scispace - formally typeset
Search or ask a question
Author

Michael Berks

Bio: Michael Berks is an academic researcher from University of Manchester. The author has contributed to research in topics: Medicine & Random forest. The author has an hindex of 9, co-authored 36 publications receiving 274 citations.

Papers
More filters
Journal ArticleDOI
TL;DR: This work has built convolutional neural networks (CNN) to predict density VAS scores from full-field digital mammograms and shows promising results for cancer risk prediction and is comparable with human performance.
Abstract: Mammographic density is an important risk factor for breast cancer. In recent research, percentage density assessed visually using visual analogue scales (VAS) showed stronger risk prediction than existing automated density measures, suggesting readers may recognize relevant image features not yet captured by hand-crafted algorithms. With deep learning, it may be possible to encapsulate this knowledge in an automatic method. We have built convolutional neural networks (CNN) to predict density VAS scores from full-field digital mammograms. The CNNs are trained using whole-image mammograms, each labeled with the average VAS score of two independent readers. Each CNN learns a mapping between mammographic appearance and VAS score so that at test time, they can predict VAS score for an unseen image. Networks were trained using 67,520 mammographic images from 16,968 women and for model selection we used a dataset of 73,128 images. Two case-control sets of contralateral mammograms of screen detected cancers and prior images of women with cancers detected subsequently, matched to controls on age, menopausal status, parity, HRT and BMI, were used for evaluating performance on breast cancer prediction. In the case-control sets, odd ratios of cancer in the highest versus lowest quintile of percentage density were 2.49 (95% CI: 1.59 to 3.96) for screen-detected cancers and 4.16 (2.53 to 6.82) for priors, with matched concordance indices of 0.587 (0.542 to 0.627) and 0.616 (0.578 to 0.655), respectively. There was no significant difference between reader VAS and predicted VAS for the prior test set (likelihood ratio chi square, p = 0.134). Our fully automated method shows promising results for cancer risk prediction and is comparable with human performance.

34 citations

Journal ArticleDOI
TL;DR: Evaluability is one of the major challenges in assessing nailfold capillaries, but the high intra- and inter-reliabilities suggest that overall image grade, capillary density and apex width have potential as outcome measures in longitudinal studies.

34 citations

Book ChapterDOI
14 Sep 2014
TL;DR: A fully automated system for extracting quantitative biomarkers from capillaroscopy images, using a layered machine learning approach, that reveals statistically significant differences between patients with (relatively benign) primary Raynaud's phenomenon, and those with potentially life-threatening systemic sclerosis.
Abstract: Nailfold capillaroscopy is an established qualitative technique in the assessment of patients displaying Raynaud’s phenomenon. We describe a fully automated system for extracting quantitative biomarkers from capillaroscopy images, using a layered machine learning approach. On an unseen set of 455 images, the system detects and locates individual capillaries as well as human experts, and makes measurements of vessel morphology that reveal statistically significant differences between patients with (relatively benign) primary Raynaud’s phenomenon, and those with potentially life-threatening systemic sclerosis.

32 citations

Journal ArticleDOI
TL;DR: Quantitative nailfold capillaroscopy, at least with a single observer, provides reliable outcome measures for clinical studies including randomised controlled trials, and within-operator image analysis and acquisition are reproducible.

24 citations

Book ChapterDOI
03 Jul 2011
TL;DR: This work adopts a discriminative learning approach based on a Dual-Tree Complex Wavelet representation and random forest classification that gives significantly better results than any of the other methods on the challenge of detecting curvilinear structure in mammograms.
Abstract: Detecting and classifying curvilinear structure is important in many image interpretation tasks. We focus on the challenging problem of detecting such structure in mammograms and deciding whether it is normal or abnormal. We adopt a discriminative learning approach based on a Dual-Tree Complex Wavelet representation and random forest classification. We present results of a quantitative comparison of our approach with three leading methods from the literature and with learning-based variants of those methods. We show that our new approach gives significantly better results than any of the other methods, achieving an area under the ROC curve Az = 0.923 for curvilinear structure detection, and Az = 0.761 for distinguishing between normal and abnormal structure (spicules). A detailed analysis suggests that some of the improvement is due to discriminative learning, and some due to the DT-CWT representation, which provides local phase information and good angular resolution.

21 citations


Cited by
More filters
Book ChapterDOI
13 Jul 2012
TL;DR: Analysis of whether there is an optimal number of trees within a Random Forest finds an experimental relationship for the AUC gain when doubling the number of Trees in any forest and states there is a threshold beyond which there is no significant gain, unless a huge computational environment is available.
Abstract: Random Forest is a computationally efficient technique that can operate quickly over large datasets. It has been used in many recent research projects and real-world applications in diverse domains. However, the associated literature provides almost no directions about how many trees should be used to compose a Random Forest. The research reported here analyzes whether there is an optimal number of trees within a Random Forest, i.e., a threshold from which increasing the number of trees would bring no significant performance gain, and would only increase the computational cost. Our main conclusions are: as the number of trees grows, it does not always mean the performance of the forest is significantly better than previous forests (fewer trees), and doubling the number of trees is worthless. It is also possible to state there is a threshold beyond which there is no significant gain, unless a huge computational environment is available. In addition, it was found an experimental relationship for the AUC gain when doubling the number of trees in any forest. Furthermore, as the number of trees grows, the full set of attributes tend to be used within a Random Forest, which may not be interesting in the biomedical domain. Additionally, datasets' density-based metrics proposed here probably capture some aspects of the VC dimension on decision trees and low-density datasets may require large capacity machines whilst the opposite also seems to be true.

697 citations

Journal ArticleDOI
15 Jul 2016
TL;DR: The conclusion of the current study was that the frequency of screening might be dependent on breast density and in such cases diagnostic techniques such as “digital mammography, ultra sonography and magnetic resonance imaging” may prove to be better detection tools.
Abstract: With the increase in breast cancer risk over the years, there are many factors estimated that lead to it. However, till date which factor is majorly involved in development of breast cancer or which factor accounts more is not clearly evident. Mammography technique accounting for 80-90% of cancer being detected is believed to be the best method of detection. While mammographic density is manifested by increased proliferation of fat, stoma, epithelium and connective tissue, it is considered to be a risk factor for development of breast cancer. The current study was thus conducted to find out whether the mammographic density is actually a risk factor for development of breast cancer and to find out the better detection tool available. For this, the methodology adopted was review of various journals and studies already published with respect to mammographic density and its risk on development of breast cancer. The conclusion of the current study as well as from another comparable study was that the frequency of screening might be dependent on breast density and in such cases diagnostic techniques such as “digital mammography, ultra sonography and magnetic resonance imaging” may prove to be better detection tools. Moreover, recent studies have also suggested that mammographic density as a marker for risk of developing breast cancer holds true however, this fact needs to be evaluated further. Article DOI: https://dx.doi.org/10.20319/lijhls.2016.22.4854 This work is licensed under the Creative Commons Attribution-Non-commercial 4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc/4.0/ or send a letter to Creative Commons, PO Box 1866, Mountain View, CA 94042, USA.

317 citations

Journal ArticleDOI
TL;DR: Experts in the field of capillaroscopy/microcirculation provide in this very consensus paper their view on image acquisition and analysis, different capillsaroscopic techniques, normal and abnormal capillARoscopic characteristics and their meaning, scoring systems and reliability of image acquisitionand interpretation.

205 citations

Journal ArticleDOI
TL;DR: The specificity of DBT and 2D was better than 2D alone but there was only marginal improvement in sensitivity, and the performance of synthetic 2D appeared to be comparable to standard 2D.
Abstract: This project was funded by the NIHR Health Technology Assessment programme and will be published in full in Health Technology Assessment; Vol. 19, No. 4. See the HTA programme website for further project information.

164 citations

01 Jan 2014
TL;DR: Significant inroads are being made into the understanding of MD, which may lead to benefits in clinical screening, assessment and treatment strategies, and the biological and genetic pathways that determine and perhaps modulate MD remain largely unresolved.
Abstract: There has been considerable recent interest in the genetic, biological and epidemiological basis of mammographic density (MD), and the search for causative links between MD and breast cancer (BC) risk. This report will critically review the current literature on MD and summarize the current evidence for its association with BC. Keywords 'mammographic dens*', 'dense mammary tissue' or 'percent dens*' were used to search the existing literature in English on PubMed and Medline. All reports were critically analyzed. The data were assigned to one of the following aspects of MD: general association with BC, its relationship with the breast hormonal milieu, the cellular basis of MD, the generic variations of MD, and its significance in the clinical setting. MD adjusted for age, and BMI is associated with increased risk of BC diagnosis, advanced tumour stage at diagnosis and increased risk of both local recurrence and second primary cancers. The MD measures that predict BC risk have high heritability, and to date several genetic markers associated with BC risk have been found to also be associated with these MD risk predictors. Change in MD could be a predictor of the extent of chemoprevention with tamoxifen. Although the biological and genetic pathways that determine and perhaps modulate MD remain largely unresolved, significant inroads are being made into the understanding of MD, which may lead to benefits in clinical screening, assessment and treatment strategies. This review provides a timely update on the current understanding of MD's association with BC risk.

161 citations