Improving Prostate Whole Gland Segmentation In T2-Weighted MRI With Synthetically Generated Data

Fernandez-Quilez, Alvaro; Larsen, Steinar Valle; Goodwin, Morten; Gulsurd, Thor Ole; Kjosavik, Svein Reidar; Oppedal, Ketil

doi:10.1109/isbi48211.2021.9433793

Cited by 11 publications

(11 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Such applications are particularly useful in the medical domain, where images can come from different machines with different characteristics and thus, a domain adaptation in which images are translated to the training domain might prove useful [118,119]. The generation of synthetic samples through GANs has also caught the interest of the medical imaging community, given the lack of annotated data and the difficulties to obtain them in the medical domain [120,121,122].…”

Section: Generative Adversarial Network (Gan)mentioning

confidence: 99%

“…In radiology, GANs have been used to synthesize medical images like chest radiographs [172], CT scans with lung nodules [173], images from brain MRI sequences [174] and prostate MRI scans [120] with improved performance in different applications. For instance, in [173] the DL algorithm achieved a better sensitivity and specificity through the addition of synthetic samples.…”

Section: Synthesismentioning

confidence: 99%

“…The main objective of contribution A [120] is to tackle the lack of segmentation annotations for the prostate capsule (data scarcity) and to provide an alternative to classic augmentation techniques (basic image manipulations). In particular, we argue that standard augmentation techniques produce highly correlated samples, limiting the variability of the data used to train the algorithm and the amount of information that the data can offer in the training process.…”

Section: Contribution A: Improving Prostate Whole Gland Segmentation ...mentioning

confidence: 99%

“…We base the second part of our framework in the pix2pix architecture [117]. Details about the training of the different architectures can be found in [120]. Finally, we evaluate our proposed framework by evaluating the quality of the segmentation results obtained by means of the U-net architecture, in the presence of synthetically generated samples, standard augmentation techniques and a plain U-net.…”

Section: Contribution A: Improving Prostate Whole Gland Segmentation ...mentioning

confidence: 99%

“…Contribution B [249] aims to tackle the data scarcity and lack of annotations in the context of lesion classification (clinical significance) and with ADC maps. In a similar fashion to Contribution A [120], we aim to provide an alternative to standard augmentation techniques (data manipulation) for PCa classification. We argue, again, that standard augmentation techniques might limit the amount of information provided to the final task under consideration and that even depending on the type of chosen basic manipulation the final results might not see any benefit from it but rather the opposite, the augmentations might distort the image in a way that the label (ground truth) might be distorted as well (Section 3.1.1, augmentation safety).…”

Section: Contribution B: Improving Prostate Cancer Triage With Gan-ba...mentioning

confidence: 99%

See 4 more Smart Citations

Deep learning for an improved diagnostic pathway of prostate cancer in a small multi-parametric magnetic resonance data regime

Fernandez-Quilez¹

View full text Add to dashboard Cite

Prostate Cancer (PCa) is the second most commonly diagnosed cancer among men, with an estimated incidence of 1.3 million new cases worldwide in 2018. The current diagnostic pathway of PCa relies on prostate-specific antigen (PSA) levels in serum. Nevertheless, PSA testing comes at the cost of under-detection of malignant lesions and a substantial over-diagnosis of indolent ones, leading to unnecessary invasive testing such biopsies and treatment in indolent PCa lesions. Magnetic Resonance Imaging (MRI) is a non-invasive technique that has emerged as a valuable tool for PCa detection, staging, early screening, treatment planning and intervention. However, analysis of MRI relies on expertise, can be time-consuming, requires specialized training and in its absence suffers from inter and intra-reader variability and sub-optimal interpretations. Deep Learning (DL) techniques have the ability to recognize complex patterns in imaging data and are able to automatize certain assessments or tasks while offering a lesser degree of subjectiveness, providing a tool that can help clinicians in their daily tasks. In spite of it, DL success has traditionally relied on the availability of large amounts of labelled data, which are rarely available in the medical field and are costly and hard to obtain due to privacy regulations of patients’ data and required specialized training, among others. This work investigates DL algorithms specially tailored to work in a limited data regime with the final objective of improving the current prostate cancer diagnostic pathway by improving the performance of DL algorithms for PCa MRI applications in a limited data regime scenario. In particular, this thesis starts by exploring Generative Adversarial Networks (GAN) to generate synthetic samples and their effect on tasks such as prostate capsule segmentation and PCa lesion significance classification (triage). Following, we explore the use of Auto-encoders (AEs) to exploit the data imbalance that is usually present in medical imaging datasets. Specifically, we propose a framework based on AEs to detect the presence of prostate lesions (tumours) by uniquely learning from control (healthy) data in an outlier detection-like fashion. This thesis also explores more recent DL paradigms that have shown promising results in natural images: generative and contrastive self-supervised learning (SSL). In both cases, we propose specific prostate MRI image manipulations for a PCa lesion classification downstream task and show the improvements offered by the techniques when compared with other initialization methods such as ImageNet pre-training. Finally, we explore data fusion techniques in order to leverage different data sources in the form of MRI sequences (orthogonal views) acquired by default during patient examinations and that are commonly ignored in DL systems. We show improvements in a PCa lesion significance classification when compared to a single input system (axial view).

show abstract

Section: Generative Adversarial Network (Gan)mentioning

confidence: 99%

Section: Synthesismentioning

confidence: 99%

Section: Contribution A: Improving Prostate Whole Gland Segmentation ...mentioning

confidence: 99%

Section: Contribution A: Improving Prostate Whole Gland Segmentation ...mentioning

confidence: 99%

Section: Contribution B: Improving Prostate Cancer Triage With Gan-ba...mentioning

confidence: 99%

See 3 more Smart Citations

Deep learning for an improved diagnostic pathway of prostate cancer in a small multi-parametric magnetic resonance data regime

Fernandez-Quilez¹

View full text Add to dashboard Cite

show abstract

On Undesired Emergent Behaviors in Compound Prostate Cancer Detection Systems

Sortland Rolfsnes,

Thangngat,

Eftestøl

et al. 2024

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Deep learning in radiology: ethics of data and on the value of algorithm transparency, interpretability and explainability

Fernandez-Quilez

2022

AI Ethics

View full text Add to dashboard Cite

AI systems are quickly being adopted in radiology and, in general, in healthcare. A myriad of systems is being proposed and developed on a daily basis for high-stake decisions that can lead to unwelcome and negative consequences. AI systems trained under the supervised learning paradigm greatly depend on the quality and amount of data used to develop them. Nevertheless, barriers in data collection and sharing limit the data accessibility and potential ethical challenges might arise due to them leading, for instance, to systems that do not offer equity in their decisions and discriminate against certain patient populations or that are vulnerable to appropriation of intellectual property, among others. This paper provides an overview of some of the ethical issues both researchers and end-users might meet during data collection and development of AI systems, as well an introduction to the current state of transparency, interpretability and explainability of the systems in radiology applications. Furthermore, we aim to provide a comprehensive summary of currently open questions and identify key issues during the development and deployment of AI systems in healthcare, with a particular focus on the radiology area.

show abstract

Improving Prostate Whole Gland Segmentation In T2-Weighted MRI With Synthetically Generated Data

Cited by 11 publications

References 19 publications

Deep learning for an improved diagnostic pathway of prostate cancer in a small multi-parametric magnetic resonance data regime

Deep learning for an improved diagnostic pathway of prostate cancer in a small multi-parametric magnetic resonance data regime

On Undesired Emergent Behaviors in Compound Prostate Cancer Detection Systems

Deep learning in radiology: ethics of data and on the value of algorithm transparency, interpretability and explainability

Contact Info

Product

Resources

About