube stops without inducing any mechanical instability but may result in a longer scan acquisition time and therefore an increased chance for patient motion artifacts.6


  • A comprehensive assessment of a prototype high ratio antiscatter grid in interventional cardiology using experimental measurements and Monte Carlo simulations

    by Rodrigo T Massera, Michiel Dehairs, Hannelore Verhoeven, Hilde Bosmans, Nicholas Marshall

    Objective. To assess the performance of a new antiscatter grid design in interventional cardiology for image quality improvement and dose reduction using experimental measurements and Monte Carlo (MC) simulation. Approach. Experimental measurements were performed on an angiography system, using a multi-layered tissue simulating composite phantom made from of poly(methyl methacrylate), aluminium and expanded polystyrene (2/0.2/0.7 cm). The total phantom thickness ranged from 20.3 cm to 40.6 cm. Four conditions were compared; (A) 105 cm source-image receptor distance (SID) without grid, (B i) 105 cm SID with grid ratio (r) and strip density (N) (r15N80), (B ii) 120 cm SID without grid, and (Biii) 120 cm SID with high ratio grid (r29N80). The system efficiency (η), defined by the signal-to-noise ratio, was compared from the B conditions against case A. These conditions were also simulated with MC techniques, allowing additional phantom compositions to be explored. Weighted image quality improvement factor (ηw (u)) was studied experimentally at a specific spatial frequency due to the SID change. Images were simulated with an anthropomorphic chest phantom for the different conditions, and the system efficiency was compared for the different anatomical regions. Main results. Good agreement was found between the η and ηw (u) methods using both measured and simulated data, with average relative differences between 2%–11%. Case B iii provided higher η values compared to A, and B i for thicknesses larger than 20.3 cm. In addition, case B iii also provided higher η values for high attenuating areas in the anthropomorphic phantom, such as behind the spine. Significance. The new antiscatter grid design provided higher system efficiency compared to the standard grid for the parameters explored in this work.