CS-CO: A Hybrid Self-Supervised Visual Representation Learning Method for H&E-stained Histopathological Images

Visual representation extraction is a fundamental problem in the field of computational histopathology. Considering the powerful representation capacity of deep learning and the scarcity of annotations, self-supervised learning has emerged as a promising approach to extract effective visual representations from unlabeled histopathological images. Although a few self-supervised learning methods have been specifically proposed for histopathological images, most of them suffer from certain defects that may hurt the versatility or representation capacity. In this work, we propose CS-CO, a hybrid self-supervised visual representation learning method tailored for H&E-stained histopathological images, which integrates advantages of both generative and discriminative approaches. The proposed method consists of two self-supervised learning stages: cross-stain prediction (CS) and contrastive learning (CO). In addition, a novel data augmentation approach named stain vector perturbation is specifically proposed to facilitate contrastive learning. Our CS-CO makes good use of domain-specific knowledge and requires no side information, which means good rationality and versatility. We evaluate and analyze the proposed CS-CO on three H&E-stained histopathological image datasets with downstream tasks of patch-level tissue classification and slide-level cancer prognosis and subtyping. Experimental results demonstrate the effectiveness and robustness of the proposed CS-CO on common computational histopathology tasks. Furthermore, we also conduct ablation studies and prove that cross-staining prediction and contrastive learning in our CS-CO can complement and enhance each other. Our code is made available at https://github.com/easonyang1996/CS-CO.     

Self-supervised driven consistency training for annotation efficient histopathology image analysis

Training a neural network with a large labeled dataset is still a dominant paradigm in computational histopathology. However, obtaining such exhaustive manual annotations is often expensive, laborious, and prone to inter and intra-observer variability. While recent self-supervised and semi-supervised methods can alleviate this need by learning unsupervised feature representations, they still struggle to generalize well to downstream tasks when the number of labeled instances is small. In this work, we overcome this challenge by leveraging both task-agnostic and task-specific unlabeled data based on two novel strategies: (i) a self-supervised pretext task that harnesses the underlying multi-resolution contextual cues in histology whole-slide images to learn a powerful supervisory signal for unsupervised representation learning; (ii) a new teacher-student semi-supervised consistency paradigm that learns to effectively transfer the pretrained representations to downstream tasks based on prediction consistency with the task-specific unlabeled data.We carry out extensive validation experiments on three histopathology benchmark datasets across two classification and one regression based tasks, i.e., tumor metastasis detection, tissue type classification, and tumor cellularity quantification. Under limited-label data, the proposed method yields tangible improvements, which is close to or even outperforming other state-of-the-art self-supervised and supervised baselines. Furthermore, we empirically show that the idea of bootstrapping the self-supervised pretrained features is an effective way to improve the task-specific semi-supervised learning on standard benchmarks. Code and pretrained models are made available at: https://github.com/srinidhiPY/SSL_CR_Histo.     

Local contrastive loss with pseudo-label based self-training for semi-supervised medical image segmentation

Supervised deep learning-based methods yield accurate results for medical image segmentation. However, they require large labeled datasets for this, and obtaining them is a laborious task that requires clinical expertise. Semi/self-supervised learning-based approaches address this limitation by exploiting unlabeled data along with limited annotated data. Recent self-supervised learning methods use contrastive loss to learn good global level representations from unlabeled images and achieve high performance in classification tasks on popular natural image datasets like ImageNet. In pixel-level prediction tasks such as segmentation, it is crucial to also learn good local level representations along with global representations to achieve better accuracy. However, the impact of the existing local contrastive loss-based methods remains limited for learning good local representations because similar and dissimilar local regions are defined based on random augmentations and spatial proximity; not based on the semantic label of local regions due to lack of large-scale expert annotations in the semi/self-supervised setting. In this paper, we propose a local contrastive loss to learn good pixel level features useful for segmentation by exploiting semantic label information obtained from pseudo-labels of unlabeled images alongside limited annotated images with ground truth (GT) labels. In particular, we define the proposed contrastive loss to encourage similar representations for the pixels that have the same pseudo-label/GT label while being dissimilar to the representation of pixels with different pseudo-label/GT label in the dataset. We perform pseudo-label based self-training and train the network by jointly optimizing the proposed contrastive loss on both labeled and unlabeled sets and segmentation loss on only the limited labeled set. We evaluated the proposed approach on three public medical datasets of cardiac and prostate anatomies, and obtain high segmentation performance with a limited labeled set of one or two 3D volumes. Extensive comparisons with the state-of-the-art semi-supervised and data augmentation methods and concurrent contrastive learning methods demonstrate the substantial improvement achieved by the proposed method. The code is made publicly available at https://github.com/krishnabits001/pseudo_label_contrastive_training.     

Source free domain adaptation for medical image segmentation with fourier style mining

Unsupervised domain adaptation (UDA) aims to exploit the knowledge learned from a labeled source dataset to solve similar tasks in a new unlabeled target domain. Existing UDA techniques typically assume that samples from source and target domains are freely accessible during the training. However, it may be impractical to access source images due to privacy concerns, especially in medical imaging scenarios with the patient information. To tackle this issue, we devise a novel source free domain adaptation framework with fourier style mining, where only a well-trained source segmentation model is available for the adaptation to the target domain. Our framework is composed of two stages: a generation stage and an adaptation stage. In the generation stage, we design a Fourier Style Mining (FSM) generator to inverse source-like images through statistic information of the pretrained source model and mutual Fourier Transform. These generated source-like images can provide source data distribution and benefit the domain alignment. In the adaptation stage, we design a Contrastive Domain Distillation (CDD) module to achieve feature-level adaptation, including a domain distillation loss to transfer relation knowledge and a domain contrastive loss to narrow down the domain gap by a self-supervised paradigm. Besides, a Compact-Aware Domain Consistency (CADC) module is proposed to enhance consistency learning by filtering out noisy pseudo labels with shape compactness metric, thus achieving output-level adaptation. Extensive experiments on cross-device and cross-centre datasets are conducted for polyp and prostate segmentation, and our method delivers impressive performance compared with state-of-the-art domain adaptation methods. The source code is available at https://github.com/CityU-AIM-Group/SFDA-FSM.     

Weakly supervised segmentation on neural compressed histopathology with self-equivariant regularization

In digital pathology, segmentation is a fundamental task for the diagnosis and treatment of diseases. Existing fully supervised methods often require accurate pixel-level annotations that are both time-consuming and laborious to generate. Typical approaches first pre-process histology images into patches to meet memory constraints and later perform stitching for segmentation; at times leading to lower performance given the lack of global context. Since image level labels are cheaper to acquire, weakly supervised learning is a more practical alternative for training segmentation algorithms. In this work, we present a weakly supervised framework for histopathology segmentation using only image-level labels by refining class activation maps (CAM) with self-supervision. First, we compress gigapixel histology images with an unsupervised contrastive learning technique to retain high-level spatial context. Second, a network is trained on the compressed images to jointly predict image-labels and refine the initial CAMs via self-supervised losses. In particular, we achieve refinement via a pixel correlation module (PCM) that leverages self-attention between the initial CAM and the input to encourage fine-grained activations. Also, we introduce a feature masking technique that performs spatial dropout on the compressed input to suppress low confidence predictions. To effectively train our model, we propose a loss function that includes a classification objective with image-labels, self-supervised regularization and entropy minimization between the CAM predictions. Experimental results on two curated datasets show that our approach is comparable to fully-supervised methods and can outperform existing state-of-the-art patch-based methods. https://github.com/PhilipChicco/wsshisto     

Semi-supervised medical image segmentation via uncertainty rectified pyramid consistency

Despite that Convolutional Neural Networks (CNNs) have achieved promising performance in many medical image segmentation tasks, they rely on a large set of labeled images for training, which is expensive and time-consuming to acquire. Semi-supervised learning has shown the potential to alleviate this challenge by learning from a large set of unlabeled images and limited labeled samples. In this work, we present a simple yet efficient consistency regularization approach for semi-supervised medical image segmentation, called Uncertainty Rectified Pyramid Consistency (URPC). Inspired by the pyramid feature network, we chose a pyramid-prediction network that obtains a set of segmentation predictions at different scales. For semi-supervised learning, URPC learns from unlabeled data by minimizing the discrepancy between each of the pyramid predictions and their average. We further present multi-scale uncertainty rectification to boost the pyramid consistency regularization, where the rectification seeks to temper the consistency loss at outlier pixels that may have substantially different predictions than the average, potentially due to upsampling errors or lack of enough labeled data. Experiments on two public datasets and an in-house clinical dataset showed that: 1) URPC can achieve large performance improvement by utilizing unlabeled data and 2) Compared with five existing semi-supervised methods, URPC achieved better or comparable results with a simpler pipeline. Furthermore, we build a semi-supervised medical image segmentation codebase to boost research on this topic: https://github.com/HiLab-git/SSL4MIS.     

Dual-distribution discrepancy with self-supervised refinement for anomaly detection in medical images

Medical anomaly detection is a crucial yet challenging task aimed at recognizing abnormal images to assist in diagnosis. Due to the high-cost annotations of abnormal images, most methods utilize only known normal images during training and identify samples deviating from the normal profile as anomalies in the testing phase. Many readily available unlabeled images containing anomalies are thus ignored in the training phase, restricting the performance. To solve this problem, we introduce one-class semi-supervised learning (OC-SSL) to utilize known normal and unlabeled images for training, and propose Dual-distribution Discrepancy for Anomaly Detection (DDAD) based on this setting. Ensembles of reconstruction networks are designed to model the distribution of normal images and the distribution of both normal and unlabeled images, deriving the normative distribution module (NDM) and unknown distribution module (UDM). Subsequently, the intra-discrepancy of NDM and inter-discrepancy between the two modules are designed as anomaly scores. Furthermore, we propose a new perspective on self-supervised learning, which is designed to refine the anomaly scores rather than directly detect anomalies. Five medical datasets, including chest X-rays, brain MRIs and retinal fundus images, are organized as benchmarks for evaluation. Experiments on these benchmarks comprehensively compare a wide range of anomaly detection methods and demonstrate that our method achieves significant gains and outperforms the state-of-the-art. Code and organized benchmarks are available at https://github.com/caiyu6666/DDAD-ASR.     

Unsupervised domain adaptation for clinician pose estimation and instance segmentation in the operating room

The fine-grained localization of clinicians in the operating room (OR) is a key component to design the new generation of OR support systems. Computer vision models for person pixel-based segmentation and body-keypoints detection are needed to better understand the clinical activities and the spatial layout of the OR. This is challenging, not only because OR images are very different from traditional vision datasets, but also because data and annotations are hard to collect and generate in the OR due to privacy concerns. To address these concerns, we first study how joint person pose estimation and instance segmentation can be performed on low resolutions images with downsampling factors from 1x to 12x. Second, to address the domain shift and the lack of annotations, we propose a novel unsupervised domain adaptation method, called AdaptOR, to adapt a model from an in-the-wild labeled source domain to a statistically different unlabeled target domain. We propose to exploit explicit geometric constraints on the different augmentations of the unlabeled target domain image to generate accurate pseudo labels and use these pseudo labels to train the model on high- and low-resolution OR images in a self-training framework. Furthermore, we propose disentangled feature normalization to handle the statistically different source and target domain data. Extensive experimental results with detailed ablation studies on the two OR datasets MVOR+ and TUM-OR-test show the effectiveness of our approach against strongly constructed baselines, especially on the low-resolution privacy-preserving OR images. Finally, we show the generality of our method as a semi-supervised learning (SSL) method on the large-scale COCO dataset, where we achieve comparable results with as few as 1% of labeled supervision against a model trained with 100% labeled supervision. Code is available at https://github.com/CAMMA-public/HPE-AdaptOR.     

Self-Supervised monocular depth and ego-Motion estimation in endoscopy: Appearance flow to the rescue

Recently, self-supervised learning technology has been applied to calculate depth and ego-motion from monocular videos, achieving remarkable performance in autonomous driving scenarios. One widely adopted assumption of depth and ego-motion self-supervised learning is that the image brightness remains constant within nearby frames. Unfortunately, the endoscopic scene does not meet this assumption because there are severe brightness fluctuations induced by illumination variations, non-Lambertian reflections and interreflections during data collection, and these brightness fluctuations inevitably deteriorate the depth and ego-motion estimation accuracy. In this work, we introduce a novel concept referred to as appearance flow to address the brightness inconsistency problem. The appearance flow takes into consideration any variations in the brightness pattern and enables us to develop a generalized dynamic image constraint. Furthermore, we build a unified self-supervised framework to estimate monocular depth and ego-motion simultaneously in endoscopic scenes, which comprises a structure module, a motion module, an appearance module and a correspondence module, to accurately reconstruct the appearance and calibrate the image brightness. Extensive experiments are conducted on the SCARED dataset and EndoSLAM dataset, and the proposed unified framework exceeds other self-supervised approaches by a large margin. To validate our framework’s generalization ability on different patients and cameras, we train our model on SCARED but test it on the SERV-CT and Hamlyn datasets without any fine-tuning, and the superior results reveal its strong generalization ability. Code is available at: https://github.com/ShuweiShao/AF-SfMLearner.     

Models Genesis

Transfer learning from natural images to medical images has been established as one of the most practical paradigms in deep learning for medical image analysis. To fit this paradigm, however, 3D imaging tasks in the most prominent imaging modalities (e.g., CT and MRI) have to be reformulated and solved in 2D, losing rich 3D anatomical information, thereby inevitably compromising its performance. To overcome this limitation, we have built a set of models, called Generic Autodidactic Models, nicknamed Models Genesis, because they are created ex nihilo (with no manual labeling), self-taught (learnt by self-supervision), and generic (served as source models for generating application-specific target models). Our extensive experiments demonstrate that our Models Genesis significantly outperform learning from scratch and existing pre-trained 3D models in all five target 3D applications covering both segmentation and classification. More importantly, learning a model from scratch simply in 3D may not necessarily yield performance better than transfer learning from ImageNet in 2D, but our Models Genesis consistently top any 2D/2.5D approaches including fine-tuning the models pre-trained from ImageNet as well as fine-tuning the 2D versions of our Models Genesis, confirming the importance of 3D anatomical information and significance of Models Genesis for 3D medical imaging. This performance is attributed to our unified self-supervised learning framework, built on a simple yet powerful observation: the sophisticated and recurrent anatomy in medical images can serve as strong yet free supervision signals for deep models to learn common anatomical representation automatically via self-supervision. As open science, all codes and pre-trained Models Genesis are available at https://github.com/MrGiovanni/ModelsGenesis.     

Meta multi-task nuclei segmentation with fewer training samples

Cells/nuclei deliver massive information of microenvironment. An automatic nuclei segmentation approach can reduce pathologists’ workload and allow precise of the microenvironment for biological and clinical researches. Existing deep learning models have achieved outstanding performance under the supervision of a large amount of labeled data. However, when data from the unseen domain comes, we still have to prepare a certain degree of manual annotations for training for each domain. Unfortunately, obtaining histopathological annotations is extremely difficult. It is high expertise-dependent and time-consuming. In this paper, we attempt to build a generalized nuclei segmentation model with less data dependency and more generalizability. To this end, we propose a meta multi-task learning (Meta-MTL) model for nuclei segmentation which requires fewer training samples. A model-agnostic meta-learning is applied as the outer optimization algorithm for the segmentation model. We introduce a contour-aware multi-task learning model as the inner model. A feature fusion and interaction block (FFIB) is proposed to allow feature communication across both tasks. Extensive experiments prove that our proposed Meta-MTL model can improve the model generalization and obtain a comparable performance with state-of-the-art models with fewer training samples. Our model can also perform fast adaptation on the unseen domain with only a few manual annotations. Code is available at https://github.com/ChuHan89/Meta-MTL4NucleiSegmentation     

Dissecting self-supervised learning methods for surgical computer vision

The field of surgical computer vision has undergone considerable breakthroughs in recent years with the rising popularity of deep neural network-based methods. However, standard fully-supervised approaches for training such models require vast amounts of annotated data, imposing a prohibitively high cost; especially in the clinical domain. Self-Supervised Learning (SSL) methods, which have begun to gain traction in the general computer vision community, represent a potential solution to these annotation costs, allowing to learn useful representations from only unlabeled data. Still, the effectiveness of SSL methods in more complex and impactful domains, such as medicine and surgery, remains limited and unexplored. In this work, we address this critical need by investigating four state-of-the-art SSL methods (MoCo v2, SimCLR, DINO, SwAV) in the context of surgical computer vision. We present an extensive analysis of the performance of these methods on the Cholec80 dataset for two fundamental and popular tasks in surgical context understanding, phase recognition and tool presence detection. We examine their parameterization, then their behavior with respect to training data quantities in semi-supervised settings. Correct transfer of these methods to surgery, as described and conducted in this work, leads to substantial performance gains over generic uses of SSL – up to 7.4% on phase recognition and 20% on tool presence detection – as well as state-of-the-art semi-supervised phase recognition approaches by up to 14%. Further results obtained on a highly diverse selection of surgical datasets exhibit strong generalization properties. The code is available at https://github.com/CAMMA-public/SelfSupSurg.     

Mutual consistency learning for semi-supervised medical image segmentation

In this paper, we propose a novel mutual consistency network (MC-Net+) to effectively exploit the unlabeled data for semi-supervised medical image segmentation. The MC-Net+ model is motivated by the observation that deep models trained with limited annotations are prone to output highly uncertain and easily mis-classified predictions in the ambiguous regions (e.g., adhesive edges or thin branches) for medical image segmentation. Leveraging these challenging samples can make the semi-supervised segmentation model training more effective. Therefore, our proposed MC-Net+ model consists of two new designs. First, the model contains one shared encoder and multiple slightly different decoders (i.e., using different up-sampling strategies). The statistical discrepancy of multiple decoders’ outputs is computed to denote the model’s uncertainty, which indicates the unlabeled hard regions. Second, we apply a novel mutual consistency constraint between one decoder’s probability output and other decoders’ soft pseudo labels. In this way, we minimize the discrepancy of multiple outputs (i.e., the model uncertainty) during training and force the model to generate invariant results in such challenging regions, aiming at regularizing the model training. We compared the segmentation results of our MC-Net+ model with five state-of-the-art semi-supervised approaches on three public medical datasets. Extension experiments with two standard semi-supervised settings demonstrate the superior performance of our model over other methods, which sets a new state of the art for semi-supervised medical image segmentation. Our code is released publicly at https://github.com/ycwu1997/MC-Net.     

Volumetric white matter tract segmentation with nested self-supervised learning using sequential pretext tasks

White matter (WM) tract segmentation based on diffusion magnetic resonance imaging (dMRI) provides an important tool for the analysis of brain development, function, and disease. Deep learning based methods of WM tract segmentation have been proposed, which greatly improve the accuracy of the segmentation. However, the training of the deep networks usually requires a large number of manual delineations of WM tracts, which can be especially difficult to obtain and unavailable in many scenarios. Therefore, in this work, we explore how to perform deep learning based WM tract segmentation when annotated training data is scarce. To this end, we seek to exploit the abundant unannotated dMRI data in the self-supervised learning framework. From the unannotated data, knowledge about image context can be learned with pretext tasks that do not require manual annotations. Specifically, a deep network can be pretrained for the pretext task, and the knowledge learned from the pretext task is then transferred to the subsequent WM tract segmentation task with only a small number of annotated scans via fine-tuning. We explore two designs of pretext tasks that are related to WM tracts. The first pretext task predicts the density map of fiber streamlines, which are representations of generic WM pathways, and the training data can be obtained automatically with tractography. The second pretext task learns to mimic the results of registration-based WM tract segmentation, which, although inaccurate, is more relevant to WM tract segmentation and provides a good target for learning context knowledge. Then, we combine the two pretext tasks and develop a nested self-supervised learning strategy. In the nested self-supervised learning strategy, the first pretext task provides initial knowledge for the second pretext task, and the knowledge learned from the second pretext task with the initial knowledge is transferred to the target WM tract segmentation task via fine-tuning. To evaluate the proposed method, experiments were performed on brain dMRI scans from the Human Connectome Project dataset with various experimental settings. The results show that the proposed method improves the performance of WM tract segmentation when tract annotations are scarce.     

Real-time echocardiography image analysis and quantification of cardiac indices

Deep learning has a huge potential to transform echocardiography in clinical practice and point of care ultrasound testing by providing real-time analysis of cardiac structure and function. Automated echocardiography analysis is benefited through use of machine learning for tasks such as image quality assessment, view classification, cardiac region segmentation, and quantification of diagnostic indices. By taking advantage of high-performing deep neural networks, we propose a novel and eicient real-time system for echocardiography analysis and quantification. Our system uses a self-supervised modality-specific representation trained using a publicly available large-scale dataset. The trained representation is used to enhance the learning of target echo tasks with relatively small datasets. We also present a novel Trilateral Attention Network (TaNet) for real-time cardiac region segmentation. The proposed network uses a module for region localization and three lightweight pathways for encoding rich low-level, textural, and high-level features. Feature embeddings from these individual pathways are then aggregated for cardiac region segmentation. This network is fine-tuned using a joint loss function and training strategy. We extensively evaluate the proposed system and its components, which are echo view retrieval, cardiac segmentation, and quantification, using four echocardiography datasets. Our experimental results show a consistent improvement in the performance of echocardiography analysis tasks with enhanced computational eiciency that charts a path toward its adoption in clinical practice. Specifically, our results show superior real-time performance in retrieving good quality echo from individual cardiac view, segmenting cardiac chambers with complex overlaps, and extracting cardiac indices that highly agree with the experts’ values. The source code of our implementation can be found in the project’s GitHub page.     

Self-rule to multi-adapt: Generalized multi-source feature learning using unsupervised domain adaptation for colorectal cancer tissue detection

Supervised learning is constrained by the availability of labeled data, which are especially expensive to acquire in the field of digital pathology. Making use of open-source data for pre-training or using domain adaptation can be a way to overcome this issue. However, pre-trained networks often fail to generalize to new test domains that are not distributed identically due to tissue stainings, types, and textures variations. Additionally, current domain adaptation methods mainly rely on fully-labeled source datasets. In this work, we propose Self-Rule to Multi-Adapt (SRMA), which takes advantage of self-supervised learning to perform domain adaptation, and removes the necessity of fully-labeled source datasets. SRMA can effectively transfer the discriminative knowledge obtained from a few labeled source domain’s data to a new target domain without requiring additional tissue annotations. Our method harnesses both domains’ structures by capturing visual similarity with intra-domain and cross-domain self-supervision. Moreover, we present a generalized formulation of our approach that allows the framework to learn from multiple source domains. We show that our proposed method outperforms baselines for domain adaptation of colorectal tissue type classification in single and multi-source settings, and further validate our approach on an in-house clinical cohort. The code and trained models are available open-source: https://github.com/christianabbet/SRA.     

Histogram of Oriented Gradients meet deep learning: A novel multi-task deep network for 2D surgical image semantic segmentation

We present our novel deep multi-task learning method for medical image segmentation. Existing multi-task methods demand ground truth annotations for both the primary and auxiliary tasks. Contrary to it, we propose to generate the pseudo-labels of an auxiliary task in an unsupervised manner. To generate the pseudo-labels, we leverage Histogram of Oriented Gradients (HOGs), one of the most widely used and powerful hand-crafted features for detection. Together with the ground truth semantic segmentation masks for the primary task and pseudo-labels for the auxiliary task, we learn the parameters of the deep network to minimize the loss of both the primary task and the auxiliary task jointly. We employed our method on two powerful and widely used semantic segmentation networks: UNet and U2Net to train in a multi-task setup. To validate our hypothesis, we performed experiments on two different medical image segmentation data sets. From the extensive quantitative and qualitative results, we observe that our method consistently improves the performance compared to the counter-part method. Moreover, our method is the winner of FetReg Endovis Sub-challenge on Semantic Segmentation organised in conjunction with MICCAI 2021. Code and implementation details are available at:https://github.com/thetna/medical_image_segmentation.     

Self-supervised learning of neighborhood embedding for longitudinal MRI

In recent years, several deep learning models recommend first to represent Magnetic Resonance Imaging (MRI) as latent features before performing a downstream task of interest (such as classification or regression). The performance of the downstream task generally improves when these latent representations are explicitly associated with factors of interest. For example, we derived such a representation for capturing brain aging by applying self-supervised learning to longitudinal MRIs and then used the resulting encoding to automatically identify diseases accelerating the aging of the brain. We now propose a refinement of this representation by replacing the linear modeling of brain aging with one that is consistent in local neighborhoods in the latent space. Called Longitudinal Neighborhood Embedding (LNE), we derive an encoding so that neighborhoods are age-consistent (i.e., brain MRIs of different subjects with similar brain ages are in close proximity of each other) and progression-consistent, i.e., the latent space is defined by a smooth trajectory field where each trajectory captures changes in brain ages between a pair of MRIs extracted from a longitudinal sequence. To make the problem computationally tractable, we further propose a strategy for mini-batch sampling so that the resulting local neighborhoods accurately approximate the ones that would be defined based on the whole cohort.We evaluate LNE on three different downstream tasks: (1) to predict chronological age from T1-w MRI of 274 healthy subjects participating in a study at SRI International; (2) to distinguish Normal Control (NC) from Alzheimer’s Disease (AD) and stable Mild Cognitive Impairment (sMCI) from progressive Mild Cognitive Impairment (pMCI) based on T1-w MRI of 632 participants of the Alzheimer’s Disease Neuroimaging Initiative (ADNI); and (3) to distinguish no-to-low from moderate-to-heavy alcohol drinkers based on fractional anisotropy derived from diffusion tensor MRIs of 764 adolescents recruited by the National Consortium on Alcohol and NeuroDevelopment in Adolescence (NCANDA). Across the three data sets, the visualization of the smooth trajectory vector fields and superior accuracy on downstream tasks demonstrate the strength of the proposed method over existing self-supervised methods in extracting information related to brain aging, which could help study the impact of substance use and neurodegenerative disorders. The code is available at https://github.com/ouyangjiahong/longitudinal-neighbourhood-embedding.     

Segmentation with mixed supervision: Confidence maximization helps knowledge distillation

Despite achieving promising results in a breadth of medical image segmentation tasks, deep neural networks (DNNs) require large training datasets with pixel-wise annotations. Obtaining these curated datasets is a cumbersome process which limits the applicability of DNNs in scenarios where annotated images are scarce. Mixed supervision is an appealing alternative for mitigating this obstacle. In this setting, only a small fraction of the data contains complete pixel-wise annotations and other images have a weaker form of supervision, e.g., only a handful of pixels are labeled. In this work, we propose a dual-branch architecture, where the upper branch (teacher) receives strong annotations, while the bottom one (student) is driven by limited supervision and guided by the upper branch. Combined with a standard cross-entropy loss over the labeled pixels, our novel formulation integrates two important terms: (i) a Shannon entropy loss defined over the less-supervised images, which encourages confident student predictions in the bottom branch; and (ii) a Kullback–Leibler (KL) divergence term, which transfers the knowledge (i.e., predictions) of the strongly supervised branch to the less-supervised branch and guides the entropy (student-confidence) term to avoid trivial solutions. We show that the synergy between the entropy and KL divergence yields substantial improvements in performance. We also discuss an interesting link between Shannon-entropy minimization and standard pseudo-mask generation, and argue that the former should be preferred over the latter for leveraging information from unlabeled pixels. We evaluate the effectiveness of the proposed formulation through a series of quantitative and qualitative experiments using two publicly available datasets. Results demonstrate that our method significantly outperforms other strategies for semantic segmentation within a mixed-supervision framework, as well as recent semi-supervised approaches. Moreover, in line with recent observations in classification, we show that the branch trained with reduced supervision and guided by the top branch largely outperforms the latter. Our code is publicly available: https://github.com/by-liu/ConfKD.     

Information bottleneck-based interpretable multitask network for breast cancer classification and segmentation

Breast cancer is one of the most common causes of death among women worldwide. Early signs of breast cancer can be an abnormality depicted on breast images (e.g., mammography or breast ultrasonography). However, reliable interpretation of breast images requires intensive labor and physicians with extensive experience. Deep learning is evolving breast imaging diagnosis by introducing a second opinion to physicians. However, most deep learning-based breast cancer analysis algorithms lack interpretability because of their black box nature, which means that domain experts cannot understand why the algorithms predict a label. In addition, most deep learning algorithms are formulated as a single-task-based model that ignores correlations between different tasks (e.g., tumor classification and segmentation). In this paper, we propose an interpretable multitask information bottleneck network (MIB-Net) to accomplish simultaneous breast tumor classification and segmentation. MIB-Net maximizes the mutual information between the latent representations and class labels while minimizing information shared by the latent representations and inputs. In contrast from existing models, our MIB-Net generates a contribution score map that offers an interpretable aid for physicians to understand the model’s decision-making process. In addition, MIB-Net implements multitask learning and further proposes a dual prior knowledge guidance strategy to enhance deep task correlation. Our evaluations are carried out on three breast image datasets in different modalities. Our results show that the proposed framework is not only able to help physicians better understand the model’s decisions but also improve breast tumor classification and segmentation accuracy over representative state-of-the-art models. Our code is available at https://github.com/jxw0810/MIB-Net.