In this work, we utilize image segmentation to visually identify blood vessels in retinal examination images. This process is typically carried out manually. However, we can employ heuristic methods and machine learning to automate or at least expedite the process. In this context, we propose a cross-platform, open-source, and responsive software that allows users to manually segment a retinal image. The purpose is to use the user-segmented image to retrain machine learning algorithms, thereby enhancing future automated segmentation results. Moreover, the software also incorporates and applies certain image filters established in the literature to improve vessel visualization. We propose the first solution of this kind in the literature. This is the inaugural integrated software that embodies the aforementioned attributes: open-source, responsive, and cross-platform. It offers a comprehensive solution encompassing manual vessel segmentation, as well as the automated execution of classification algorithms to refine predictive models.
The Segment Anything Model (SAM) represents a state-of-the-art research advancement in natural image segmentation, achieving impressive results with input prompts such as points and bounding boxes. However, our evaluation and recent research indicate that directly applying the pretrained SAM to medical image segmentation does not yield satisfactory performance. This limitation primarily arises from significant domain gap between natural images and medical images. To bridge this gap, we introduce SAM-Med2D, the most comprehensive studies on applying SAM to medical 2D images. Specifically, we first collect and curate approximately 4.6M images and 19.7M masks from public and private datasets, constructing a large-scale medical image segmentation dataset encompassing various modalities and objects. Then, we comprehensively fine-tune SAM on this dataset and turn it into SAM-Med2D. Unlike previous methods that only adopt bounding box or point prompts as interactive segmentation approach, we adapt SAM to medical image segmentation through more comprehensive prompts involving bounding boxes, points, and masks. We additionally fine-tune the encoder and decoder of the original SAM to obtain a well-performed SAM-Med2D, leading to the most comprehensive fine-tuning strategies to date. Finally, we conducted a comprehensive evaluation and analysis to investigate the performance of SAM-Med2D in medical image segmentation across various modalities, anatomical structures, and organs. Concurrently, we validated the generalization capability of SAM-Med2D on 9 datasets from MICCAI 2023 challenge. Overall, our approach demonstrated significantly superior performance and generalization capability compared to SAM.
Unsupervised anomaly segmentation aims to detect patterns that are distinct from any patterns processed during training, commonly called abnormal or out-of-distribution patterns, without providing any associated manual segmentations. Since anomalies during deployment can lead to model failure, detecting the anomaly can enhance the reliability of models, which is valuable in high-risk domains like medical imaging. This paper introduces Masked Modality Cycles with Conditional Diffusion (MMCCD), a method that enables segmentation of anomalies across diverse patterns in multimodal MRI. The method is based on two fundamental ideas. First, we propose the use of cyclic modality translation as a mechanism for enabling abnormality detection. Image-translation models learn tissue-specific modality mappings, which are characteristic of tissue physiology. Thus, these learned mappings fail to translate tissues or image patterns that have never been encountered during training, and the error enables their segmentation. Furthermore, we combine image translation with a masked conditional diffusion model, which attempts to `imagine' what tissue exists under a masked area, further exposing unknown patterns as the generative model fails to recreate them. We evaluate our method on a proxy task by training on healthy-looking slices of BraTS2021 multi-modality MRIs and testing on slices with tumors. We show that our method compares favorably to previous unsupervised approaches based on image reconstruction and denoising with autoencoders and diffusion models.
Both CNN-based and Transformer-based object detection with bounding box representation have been extensively studied in computer vision and medical image analysis, but circular object detection in medical images is still underexplored. Inspired by the recent anchor free CNN-based circular object detection method (CircleNet) for ball-shape glomeruli detection in renal pathology, in this paper, we present CircleFormer, a Transformer-based circular medical object detection with dynamic anchor circles. Specifically, queries with circle representation in Transformer decoder iteratively refine the circular object detection results, and a circle cross attention module is introduced to compute the similarity between circular queries and image features. A generalized circle IoU (gCIoU) is proposed to serve as a new regression loss of circular object detection as well. Moreover, our approach is easy to generalize to the segmentation task by adding a simple segmentation branch to CircleFormer. We evaluate our method in circular nuclei detection and segmentation on the public MoNuSeg dataset, and the experimental results show that our method achieves promising performance compared with the state-of-the-art approaches. The effectiveness of each component is validated via ablation studies as well. Our code is released at https://github.com/zhanghx-iim-ahu/CircleFormer.
Small liver lesions common to colorectal liver metastases (CRLMs) are challenging for convolutional neural network (CNN) segmentation models, especially when we have a wide range of slice thicknesses in the computed tomography (CT) scans. Slice thickness of CT images may vary by clinical indication. For example, thinner slices are used for presurgical planning when fine anatomic details of small vessels are required. While keeping the effective radiation dose in patients as low as possible, various slice thicknesses are employed in CRLMs due to their limitations. However, differences in slice thickness across CTs lead to significant performance degradation in CT segmentation models based on CNNs. This paper proposes a novel unsupervised attention-based interpolation model to generate intermediate slices from consecutive triplet slices in CT scans. We integrate segmentation loss during the interpolation model's training to leverage segmentation labels in existing slices to generate middle ones. Unlike common interpolation techniques in CT volumes, our model highlights the regions of interest (liver and lesions) inside the abdominal CT scans in the interpolated slice. Moreover, our model's outputs are consistent with the original input slices while increasing the segmentation performance in two cutting-edge 3D segmentation pipelines. We tested the proposed model on the CRLM dataset to upsample subjects with thick slices and create isotropic volume for our segmentation model. The produced isotropic dataset increases the Dice score in the segmentation of lesions and outperforms other interpolation approaches in terms of interpolation metrics.
Multi-centre colonoscopy images from various medical centres exhibit distinct complicating factors and overlays that impact the image content, contingent on the specific acquisition centre. Existing Deep Segmentation networks struggle to achieve adequate generalizability in such data sets, and the currently available data augmentation methods do not effectively address these sources of data variability. As a solution, we introduce an innovative data augmentation approach centred on interpretability saliency maps, aimed at enhancing the generalizability of Deep Learning models within the realm of multi-centre colonoscopy image segmentation. The proposed augmentation technique demonstrates increased robustness across different segmentation models and domains. Thorough testing on a publicly available multi-centre dataset for polyp detection demonstrates the effectiveness and versatility of our approach, which is observed both in quantitative and qualitative results. The code is publicly available at: https://github.com/nki-radiology/interpretability_augmentation
A person's facial hairstyle, such as presence and size of beard, can significantly impact face recognition accuracy. There are publicly-available deep networks that achieve reasonable accuracy at binary attribute classification, such as beard / no beard, but few if any that segment the facial hair region. To investigate the effect of facial hair in a rigorous manner, we first created a set of fine-grained facial hair annotations to train a segmentation model and evaluate its accuracy across African-American and Caucasian face images. We then use our facial hair segmentations to categorize image pairs according to the degree of difference or similarity in the facial hairstyle. We find that the False Match Rate (FMR) for image pairs with different categories of facial hairstyle varies by a factor of over 10 for African-American males and over 25 for Caucasian males. To reduce the bias across image pairs with different facial hairstyles, we propose a scheme for adaptive thresholding based on facial hairstyle similarity. Evaluation on a subject-disjoint set of images shows that adaptive similarity thresholding based on facial hairstyles of the image pair reduces the ratio between the highest and lowest FMR across facial hairstyle categories for African-American from 10.7 to 1.8 and for Caucasians from 25.9 to 1.3. Facial hair annotations and facial hair segmentation model will be publicly available.
Referring image segmentation, the task of segmenting any arbitrary entities described in free-form texts, opens up a variety of vision applications. However, manual labeling of training data for this task is prohibitively costly, leading to lack of labeled data for training. We address this issue by a weakly supervised learning approach using text descriptions of training images as the only source of supervision. To this end, we first present a new model that discovers semantic entities in input image and then combines such entities relevant to text query to predict the mask of the referent. We also present a new loss function that allows the model to be trained without any further supervision. Our method was evaluated on four public benchmarks for referring image segmentation, where it clearly outperformed the existing method for the same task and recent open-vocabulary segmentation models on all the benchmarks.
We present 'CongNaMul', a comprehensive dataset designed for various tasks in soybean sprouts image analysis. The CongNaMul dataset is curated to facilitate tasks such as image classification, semantic segmentation, decomposition, and measurement of length and weight. The classification task provides four classes to determine the quality of soybean sprouts: normal, broken, spotted, and broken and spotted, for the development of AI-aided automatic quality inspection technology. For semantic segmentation, images with varying complexity, from single sprout images to images with multiple sprouts, along with human-labelled mask images, are included. The label has 4 different classes: background, head, body, tail. The dataset also provides images and masks for the image decomposition task, including two separate sprout images and their combined form. Lastly, 5 physical features of sprouts (head length, body length, body thickness, tail length, weight) are provided for image-based measurement tasks. This dataset is expected to be a valuable resource for a wide range of research and applications in the advanced analysis of images of soybean sprouts. Also, we hope that this dataset can assist researchers studying classification, semantic segmentation, decomposition, and physical feature measurement in other industrial fields, in evaluating their models. The dataset is available at the authors' repository. (https://bhban.kr/data)
Thyroid nodule segmentation is a crucial step in the diagnostic procedure of physicians and computer-aided diagnosis systems. Mostly, current studies treat segmentation and diagnosis as independent tasks without considering the correlation between these tasks. The sequence steps of these independent tasks in computer-aided diagnosis systems may lead to the accumulation of errors. Therefore, it is worth combining them as a whole through exploring the relationship between thyroid nodule segmentation and diagnosis. According to the thyroid imaging reporting and data system (TI-RADS), the assessment of shape and margin characteristics is the prerequisite for the discrimination of benign and malignant thyroid nodules. These characteristics can be observed in the thyroid nodule segmentation masks. Inspired by the diagnostic procedure of TI-RADS, this paper proposes a shape-margin knowledge augmented network (SkaNet) for simultaneously thyroid nodule segmentation and diagnosis. Due to the similarity in visual features between segmentation and diagnosis, SkaNet shares visual features in the feature extraction stage and then utilizes a dual-branch architecture to perform thyroid nodule segmentation and diagnosis tasks simultaneously. To enhance effective discriminative features, an exponential mixture module is devised, which incorporates convolutional feature maps and self-attention maps by exponential weighting. Then, SkaNet is jointly optimized by a knowledge augmented multi-task loss function with a constraint penalty term. It embeds shape and margin characteristics through numerical computation and models the relationship between the thyroid nodule diagnosis results and segmentation masks.
Until recently, the Video Instance Segmentation (VIS) community operated under the common belief that offline methods are generally superior to a frame by frame online processing. However, the recent success of online methods questions this belief, in particular, for challenging and long video sequences. We understand this work as a rebuttal of those recent observations and an appeal to the community to focus on dedicated near-online VIS approaches. To support our argument, we present a detailed analysis on different processing paradigms and the new end-to-end trainable NOVIS (Near-Online Video Instance Segmentation) method. Our transformer-based model directly predicts spatio-temporal mask volumes for clips of frames and performs instance tracking between clips via overlap embeddings. NOVIS represents the first near-online VIS approach which avoids any handcrafted tracking heuristics. We outperform all existing VIS methods by large margins and provide new state-of-the-art results on both YouTube-VIS (2019/2021) and the OVIS benchmarks.
As recent advances in AI are causing the decline of conventional diagnostic methods, the realization of end-to-end diagnosis is fast approaching. Ultrasound image segmentation is an important step in the diagnostic process. An accurate and robust segmentation model accelerates the process and reduces the burden of sonographers. In contrast to previous research, we take two inherent features of ultrasound images into consideration: (1) different organs and tissues vary in spatial sizes, (2) the anatomical structures inside human body form a relatively constant spatial relationship. Based on those two ideas, we propose a new image segmentation model combining Feature Pyramid Network (FPN) and Spatial Recurrent Neural Network (SRNN). We discuss why we use FPN to extract anatomical structures of different scales and how SRNN is implemented to extract the spatial context features in abdominal ultrasound images.
We present DySample, an ultra-lightweight and effective dynamic upsampler. While impressive performance gains have been witnessed from recent kernel-based dynamic upsamplers such as CARAFE, FADE, and SAPA, they introduce much workload, mostly due to the time-consuming dynamic convolution and the additional sub-network used to generate dynamic kernels. Further, the need for high-res feature guidance of FADE and SAPA somehow limits their application scenarios. To address these concerns, we bypass dynamic convolution and formulate upsampling from the perspective of point sampling, which is more resource-efficient and can be easily implemented with the standard built-in function in PyTorch. We first showcase a naive design, and then demonstrate how to strengthen its upsampling behavior step by step towards our new upsampler, DySample. Compared with former kernel-based dynamic upsamplers, DySample requires no customized CUDA package and has much fewer parameters, FLOPs, GPU memory, and latency. Besides the light-weight characteristics, DySample outperforms other upsamplers across five dense prediction tasks, including semantic segmentation, object detection, instance segmentation, panoptic segmentation, and monocular depth estimation. Code is available at https://github.com/tiny-smart/dysample.
Segment Anything Model (SAM) is one of the pioneering prompt-based foundation models for image segmentation and has been rapidly adopted for various medical imaging applications. However, in clinical settings, creating effective prompts is notably challenging and time-consuming, requiring the expertise of domain specialists such as physicians. This requirement significantly diminishes SAM's primary advantage, its interactive capability with end users, in medical applications. Moreover, recent studies have indicated that SAM, originally designed for 2D natural images, performs suboptimally on 3D medical image segmentation tasks. This subpar performance is attributed to the domain gaps between natural and medical images and the disparities in spatial arrangements between 2D and 3D images, particularly in multi-organ segmentation applications. To overcome these challenges, we present a novel technique termed AutoProSAM. This method automates 3D multi-organ CT-based segmentation by leveraging SAM's foundational model capabilities without relying on domain experts for prompts. The approach utilizes parameter-efficient adaptation techniques to adapt SAM for 3D medical imagery and incorporates an effective automatic prompt learning paradigm specific to this domain. By eliminating the need for manual prompts, it enhances SAM's capabilities for 3D medical image segmentation and achieves state-of-the-art (SOTA) performance in CT-based multi-organ segmentation tasks. The code is in this {\href{https://github.com/ChengyinLee/AutoProSAM_2024}{link}}.
Active Learning (AL) for semantic segmentation is challenging due to heavy class imbalance and different ways of defining "sample" (pixels, areas, etc.), leaving the interpretation of the data distribution ambiguous. We propose "Maturity-Aware Distribution Breakdown-based Active Learning'' (MADBAL), an AL method that benefits from a hierarchical approach to define a multiview data distribution, which takes into account the different "sample" definitions jointly, hence able to select the most impactful segmentation pixels with comprehensive understanding. MADBAL also features a novel uncertainty formulation, where AL supporting modules are included to sense the features' maturity whose weighted influence continuously contributes to the uncertainty detection. In this way, MADBAL makes significant performance leaps even in the early AL stage, hence reducing the training burden significantly. It outperforms state-of-the-art methods on Cityscapes and PASCAL VOC datasets as verified in our extensive experiments.
We address the task of supervised action segmentation which aims to partition a video into non-overlapping segments, each representing a different action. Recent works apply transformers to perform temporal modeling at the frame-level, which suffer from high computational cost and cannot well capture action dependencies over long temporal horizons. To address these issues, we propose an efficient BI-level Temporal modeling (BIT) framework that learns explicit action tokens to represent action segments, in parallel performs temporal modeling on frame and action levels, while maintaining a low computational cost. Our model contains (i) a frame branch that uses convolution to learn frame-level relationships, (ii) an action branch that uses transformer to learn action-level dependencies with a small set of action tokens and (iii) cross-attentions to allow communication between the two branches. We apply and extend a set-prediction objective to allow each action token to represent one or multiple action segments, thus can avoid learning a large number of tokens over long videos with many segments. Thanks to the design of our action branch, we can also seamlessly leverage textual transcripts of videos (when available) to help action segmentation by using them to initialize the action tokens. We evaluate our model on four video datasets (two egocentric and two third-person) for action segmentation with and without transcripts, showing that BIT significantly improves the state-of-the-art accuracy with much lower computational cost (30 times faster) compared to existing transformer-based methods.
Since real-world training datasets cannot properly sample the long tail of the underlying data distribution, corner cases and rare out-of-domain samples can severely hinder the performance of state-of-the-art models. This problem becomes even more severe for dense tasks, such as 3D semantic segmentation, where points of non-standard objects can be confidently associated to the wrong class. In this work, we focus on improving the generalization to out-of-domain data. We achieve this by augmenting the training set with adversarial examples. First, we learn a set of vectors that deform the objects in an adversarial fashion. To prevent the adversarial examples from being too far from the existing data distribution, we preserve their plausibility through a series of constraints, ensuring sensor-awareness and shapes smoothness. Then, we perform adversarial augmentation by applying the learned sample-independent vectors to the available objects when training a model. We conduct extensive experiments across a variety of scenarios on data from KITTI, Waymo, and CrashD for 3D object detection, and on data from SemanticKITTI, Waymo, and nuScenes for 3D semantic segmentation. Despite training on a standard single dataset, our approach substantially improves the robustness and generalization of both 3D object detection and 3D semantic segmentation methods to out-of-domain data.
The coincidence similarity index, based on a combination of the Jaccard and overlap similarity indices, has noticeable properties in comparing and classifying data, including enhanced selectivity and sensitivity, intrinsic normalization, and robustness to data perturbations and outliers. These features allow multiset neurons, which are based on the coincidence similarity operation, to perform effective pattern recognition applications, including the challenging task of image segmentation. A few prototype points have been used in previous related approaches to represent each pattern to be identified, each of them being associated with respective multiset neurons. The segmentation of the regions can then proceed by taking into account the outputs of these neurons. The present work describes multilayer multiset neuronal networks incorporating two or more layers of coincidence similarity neurons. In addition, as a means to improve performance, this work also explores the utilization of counter-prototype points, which are assigned to the image regions to be avoided. This approach is shown to allow effective segmentation of complex regions despite considering only one prototype and one counter-prototype point. As reported here, the balanced accuracy landscapes to be optimized in order to identify the weight of the neurons in subsequent layers have been found to be relatively smooth, while typically involving more than one attraction basin. The use of a simple gradient-based optimization methodology has been demonstrated to effectively train the considered neural networks with several architectures, at least for the given data type, configuration of parameters, and network architecture.
Transfer learning improves the performance of deep learning models by initializing them with parameters pre-trained on larger datasets. Intuitively, transfer learning is more effective when pre-training is on the in-domain datasets. A recent study by NASA has demonstrated that the microstructure segmentation with encoder-decoder algorithms benefits more from CNN encoders pre-trained on microscopy images than from those pre-trained on natural images. However, CNN models only capture the local spatial relations in images. In recent years, attention networks such as Transformers are increasingly used in image analysis to capture the long-range relations between pixels. In this study, we compare the segmentation performance of Transformer and CNN models pre-trained on microscopy images with those pre-trained on natural images. Our result partially confirms the NASA study that the segmentation performance of out-of-distribution images (taken under different imaging and sample conditions) is significantly improved when pre-training on microscopy images. However, the performance gain for one-shot and few-shot learning is more modest with Transformers. We also find that for image segmentation, the combination of pre-trained Transformers and CNN encoders are consistently better than pre-trained CNN encoders alone. Our dataset (of about 50,000 images) combines the public portion of the NASA dataset with additional images we collected. Even with much less training data, our pre-trained models have significantly better performance for image segmentation. This result suggests that Transformers and CNN complement each other and when pre-trained on microscopy images, they are more beneficial to the downstream tasks.
We present a novel confidence refinement scheme that enhances pseudo labels in semi-supervised semantic segmentation. Unlike existing methods, which filter pixels with low-confidence predictions in isolation, our approach leverages the spatial correlation of labels in segmentation maps by grouping neighboring pixels and considering their pseudo labels collectively. With this contextual information, our method, named S4MC, increases the amount of unlabeled data used during training while maintaining the quality of the pseudo labels, all with negligible computational overhead. Through extensive experiments on standard benchmarks, we demonstrate that S4MC outperforms existing state-of-the-art semi-supervised learning approaches, offering a promising solution for reducing the cost of acquiring dense annotations. For example, S4MC achieves a 1.39 mIoU improvement over the prior art on PASCAL VOC 12 with 366 annotated images. The code to reproduce our experiments is available at https://s4mcontext.github.io/