Image and Video Processing (eess.IV)

Abstract: Skin lesions known as naevi exhibit diverse characteristics such as size, shape, and colouration. The concept of an "Ugly Duckling Naevus" comes into play when monitoring for melanoma, referring to a lesion with distinctive features that sets it apart from other lesions in the vicinity. As lesions within the same individual typically share similarities and follow a predictable pattern, an ugly duckling naevus stands out as unusual and may indicate the presence of a cancerous melanoma. Computer-aided diagnosis (CAD) has become a significant player in the research and development field, as it combines machine learning techniques with a variety of patient analysis methods. Its aim is to increase accuracy and simplify decision-making, all while responding to the shortage of specialized professionals. These automated systems are especially important in skin cancer diagnosis where specialist availability is limited. As a result, their use could lead to life-saving benefits and cost reductions within healthcare. Given the drastic change in survival when comparing early stage to late-stage melanoma, early detection is vital for effective treatment and patient outcomes. Machine learning (ML) and deep learning (DL) techniques have gained popularity in skin cancer classification, effectively addressing challenges, and providing results equivalent to that of specialists. This article extensively covers modern Machine Learning and Deep Learning algorithms for detecting melanoma and suspicious naevi. It begins with general information on skin cancer and different types of naevi, then introduces AI, ML, DL, and CAD. The article then discusses the successful applications of various ML techniques like convolutional neural networks (CNN) for melanoma detection compared to dermatologists' performance. Lastly, it examines ML methods for UD naevus detection and identifying suspicious naevi.

Abstract: Deep learning models have revolutionized the field of medical image analysis, offering significant promise for improved diagnostics and patient care. However, their performance can be misleadingly optimistic due to a hidden pitfall called 'data leakage'. In this study, we investigate data leakage in 3D medical imaging, specifically using 3D Convolutional Neural Networks (CNNs) for brain MRI analysis. While 3D CNNs appear less prone to leakage than 2D counterparts, improper data splitting during cross-validation (CV) can still pose issues, especially with longitudinal imaging data containing repeated scans from the same subject. We explore the impact of different data splitting strategies on model performance for longitudinal brain MRI analysis and identify potential data leakage concerns. GradCAM visualization helps reveal shortcuts in CNN models caused by identity confounding, where the model learns to identify subjects along with diagnostic features. Our findings, consistent with prior research, underscore the importance of subject-wise splitting and evaluating our model further on hold-out data from different subjects to ensure the integrity and reliability of deep learning models in medical image analysis.

Abstract: Thyroid disorders are most commonly diagnosed using high-resolution Ultrasound (US). Longitudinal nodule tracking is a pivotal diagnostic protocol for monitoring changes in pathological thyroid morphology. This task, however, imposes a substantial cognitive load on clinicians due to the inherent challenge of maintaining a mental 3D reconstruction of the organ. We thus present a framework for automated US image slice localization within a 3D shape representation to ease how such sonographic diagnoses are carried out. Our proposed method learns a common latent embedding space between US image patches and the 3D surface of an individual's thyroid shape, or a statistical aggregation in the form of a statistical shape model (SSM), via contrastive metric learning. Using cross-modality registration and Procrustes analysis, we leverage features from our model to register US slices to a 3D mesh representation of the thyroid shape. We demonstrate that our multi-modal registration framework can localize images on the 3D surface topology of a patient-specific organ and the mean shape of an SSM. Experimental results indicate slice positions can be predicted within an average of 1.2 mm of the ground-truth slice location on the patient-specific 3D anatomy and 4.6 mm on the SSM, exemplifying its usefulness for slice localization during sonographic acquisitions. Code is publically available: \href{https://github.com/vuenc/slice-to-shape}{https://github.com/vuenc/slice-to-shape}

Abstract: Speed-of-sound (SoS) is an emerging ultrasound contrast modality, where pulse-echo techniques using conventional transducers offer multiple benefits. For estimating tissue SoS distributions, spatial domain reconstruction from relative speckle shifts between different beamforming sequences is a promising approach. This operates based on a forward model that relates the sought local values of SoS to observed speckle shifts, for which the associated image reconstruction inverse problem is solved. The reconstruction accuracy thus highly depends on the hand-crafted forward imaging model. In this work, we propose to learn the SoS imaging model based on data. We introduce a convolutional formulation of the pulse-echo SoS imaging problem such that the entire field-of-view requires a single unified kernel, the learning of which is then tractable and robust. We present least-squares estimation of such convolutional kernel, which can further be constrained and regularized for numerical stability. In experiments, we show that a forward model learned from k-Wave simulations improves the median contrast of SoS reconstructions by 63%, compared to a conventional hand-crafted line-based wave-path model. This simulation-learned model generalizes successfully to acquired phantom data, nearly doubling the SoS contrast compared to the conventional hand-crafted alternative. We demonstrate equipment-specific and small-data regime feasibility by learning a forward model from a single phantom image, where our learned model quadruples the SoS contrast compared to the conventional hand-crafted model. On in-vivo data, the simulation- and phantom-learned models respectively exhibit impressive 7 and 10 folds contrast improvements over the conventional model.

Abstract: In Computed Tomography (CT), an image of the interior structure of an object is computed from a set of acquired projection images. The quality of these reconstructed images is essential for accurate analysis, but this quality can be degraded by a variety of imaging artifacts. To improve reconstruction quality, the acquired projection images are often processed by a pipeline consisting of multiple artifact-removal steps applied in various image domains (e.g., outlier removal on projection images and denoising of reconstruction images). These artifact-removal methods exploit the fact that certain artifacts are easier to remove in a certain domain compared with other domains. Recently, deep learning methods have shown promising results for artifact removal for CT images. However, most existing deep learning methods for CT are applied as a post-processing method after reconstruction. Therefore, artifacts that are relatively difficult to remove in the reconstruction domain may not be effectively removed by these methods. As an alternative, we propose a multi-stage deep learning method for artifact removal, in which neural networks are applied to several domains, similar to a classical CT processing pipeline. We show that the neural networks can be effectively trained in succession, resulting in easy-to-use and computationally efficient training. Experiments on both simulated and real-world experimental datasets show that our method is effective in reducing artifacts and superior to deep learning-based post-processing.

Abstract: In this work, an image translation model is implemented to produce synthetic amyloid-beta PET images from structural MRI that are quantitatively accurate. Image pairs of amyloid-beta PET and structural MRI were used to train the model. We found that the synthetic PET images could be produced with a high degree of similarity to truth in terms of shape, contrast and overall high SSIM and PSNR. This work demonstrates that performing structural to quantitative image translation is feasible to enable the access amyloid-beta information from only MRI.