Unsupervised Positron Emission Tomography Tumor Segmentation via GAN based Adversarial Auto-Encoder

Wu, Xinheng; Bi, Lei; Fulham, Michael; Kim, Jin Man

doi:10.1109/icarcv50220.2020.9305364

Cited by 8 publications

(3 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As further developments, we plan to investigate innovative improvements to further improve the performance with fuzzy clustering, by using (i) more sophisticated membership functions, and (ii) more advanced pre-and post-processing steps. Moreover, investigating and comparing the latest machine learning techniques, such as Generative Adversarial Networks (GANs), for unsupervised detection and segmentation [57,58] would be relevant with a sufficient amount of data for training and test. Finally, the implementation of multiparametric or multimodal approaches [59], by using different types of co-registered medical images-i.e., Diffusion Weighted Imaging (DWI) and Positron Emission Tomography (PET)/MRI-probably would allow us to improve the detection performance [60,61].…”

Section: Discussionmentioning

confidence: 99%

On Unsupervised Methods for Medical Image Segmentation: Investigating Classic Approaches in Breast Cancer DCE-MRI

et al. 2021

View full text Add to dashboard Cite

Unsupervised segmentation techniques, which do not require labeled data for training and can be more easily integrated into the clinical routine, represent a valid solution especially from a clinical feasibility perspective. Indeed, large-scale annotated datasets are not always available, undermining their immediate implementation and use in the clinic. Breast cancer is the most common cause of cancer death in women worldwide. In this study, breast lesion delineation in Dynamic Contrast Enhanced MRI (DCE-MRI) series was addressed by means of four popular unsupervised segmentation approaches: Split-and-Merge combined with Region Growing (SMRG), k-means, Fuzzy C-Means (FCM), and spatial FCM (sFCM). They represent well-established pattern recognition techniques that are still widely used in clinical research. Starting from the basic versions of these segmentation approaches, during our analysis, we identified the shortcomings of each of them, proposing improved versions, as well as developing ad hoc pre- and post-processing steps. The obtained experimental results, in terms of area-based—namely, Dice Index (DI), Jaccard Index (JI), Sensitivity, Specificity, False Positive Ratio (FPR), False Negative Ratio (FNR)—and distance-based metrics—Mean Absolute Distance (MAD), Maximum Distance (MaxD), Hausdorff Distance (HD)—encourage the use of unsupervised machine learning techniques in medical image segmentation. In particular, fuzzy clustering approaches (namely, FCM and sFCM) achieved the best performance. In fact, for area-based metrics, they obtained DI = 78.23% ± 6.50 (sFCM), JI = 65.90% ± 8.14 (sFCM), sensitivity = 77.84% ± 8.72 (FCM), specificity = 87.10% ± 8.24 (sFCM), FPR = 0.14 ± 0.12 (sFCM), and FNR = 0.22 ± 0.09 (sFCM). Concerning distance-based metrics, they obtained MAD = 1.37 ± 0.90 (sFCM), MaxD = 4.04 ± 2.87 (sFCM), and HD = 2.21 ± 0.43 (FCM). These experimental findings suggest that further research would be useful for advanced fuzzy logic techniques specifically tailored to medical image segmentation.

show abstract

Section: Discussionmentioning

confidence: 99%

On Unsupervised Methods for Medical Image Segmentation: Investigating Classic Approaches in Breast Cancer DCE-MRI

et al. 2021

View full text Add to dashboard Cite

show abstract

“…TernausNet [12], a U-Net variant, reshapes the U-Net encoder to match the VGG11 architecture, allowing it to leverage pre-trained VGG11 [19] model weights for faster convergence and improved segmentation outcomes. While most medical imaging segmentation models are trained using supervised learning, weakly supervised segmentation methods such as VoxelMorph augmented segmentation [27], ACNN [16], CCNN [13], graph-based unsupervised segmentation [15], and GAN-based unsupervised segmentation [23,24] also achieve comparable results. For the segmentation of the 4CH, 2CH, SA, and aorta view CMR imaging dataset from the UK Biobank, Bai et al [2] provide a supervised segmentation model.…”

Section: Segmentation Of the Cardiac Magnetic Resonance Imagesmentioning

confidence: 99%

Recognizing Mitral Regurgitation Through Machine Learning Algorithms in Cardiac Imaging

Singh,

Anwar,

Vavekanand

2024

Preprint

View full text Add to dashboard Cite

Mitral regurgitation (MR) is a serious heart valve disease that can have devastating consequences if left untreated. Traditional diagnostic methods are expensive and labor-intensive, requiring specialized clinical expertise, creating barriers to effective MR screening. To address these challenges, a novel semi-supervised model called CUSSP is proposed. CUSSP is designed to process cardiac imaging slices from the 4-chamber view of the heart, utilizing standard computer vision techniques and contrastive models to learn from large amounts of unlabeled data. This approach enables the model to leverage vast amounts of available imaging data, even when labeled data is scarce. CUSSP creates the first automated MR classification system, revolutionizing the diagnosis and treatment of this critical heart condition. The significance of CUSSP lies in its ability to overcome the limitations of traditional diagnostic methods. By automating the process, healthcare professionals can focus on high-value tasks while ensuring accurate and timely diagnoses. The performance of CUSSP is impressive, achieving an F1 score of 0.69 and a ROC-AUC score of 0.88 on a test set of 179 labeled sequences, comprising 154 non-MR and 25 MR cases. The model is trained using a semi-supervised approach, combining the strengths of both supervised and unsupervised learning. By incorporating specialized classifiers, CUSSP achieves high accuracy and robustness, even in the presence of limited labeled data. CUSSP represents a significant breakthrough in MR diagnosis, offering a novel semi-supervised approach to automated classification. By leveraging computer vision techniques, contrastive models, and specialized classifiers, CUSSP achieves high accuracy and robustness, establishing a new benchmark for MR diagnosis.

show abstract

“…The proposed model got a mean dice score of 0.683 when tested on 100 lung CT images. Besides this, table 2 includes some more GAN-based techniques[75]-[84] applied to medical images.…”

mentioning

confidence: 99%

Convolutional Neural Networks for Medical Image Segmentation and Classification: A Review

M J

2023

jist

View full text Add to dashboard Cite

Medical imaging refers to the process of obtaining images of internal organs for therapeutic purposes such as discovering or studying diseases. The primary objective of medical image analysis is to improve the efficacy of clinical research and treatment options. Deep learning has revamped medical image analysis, yielding excellent results in image processing tasks such as registration, segmentation, feature extraction, and classification. The prime motivations for this are the availability of computational resources and the resurgence of deep Convolutional Neural Networks. Deep learning techniques are good at observing hidden patterns in images and supporting clinicians in achieving diagnostic perfection. It has proven to be the most effective method for organ segmentation, cancer detection, disease categorization, and computer-assisted diagnosis. Many deep learning approaches have been published to analyze medical images for various diagnostic purposes. In this paper, we review the works exploiting current state-of-the-art deep learning approaches in medical image processing. We begin the survey by providing a synopsis of research works in medical imaging based on convolutional neural networks. Second, we discuss popular pre-trained models and General Adversarial Networks that aid in improving convolutional networks' performance. Finally, to ease direct evaluation, we compile the performance metrics of deep learning models focusing on covid-19 detection and child bone age prediction.

show abstract

Unsupervised Positron Emission Tomography Tumor Segmentation via GAN based Adversarial Auto-Encoder

Cited by 8 publications

References 11 publications

On Unsupervised Methods for Medical Image Segmentation: Investigating Classic Approaches in Breast Cancer DCE-MRI

On Unsupervised Methods for Medical Image Segmentation: Investigating Classic Approaches in Breast Cancer DCE-MRI

Recognizing Mitral Regurgitation Through Machine Learning Algorithms in Cardiac Imaging

Convolutional Neural Networks for Medical Image Segmentation and Classification: A Review

Contact Info

Product

Resources

About