Seong Tae Kim

Tobias Czempiel, Coco Rogers, Matthias Keicher et al. · stanford

Quantifying COVID-19 infection over time is an important task to manage the hospitalization of patients during a global pandemic. Recently, deep learning-based approaches have been proposed to help radiologists automatically quantify COVID-19 pathologies on longitudinal CT scans. However, the learning process of deep learning methods demands extensive training data to learn the complex characteristics of infected regions over longitudinal scans. It is challenging to collect a large-scale dataset, especially for longitudinal training. In this study, we want to address this problem by proposing a new self-supervised learning method to effectively train longitudinal networks for the quantification of COVID-19 infections. For this purpose, longitudinal self-supervision schemes are explored on clinical longitudinal COVID-19 CT scans. Experimental results show that the proposed method is effective, helping the model better exploit the semantics of longitudinal data and improve two COVID-19 quantification tasks.

Chaoning Zhang, Chenshuang Zhang, Chenghao Li et al.

OpenAI has recently released GPT-4 (a.k.a. ChatGPT plus), which is demonstrated to be one small step for generative AI (GAI), but one giant leap for artificial general intelligence (AGI). Since its official release in November 2022, ChatGPT has quickly attracted numerous users with extensive media coverage. Such unprecedented attention has also motivated numerous researchers to investigate ChatGPT from various aspects. According to Google scholar, there are more than 500 articles with ChatGPT in their titles or mentioning it in their abstracts. Considering this, a review is urgently needed, and our work fills this gap. Overall, this work is the first to survey ChatGPT with a comprehensive review of its underlying technology, applications, and challenges. Moreover, we present an outlook on how ChatGPT might evolve to realize general-purpose AIGC (a.k.a. AI-generated content), which will be a significant milestone for the development of AGI.

Felix Buchert, Nassir Navab, Seong Tae Kim

The availability of large labeled datasets is the key component for the success of deep learning. However, annotating labels on large datasets is generally time-consuming and expensive. Active learning is a research area that addresses the issues of expensive labeling by selecting the most important samples for labeling. Diversity-based sampling algorithms are known as integral components of representation-based approaches for active learning. In this paper, we introduce a new diversity-based initial dataset selection algorithm to select the most informative set of samples for initial labeling in the active learning setting. Self-supervised representation learning is used to consider the diversity of samples in the initial dataset selection algorithm. Also, we propose a novel active learning query strategy, which uses diversity-based sampling on consistency-based embeddings. By considering the consistency information with the diversity in the consistency-based embedding scheme, the proposed method could select more informative samples for labeling in the semi-supervised learning setting. Comparative experiments show that the proposed method achieves compelling results on CIFAR-10 and Caltech-101 datasets compared with previous active learning approaches by utilizing the diversity of unlabeled data.

Huy Q. Le, Loc X. Nguyen, Yu Qiao et al.

Federated Learning (FL) enables decentralized model training across multiple clients without exposing private data, making it ideal for privacy-sensitive applications. However, in real-world FL scenarios, clients often hold data from distinct domains, leading to severe domain shift and degraded global model performance. To address this, prototype learning has been emerged as a promising solution, which leverages class-wise feature representations. Yet, existing methods face two key limitations: (1) Existing prototype-based FL methods typically construct a $\textit{single global prototype}$ per class by aggregating local prototypes from all clients without preserving domain information. (2) Current feature-prototype alignment is $\textit{domain-agnostic}$, forcing clients to align with global prototypes regardless of domain origin. To address these challenges, we propose Federated Domain-Aware Prototypes (FedDAP) to construct domain-specific global prototypes by aggregating local client prototypes within the same domain using a similarity-weighted fusion mechanism. These global domain-specific prototypes are then used to guide local training by aligning local features with prototypes from the same domain, while encouraging separation from prototypes of different domains. This dual alignment enhances domain-specific learning at the local level and enables the global model to generalize across diverse domains. Finally, we conduct extensive experiments on three different datasets: DomainNet, Office-10, and PACS to demonstrate the effectiveness of our proposed framework to address the domain shift challenges. The code is available at https://github.com/quanghuy6997/FedDAP.

Samra Irshad, Douglas P. S. Gomes, Seong Tae Kim

Quantitative assessment of the abdominal region from clinically acquired CT scans requires the simultaneous segmentation of abdominal organs. Thanks to the availability of high-performance computational resources, deep learning-based methods have resulted in state-of-the-art performance for the segmentation of 3D abdominal CT scans. However, the complex characterization of organs with fuzzy boundaries prevents the deep learning methods from accurately segmenting these anatomical organs. Specifically, the voxels on the boundary of organs are more vulnerable to misprediction due to the highly-varying intensity of inter-organ boundaries. This paper investigates the possibility of improving the abdominal image segmentation performance of the existing 3D encoder-decoder networks by leveraging organ-boundary prediction as a complementary task. To address the problem of abdominal multi-organ segmentation, we train the 3D encoder-decoder network to simultaneously segment the abdominal organs and their corresponding boundaries in CT scans via multi-task learning. The network is trained end-to-end using a loss function that combines two task-specific losses, i.e., complete organ segmentation loss and boundary prediction loss. We explore two different network topologies based on the extent of weights shared between the two tasks within a unified multi-task framework. To evaluate the utilization of complementary boundary prediction task in improving the abdominal multi-organ segmentation, we use three state-of-the-art encoder-decoder networks: 3D UNet, 3D UNet++, and 3D Attention-UNet. The effectiveness of utilizing the organs' boundary information for abdominal multi-organ segmentation is evaluated on two publically available abdominal CT datasets. A maximum relative improvement of 3.5% and 3.6% is observed in Mean Dice Score for Pancreas-CT and BTCV datasets, respectively.

Yong Hyun Ahn, Gyeong-Moon Park, Seong Tae Kim

It is important to quantify the uncertainty of input samples, especially in mission-critical domains such as autonomous driving and healthcare, where failure predictions on out-of-distribution (OOD) data are likely to cause big problems. OOD detection problem fundamentally begins in that the model cannot express what it is not aware of. Post-hoc OOD detection approaches are widely explored because they do not require an additional re-training process which might degrade the model's performance and increase the training cost. In this study, from the perspective of neurons in the deep layer of the model representing high-level features, we introduce a new aspect for analyzing the difference in model outputs between in-distribution data and OOD data. We propose a novel method, Leveraging Important Neurons (LINe), for post-hoc Out of distribution detection. Shapley value-based pruning reduces the effects of noisy outputs by selecting only high-contribution neurons for predicting specific classes of input data and masking the rest. Activation clipping fixes all values above a certain threshold into the same value, allowing LINe to treat all the class-specific features equally and just consider the difference between the number of activated feature differences between in-distribution and OOD data. Comprehensive experiments verify the effectiveness of the proposed method by outperforming state-of-the-art post-hoc OOD detection methods on CIFAR-10, CIFAR-100, and ImageNet datasets.

Youngmin Oh, Hyung-Il Kim, Seong Tae Kim et al.

Monocular 3D object detection is an important challenging task in autonomous driving. Existing methods mainly focus on performing 3D detection in ideal weather conditions, characterized by scenarios with clear and optimal visibility. However, the challenge of autonomous driving requires the ability to handle changes in weather conditions, such as foggy weather, not just clear weather. We introduce MonoWAD, a novel weather-robust monocular 3D object detector with a weather-adaptive diffusion model. It contains two components: (1) the weather codebook to memorize the knowledge of the clear weather and generate a weather-reference feature for any input, and (2) the weather-adaptive diffusion model to enhance the feature representation of the input feature by incorporating a weather-reference feature. This serves an attention role in indicating how much improvement is needed for the input feature according to the weather conditions. To achieve this goal, we introduce a weather-adaptive enhancement loss to enhance the feature representation under both clear and foggy weather conditions. Extensive experiments under various weather conditions demonstrate that MonoWAD achieves weather-robust monocular 3D object detection. The code and dataset are released at https://github.com/VisualAIKHU/MonoWAD.

Ashkan Khakzar, Yawei Li, Yang Zhang et al.

One challenging property lurking in medical datasets is the imbalanced data distribution, where the frequency of the samples between the different classes is not balanced. Training a model on an imbalanced dataset can introduce unique challenges to the learning problem where a model is biased towards the highly frequent class. Many methods are proposed to tackle the distributional differences and the imbalanced problem. However, the impact of these approaches on the learned features is not well studied. In this paper, we look deeper into the internal units of neural networks to observe how handling data imbalance affects the learned features. We study several popular cost-sensitive approaches for handling data imbalance and analyze the feature maps of the convolutional neural networks from multiple perspectives: analyzing the alignment of salient features with pathologies and analyzing the pathology-related concepts encoded by the networks. Our study reveals differences and insights regarding the trained models that are not reflected by quantitative metrics such as AUROC and AP and show up only by looking at the models through a lens.

Soyoun Won, Sung-Ho Bae, Seong Tae Kim

Mixed sample data augmentation strategies are actively used when training deep neural networks (DNNs). Recent studies suggest that they are effective at various tasks. However, the impact of mixed sample data augmentation on model interpretability has not been widely studied. In this paper, we explore the relationship between model interpretability and mixed sample data augmentation, specifically in terms of feature attribution maps. To this end, we introduce a new metric that allows a comparison of model interpretability while minimizing the impact of occlusion robustness of the model. Experimental results show that several mixed sample data augmentation decreases the interpretability of the model and label mixing during data augmentation plays a significant role in this effect. This new finding suggests it is important to carefully adopt the mixed sample data augmentation method, particularly in applications where attribution map-based interpretability is important.

Hyeon Bae Kim, Yong Hyun Ahn, Seong Tae Kim

Recent advancements in deep neural networks have shown promise in aiding disease diagnosis and medical decision-making. However, ensuring transparent decision-making processes of AI models in compliance with regulations requires a comprehensive understanding of the model's internal workings. However, previous methods heavily rely on expensive pixel-wise annotated datasets for interpreting the model, presenting a significant drawback in medical domains. In this paper, we propose a novel medical neuron concept annotation method, named Mask-free Medical Model Interpretation (MAMMI), addresses these challenges. By using a vision-language model, our method relaxes the need for pixel-level masks for neuron concept annotation. MAMMI achieves superior performance compared to other interpretation methods, demonstrating its efficacy in providing rich representations for neurons in medical image analysis. Our experiments on a model trained on NIH chest X-rays validate the effectiveness of MAMMI, showcasing its potential for transparent clinical decision-making in the medical domain. The code is available at https://github.com/ailab-kyunghee/MAMMI.

Hao Ding, Yuqian Zhang, Tuxun Lu et al.

Surgical data science has seen rapid advancement due to the excellent performance of end-to-end deep neural networks (DNNs) for surgical video analysis. Despite their successes, end-to-end DNNs have been proven susceptible to even minor corruptions, substantially impairing the model's performance. This vulnerability has become a major concern for the translation of cutting-edge technology, especially for high-stakes decision-making in surgical data science. We introduce SegSTRONG-C, a benchmark and challenge in surgical data science dedicated, aiming to better understand model deterioration under unforeseen but plausible non-adversarial corruption and the capabilities of contemporary methods that seek to improve it. Through comprehensive baseline experiments and participating submissions from widespread community engagement, SegSTRONG-C reveals key themes for model failure and identifies promising directions for improving robustness. The performance of challenge winners, achieving an average 0.9394 DSC and 0.9301 NSD across the unreleased test sets with corruption types: bleeding, smoke, and low brightness, shows inspiring improvement of 0.1471 DSC and 0.2584 NSD in average comparing to strongest baseline methods with UNet architecture trained with AutoAugment. In conclusion, the SegSTRONG-C challenge has identified some practical approaches for enhancing model robustness, yet most approaches relied on conventional techniques that have known, and sometimes quite severe, limitations. Looking ahead, we advocate for expanding intellectual diversity and creativity in non-adversarial robustness beyond data augmentation or training scale, calling for new paradigms that enhance universal robustness to corruptions and may enable richer applications in surgical data science.

Jongseo Lee, Wooil Lee, Gyeong-Moon Park et al.

Effective explanations of video action recognition models should disentangle how movements unfold over time from the surrounding spatial context. However, existing methods based on saliency produce entangled explanations, making it unclear whether predictions rely on motion or spatial context. Language-based approaches offer structure but often fail to explain motions due to their tacit nature -- intuitively understood but difficult to verbalize. To address these challenges, we propose Disentangled Action aNd Context concept-based Explainable (DANCE) video action recognition, a framework that predicts actions through disentangled concept types: motion dynamics, objects, and scenes. We define motion dynamics concepts as human pose sequences. We employ a large language model to automatically extract object and scene concepts. Built on an ante-hoc concept bottleneck design, DANCE enforces prediction through these concepts. Experiments on four datasets -- KTH, Penn Action, HAA500, and UCF-101 -- demonstrate that DANCE significantly improves explanation clarity with competitive performance. We validate the superior interpretability of DANCE through a user study. Experimental results also show that DANCE is beneficial for model debugging, editing, and failure analysis.

Su Hyeon Lim, Minkuk Kim, Hyeon Bae Kim et al.

Visual Question Answering with Natural Language Explanation (VQA-NLE) task is challenging due to its high demand for reasoning-based inference. Recent VQA-NLE studies focus on enhancing model networks to amplify the model's reasoning capability but this approach is resource-consuming and unstable. In this work, we introduce a new VQA-NLE model, ReRe (Retrieval-augmented natural language Reasoning), using leverage retrieval information from the memory to aid in generating accurate answers and persuasive explanations without relying on complex networks and extra datasets. ReRe is an encoder-decoder architecture model using a pre-trained clip vision encoder and a pre-trained GPT-2 language model as a decoder. Cross-attention layers are added in the GPT-2 for processing retrieval features. ReRe outperforms previous methods in VQA accuracy and explanation score and shows improvement in NLE with more persuasive, reliability.

Jongmin Shin, Enki Cho, Ka Young Kim et al.

Surgical scene understanding is crucial for computer-assisted intervention systems, requiring visual comprehension of surgical scenes that involves diverse elements such as surgical tools, anatomical structures, and their interactions. To effectively represent the complex information in surgical scenes, graph-based approaches have been explored to structurally model surgical entities and their relationships. Previous surgical scene graph studies have demonstrated the feasibility of representing surgical scenes using graphs. However, certain aspects of surgical scenes-such as diverse combinations of tool-action-target and the identity of the hand operating the tool-remain underexplored in graph-based representations, despite their importance. To incorporate these aspects into graph representations, we propose Endoscapes-SG201 dataset, which includes annotations for tool-action-target combinations and hand identity. We also introduce SSG-Com, a graph-based method designed to learn and represent these critical elements. Through experiments on downstream tasks such as critical view of safety assessment and action triplet recognition, we demonstrated the importance of integrating these essential scene graph components, highlighting their significant contribution to surgical scene understanding. The code and dataset are available at https://github.com/ailab-kyunghee/SSG-Com

Ka Young Kim, Hyeon Bae Kim, Seong Tae Kim

Surgical phase recognition plays a crucial role in surgical workflow analysis, enabling various applications such as surgical monitoring, skill assessment, and workflow optimization. Despite significant advancements in deep learning-based surgical phase recognition, these models remain inherently opaque, making it difficult to understand how they make decisions. This lack of interpretability hinders trust and makes it challenging to debug the model. To address this challenge, we propose SurgX, a novel concept-based explanation framework that enhances the interpretability of surgical phase recognition models by associating neurons with relevant concepts. In this paper, we introduce the process of selecting representative example sequences for neurons, constructing a concept set tailored to the surgical video dataset, associating neurons with concepts and identifying neurons crucial for predictions. Through extensive experiments on two surgical phase recognition models, we validate our method and analyze the explanation for prediction. This highlights the potential of our method in explaining surgical phase recognition. The code is available at https://github.com/ailab-kyunghee/SurgX

Min-Yeong Park, Won-Jeong Lee, Seong Tae Kim et al.

Recently, forecasting future abnormal events has emerged as an important scenario to tackle real-world necessities. However, the solution of predicting specific future time points when anomalies will occur, known as Anomaly Prediction (AP), remains under-explored. Existing methods dealing with time series data fail in AP, focusing only on immediate anomalies or failing to provide precise predictions for future anomalies. To address the AP task, we propose a novel framework called Anomaly to Prompt (A2P), comprised of Anomaly-Aware Forecasting (AAF) and Synthetic Anomaly Prompting (SAP). To enable the forecasting model to forecast abnormal time points, we adopt a strategy to learn the relationships of anomalies. For the robust detection of anomalies, our proposed SAP introduces a learnable Anomaly Prompt Pool (APP) that simulates diverse anomaly patterns using signal adaptive prompt. Comprehensive experiments on multiple real-world datasets demonstrate the superiority of A2P over state-of-the-art methods, showcasing its ability to predict future anomalies. Our implementation code is available at https://github.com/KU-VGI/AP.

Minkuk Kim, Hyeon Bae Kim, Jinyoung Moon et al.

There has been significant attention to the research on dense video captioning, which aims to automatically localize and caption all events within untrimmed video. Several studies introduce methods by designing dense video captioning as a multitasking problem of event localization and event captioning to consider inter-task relations. However, addressing both tasks using only visual input is challenging due to the lack of semantic content. In this study, we address this by proposing a novel framework inspired by the cognitive information processing of humans. Our model utilizes external memory to incorporate prior knowledge. The memory retrieval method is proposed with cross-modal video-to-text matching. To effectively incorporate retrieved text features, the versatile encoder and the decoder with visual and textual cross-attention modules are designed. Comparative experiments have been conducted to show the effectiveness of the proposed method on ActivityNet Captions and YouCook2 datasets. Experimental results show promising performance of our model without extensive pretraining from a large video dataset.

Jongmin Shin, Ka Young Kim, Eunki Cho et al.

Purpose: Vision-language models (VLMs) have shown promising performance in surgical visual question answering (VQA). However, existing surgical VQA datasets often contain linguistic shortcuts, where question phrasing implicitly constrains the answer space. It remains unclear whether reported performance reflects visual understanding or reliance on such linguistic shortcuts. Methods: We introduce SurgCheck, a diagnostic benchmark for quantifying linguistic shortcut reliance in surgical VQA. SurgCheck employs a paired-question design in which each surgical frame is associated with an original question containing entity names and a less-biased counterpart that removes these names while preserving identical visual content and ground-truth answers. The resulting performance gap provides a diagnostic signal of shortcut reliance. To ensure that the less-biased question remains well-defined even without entity names, four grounding cues are incorporated: bounding box, arrow, spatial position, and periphrasis. We evaluate both general-purpose and surgical-specific VLMs under zero-shot and fine-tuned settings on SurgCheck. To evaluate open-ended zero-shot responses, we introduce an LLM-as-a-judge evaluation protocol. Results: Using SurgCheck, we observe consistent performance degradation on less-biased questions across five VLMs, despite identical visual inputs. Text-only ablation reveals minimal performance drops for action and target prediction, indicating that action and target prediction is largely driven by linguistic shortcuts rather than visual reasoning. Conclusion: SurgCheck provides a controlled diagnostic framework that exposes failure modes masked by linguistic bias in existing surgical VQA benchmarks. Our findings demonstrate that strong benchmark performance does not necessarily imply faithful visual understanding, underscoring the need for bias-aware evaluation in surgical VQA.

Jongseo Lee, Geo Ahn, Seong Tae Kim et al.

For automatic human figure drawing (HFD) assessment tasks, such as diagnosing autism spectrum disorder (ASD) using HFD images, the clarity and explainability of a model decision are crucial. Existing pixel-level attribution-based explainable AI (XAI) approaches demand considerable effort from users to interpret the semantic information of a region in an image, which can be often time-consuming and impractical. To overcome this challenge, we propose a part contribution evaluation based model explanation (PCEvE) framework. On top of the part detection, we measure the Shapley Value of each individual part to evaluate the contribution to a model decision. Unlike existing attribution-based XAI approaches, the PCEvE provides a straightforward explanation of a model decision, i.e., a part contribution histogram. Furthermore, the PCEvE expands the scope of explanations beyond the conventional sample-level to include class-level and task-level insights, offering a richer, more comprehensive understanding of model behavior. We rigorously validate the PCEvE via extensive experiments on multiple HFD assessment datasets. Also, we sanity-check the proposed method with a set of controlled experiments. Additionally, we demonstrate the versatility and applicability of our method to other domains by applying it to a photo-realistic dataset, the Stanford Cars.

Yong Hyun Ahn, Hyeon Bae Kim, Seong Tae Kim

Recent advancements in neural networks have showcased their remarkable capabilities across various domains. Despite these successes, the "black box" problem still remains. Addressing this, we propose a novel framework, WWW, that offers the 'what', 'where', and 'why' of the neural network decisions in human-understandable terms. Specifically, WWW utilizes adaptive selection for concept discovery, employing adaptive cosine similarity and thresholding techniques to effectively explain 'what'. To address the 'where' and 'why', we proposed a novel combination of neuron activation maps (NAMs) with Shapley values, generating localized concept maps and heatmaps for individual inputs. Furthermore, WWW introduces a method for predicting uncertainty, leveraging heatmap similarities to estimate 'how' reliable the prediction is. Experimental evaluations of WWW demonstrate superior performance in both quantitative and qualitative metrics, outperforming existing methods in interpretability. WWW provides a unified solution for explaining 'what', 'where', and 'why', introducing a method for localized explanations from global interpretations and offering a plug-and-play solution adaptable to various architectures.

Chu Myaet Thwal, Minh N. H. Nguyen, Ye Lin Tun et al.

Federated learning (FL) has emerged as a promising approach to collaboratively train machine learning models across multiple edge devices while preserving privacy. The success of FL hinges on the efficiency of participating models and their ability to handle the unique challenges of distributed learning. While several variants of Vision Transformer (ViT) have shown great potential as alternatives to modern convolutional neural networks (CNNs) for centralized training, the unprecedented size and higher computational demands hinder their deployment on resource-constrained edge devices, challenging their widespread application in FL. Since client devices in FL typically have limited computing resources and communication bandwidth, models intended for such devices must strike a balance between model size, computational efficiency, and the ability to adapt to the diverse and non-IID data distributions encountered in FL. To address these challenges, we propose OnDev-LCT: Lightweight Convolutional Transformers for On-Device vision tasks with limited training data and resources. Our models incorporate image-specific inductive biases through the LCT tokenizer by leveraging efficient depthwise separable convolutions in residual linear bottleneck blocks to extract local features, while the multi-head self-attention (MHSA) mechanism in the LCT encoder implicitly facilitates capturing global representations of images. Extensive experiments on benchmark image datasets indicate that our models outperform existing lightweight vision models while having fewer parameters and lower computational demands, making them suitable for FL scenarios with data heterogeneity and communication bottlenecks.

Minkuk Kim, Hyeon Bae Kim, Jinyoung Moon et al.

With the growing demand for solutions to real-world video challenges, interest in dense video captioning (DVC) has been on the rise. DVC involves the automatic captioning and localization of untrimmed videos. Several studies highlight the challenges of DVC and introduce improved methods utilizing prior knowledge, such as pre-training and external memory. In this research, we propose a model that leverages the prior knowledge of human-oriented hierarchical compact memory inspired by human memory hierarchy and cognition. To mimic human-like memory recall, we construct a hierarchical memory and a hierarchical memory reading module. We build an efficient hierarchical compact memory by employing clustering of memory events and summarization using large language models. Comparative experiments demonstrate that this hierarchical memory recall process improves the performance of DVC by achieving state-of-the-art performance on YouCook2 and ViTT datasets.

Joohyun Chang, Soyeon Hong, Hyogun Lee et al.

In this work, we tackle the egocentric visual query localization (VQL), where a model should localize the query object in a long-form egocentric video. Frequent and abrupt viewpoint changes in egocentric videos cause significant object appearance variations and partial occlusions, making it difficult for existing methods to achieve accurate localization. To tackle these challenges, we introduce Hierarchical, Egocentric and RObust Visual Query Localization (HERO-VQL), a novel method inspired by human cognitive process in object recognition. We propose i) Top-down Attention Guidance (TAG) and ii) Egocentric Augmentation based Consistency Training (EgoACT). Top-down Attention Guidance refines the attention mechanism by leveraging the class token for high-level context and principal component score maps for fine-grained localization. To enhance learning in diverse and challenging matching scenarios, EgoAug enhances query diversity by replacing the query with a randomly selected corresponding object from groundtruth annotations and simulates extreme viewpoint changes by reordering video frames. Additionally, CT loss enforces stable object localization across different augmentation scenarios. Extensive experiments on VQ2D dataset validate that HERO-VQL effectively handles egocentric challenges, significantly outperforming baselines.

Jongseo Lee, Wooil Lee, Gyeong-Moon Park et al.

Human action recognition (HAR) has achieved impressive results with deep learning models, but their decision-making process remains opaque due to their black-box nature. Ensuring interpretability is crucial, especially for real-world applications requiring transparency and accountability. Existing video XAI methods primarily rely on feature attribution or static textual concepts, both of which struggle to capture motion dynamics and temporal dependencies essential for action understanding. To address these challenges, we propose Pose Concept Bottleneck for Explainable Action Recognition (PCBEAR), a novel concept bottleneck framework that introduces human pose sequences as motion-aware, structured concepts for video action recognition. Unlike methods based on pixel-level features or static textual descriptions, PCBEAR leverages human skeleton poses, which focus solely on body movements, providing robust and interpretable explanations of motion dynamics. We define two types of pose-based concepts: static pose concepts for spatial configurations at individual frames, and dynamic pose concepts for motion patterns across multiple frames. To construct these concepts, PCBEAR applies clustering to video pose sequences, allowing for automatic discovery of meaningful concepts without manual annotation. We validate PCBEAR on KTH, Penn-Action, and HAA500, showing that it achieves high classification performance while offering interpretable, motion-driven explanations. Our method provides both strong predictive performance and human-understandable insights into the model's reasoning process, enabling test-time interventions for debugging and improving model behavior.

Abdullah, Ameer Hamza, Seong Tae Kim

Medical report generation is the task of automatically writing radiology reports for chest X-ray images. Manually composing these reports is a time-consuming process that is also prone to human errors. Generating medical reports can therefore help reduce the burden on radiologists. In other words, we can promote greater clinical automation in the medical domain. In this work, we propose a new framework leveraging vision-enabled Large Language Models (LLM) for the task of medical report generation. We introduce a lightweight solution that achieves better or comparative performance as compared to previous solutions on the task of medical report generation. We conduct extensive experiments exploring different model sizes and enhancement approaches, such as prefix tuning to improve the text generation abilities of the LLMs. We evaluate our approach on a prominent large-scale radiology report dataset - MIMIC-CXR. Our results demonstrate the capability of our resource-efficient framework to generate patient-specific reports with strong medical contextual understanding and high precision.

Kyu Ri Park, Jiyoung Park, Seong Tae Kim et al.

Change captioning aims to describe changes between a pair of images. However, existing works rely on visual features alone, which often fail to capture subtle but meaningful changes because they lack the ability to represent explicitly structured information such as object relationships and compositional semantics. To alleviate this, we present CORTEX (COmpositional Reasoning-aware TEXt-guided), a novel framework that integrates complementary textual cues to enhance change understanding. In addition to capturing cues from pixel-level differences, CORTEX utilizes scene-level textual knowledge provided by Vision Language Models (VLMs) to extract richer image text signals that reveal underlying compositional reasoning. CORTEX consists of three key modules: (i) an Image-level Change Detector that identifies low-level visual differences between paired images, (ii) a Reasoning-aware Text Extraction (RTE) module that use VLMs to generate compositional reasoning descriptions implicit in visual features, and (iii) an Image-Text Dual Alignment (ITDA) module that aligns visual and textual features for fine-grained relational reasoning. This enables CORTEX to reason over visual and textual features and capture changes that are otherwise ambiguous in visual features alone.

Tobias Rueckert, David Rauber, Raphaela Maerkl et al.

Reliable recognition and localization of surgical instruments in endoscopic video recordings are foundational for a wide range of applications in computer- and robot-assisted minimally invasive surgery (RAMIS), including surgical training, skill assessment, and autonomous assistance. However, robust performance under real-world conditions remains a significant challenge. Incorporating surgical context - such as the current procedural phase - has emerged as a promising strategy to improve robustness and interpretability. To address these challenges, we organized the Surgical Procedure Phase, Keypoint, and Instrument Recognition (PhaKIR) sub-challenge as part of the Endoscopic Vision (EndoVis) challenge at MICCAI 2024. We introduced a novel, multi-center dataset comprising thirteen full-length laparoscopic cholecystectomy videos collected from three distinct medical institutions, with unified annotations for three interrelated tasks: surgical phase recognition, instrument keypoint estimation, and instrument instance segmentation. Unlike existing datasets, ours enables joint investigation of instrument localization and procedural context within the same data while supporting the integration of temporal information across entire procedures. We report results and findings in accordance with the BIAS guidelines for biomedical image analysis challenges. The PhaKIR sub-challenge advances the field by providing a unique benchmark for developing temporally aware, context-driven methods in RAMIS and offers a high-quality resource to support future research in surgical scene understanding.

Abdullah Abdullah, Seong Tae Kim

Vision-Language Models (VLMs) have demonstrated remarkable success in natural language generation, excelling at instruction following and structured output generation. Knowledge graphs play a crucial role in radiology, serving as valuable sources of factual information and enhancing various downstream tasks. However, generating radiology-specific knowledge graphs presents significant challenges due to the specialized language of radiology reports and the limited availability of domain-specific data. Existing solutions are predominantly unimodal, meaning they generate knowledge graphs only from radiology reports while excluding radiographic images. Additionally, they struggle with long-form radiology data due to limited context length. To address these limitations, we propose a novel multimodal VLM-based framework for knowledge graph generation in radiology. Our approach outperforms previous methods and introduces the first multimodal solution for radiology knowledge graph generation.

Samra Irshad, Seungkyu Lee, Nassir Navab et al.

The presence of adversarial examples in the physical world poses significant challenges to the deployment of Deep Neural Networks in safety-critical applications such as autonomous driving. Most existing methods for crafting physical-world adversarial examples are ad-hoc, relying on temporary modifications like shadows, laser beams, or stickers that are tailored to specific scenarios. In this paper, we introduce a new class of physical-world adversarial examples, AdvWT, which draws inspiration from the naturally occurring phenomenon of `wear and tear', an inherent property of physical objects. Unlike manually crafted perturbations, `wear and tear' emerges organically over time due to environmental degradation, as seen in the gradual deterioration of outdoor signboards. To achieve this, AdvWT follows a two-step approach. First, a GAN-based, unsupervised image-to-image translation network is employed to model these naturally occurring damages, particularly in the context of outdoor signboards. The translation network encodes the characteristics of damaged signs into a latent `damage style code'. In the second step, we introduce adversarial perturbations into the style code, strategically optimizing its transformation process. This manipulation subtly alters the damage style representation, guiding the network to generate adversarial images where the appearance of damages remains perceptually realistic, while simultaneously ensuring their effectiveness in misleading neural networks. Through comprehensive experiments on two traffic sign datasets, we show that AdvWT effectively misleads DNNs in both digital and physical domains. AdvWT achieves an effective attack success rate, greater robustness, and a more natural appearance compared to existing physical-world adversarial examples. Additionally, integrating AdvWT into training enhances a model's generalizability to real-world damaged signs.

Yang Zhang, Ashkan Khakzar, Yawei Li et al.

One principal approach for illuminating a black-box neural network is feature attribution, i.e. identifying the importance of input features for the network's prediction. The predictive information of features is recently proposed as a proxy for the measure of their importance. So far, the predictive information is only identified for latent features by placing an information bottleneck within the network. We propose a method to identify features with predictive information in the input domain. The method results in fine-grained identification of input features' information and is agnostic to network architecture. The core idea of our method is leveraging a bottleneck on the input that only lets input features associated with predictive latent features pass through. We compare our method with several feature attribution methods using mainstream feature attribution evaluation experiments. The code is publicly available.

Michelle Xiao-Lin Foo, Seong Tae Kim, Magdalini Paschali et al.

Consistent segmentation of COVID-19 patient's CT scans across multiple time points is essential to assess disease progression and response to therapy accurately. Existing automatic and interactive segmentation models for medical images only use data from a single time point (static). However, valuable segmentation information from previous time points is often not used to aid the segmentation of a patient's follow-up scans. Also, fully automatic segmentation techniques frequently produce results that would need further editing for clinical use. In this work, we propose a new single network model for interactive segmentation that fully utilizes all available past information to refine the segmentation of follow-up scans. In the first segmentation round, our model takes 3D volumes of medical images from two-time points (target and reference) as concatenated slices with the additional reference time point segmentation as a guide to segment the target scan. In subsequent segmentation refinement rounds, user feedback in the form of scribbles that correct the segmentation and the target's previous segmentation results are additionally fed into the model. This ensures that the segmentation information from previous refinement rounds is retained. Experimental results on our in-house multiclass longitudinal COVID-19 dataset show that the proposed model outperforms its static version and can assist in localizing COVID-19 infections in patient's follow-up scans.

Ashkan Khakzar, Sabrina Musatian, Jonas Buchberger et al.

Convolutional neural networks are showing promise in the automatic diagnosis of thoracic pathologies on chest x-rays. Their black-box nature has sparked many recent works to explain the prediction via input feature attribution methods (aka saliency methods). However, input feature attribution methods merely identify the importance of input regions for the prediction and lack semantic interpretation of model behavior. In this work, we first identify the semantics associated with internal units (feature maps) of the network. We proceed to investigate the following questions; Does a regression model that is only trained with COVID-19 severity scores implicitly learn visual patterns associated with thoracic pathologies? Does a network that is trained on weakly labeled data (e.g. healthy, unhealthy) implicitly learn pathologies? Moreover, we investigate the effect of pretraining and data imbalance on the interpretability of learned features. In addition to the analysis, we propose semantic attribution to semantically explain each prediction. We present our findings using publicly available chest pathologies (CheXpert, NIH ChestX-ray8) and COVID-19 datasets (BrixIA, and COVID-19 chest X-ray segmentation dataset). The Code is publicly available.

Ashkan Khakzar, Yang Zhang, Wejdene Mansour et al.

Neural networks have demonstrated remarkable performance in classification and regression tasks on chest X-rays. In order to establish trust in the clinical routine, the networks' prediction mechanism needs to be interpretable. One principal approach to interpretation is feature attribution. Feature attribution methods identify the importance of input features for the output prediction. Building on Information Bottleneck Attribution (IBA) method, for each prediction we identify the chest X-ray regions that have high mutual information with the network's output. Original IBA identifies input regions that have sufficient predictive information. We propose Inverse IBA to identify all informative regions. Thus all predictive cues for pathologies are highlighted on the X-rays, a desirable property for chest X-ray diagnosis. Moreover, we propose Regression IBA for explaining regression models. Using Regression IBA we observe that a model trained on cumulative severity score labels implicitly learns the severity of different X-ray regions. Finally, we propose Multi-layer IBA to generate higher resolution and more detailed attribution/saliency maps. We evaluate our methods using both human-centric (ground-truth-based) interpretability metrics, and human-independent feature importance metrics on NIH Chest X-ray8 and BrixIA datasets. The Code is publicly available.

Ashkan Khakzar, Soroosh Baselizadeh, Saurabh Khanduja et al.

Is critical input information encoded in specific sparse pathways within the neural network? In this work, we discuss the problem of identifying these critical pathways and subsequently leverage them for interpreting the network's response to an input. The pruning objective -- selecting the smallest group of neurons for which the response remains equivalent to the original network -- has been previously proposed for identifying critical pathways. We demonstrate that sparse pathways derived from pruning do not necessarily encode critical input information. To ensure sparse pathways include critical fragments of the encoded input information, we propose pathway selection via neurons' contribution to the response. We proceed to explain how critical pathways can reveal critical input features. We prove that pathways selected via neuron contribution are locally linear (in an L2-ball), a property that we use for proposing a feature attribution method: "pathway gradient". We validate our interpretation method using mainstream evaluation experiments. The validation of pathway gradient interpretation method further confirms that selected pathways using neuron contributions correspond to critical input features. The code is publicly available.

Aadhithya Sankar, Matthias Keicher, Rami Eisawy et al.

Disentangled representations can be useful in many downstream tasks, help to make deep learning models more interpretable, and allow for control over features of synthetically generated images that can be useful in training other models that require a large number of labelled or unlabelled data. Recently, flow-based generative models have been proposed to generate realistic images by directly modeling the data distribution with invertible functions. In this work, we propose a new flow-based generative model framework, named GLOWin, that is end-to-end invertible and able to learn disentangled representations. Feature disentanglement is achieved by factorizing the latent space into components such that each component learns the representation for one generative factor. Comprehensive experiments have been conducted to evaluate the proposed method on a public brain tumor MR dataset. Quantitative and qualitative results suggest that the proposed method is effective in disentangling the features from complex medical images.

Seong Tae Kim, Leili Goli, Magdalini Paschali et al.

Chest computed tomography (CT) has played an essential diagnostic role in assessing patients with COVID-19 by showing disease-specific image features such as ground-glass opacity and consolidation. Image segmentation methods have proven to help quantify the disease burden and even help predict the outcome. The availability of longitudinal CT series may also result in an efficient and effective method to reliably assess the progression of COVID-19, monitor the healing process and the response to different therapeutic strategies. In this paper, we propose a new framework to identify infection at a voxel level (identification of healthy lung, consolidation, and ground-glass opacity) and visualize the progression of COVID-19 using sequential low-dose non-contrast CT scans. In particular, we devise a longitudinal segmentation network that utilizes the reference scan information to improve the performance of disease identification. Experimental results on a clinical longitudinal dataset collected in our institution show the effectiveness of the proposed method compared to the static deep neural networks for disease quantification.

Tobias Czempiel, Magdalini Paschali, Daniel Ostler et al.

In this paper we introduce OperA, a transformer-based model that accurately predicts surgical phases from long video sequences. A novel attention regularization loss encourages the model to focus on high-quality frames during training. Moreover, the attention weights are utilized to identify characteristic high attention frames for each surgical phase, which could further be used for surgery summarization. OperA is thoroughly evaluated on two datasets of laparoscopic cholecystectomy videos, outperforming various state-of-the-art temporal refinement approaches.

Abinav Ravi Venkatakrishnan, Seong Tae Kim, Rami Eisawy et al.

Medical imaging data suffers from the limited availability of annotation because annotating 3D medical data is a time-consuming and expensive task. Moreover, even if the annotation is available, supervised learning-based approaches suffer highly imbalanced data. Most of the scans during the screening are from normal subjects, but there are also large variations in abnormal cases. To address these issues, recently, unsupervised deep anomaly detection methods that train the model on large-sized normal scans and detect abnormal scans by calculating reconstruction error have been reported. In this paper, we propose a novel self-supervised learning technique for anomaly detection. Our architecture largely consists of two parts: 1) Reconstruction and 2) predicting geometric transformations. By training the network to predict geometric transformations, the model could learn better image features and distribution of normal scans. In the test time, the geometric transformation predictor can assign the anomaly score by calculating the error between geometric transformation and prediction. Moreover, we further use self-supervised learning with context restoration for pretraining our model. By comparative experiments on clinical brain CT scans, the effectiveness of the proposed method has been verified.

Hakmin Lee, Hong Joo Lee, Seong Tae Kim et al.

Deep neural networks have achieved substantial achievements in several computer vision areas, but have vulnerabilities that are often fooled by adversarial examples that are not recognized by humans. This is an important issue for security or medical applications. In this paper, we propose an ensemble model training framework with random layer sampling to improve the robustness of deep neural networks. In the proposed training framework, we generate various sampled model through the random layer sampling and update the weight of the sampled model. After the ensemble models are trained, it can hide the gradient efficiently and avoid the gradient-based attack by the random layer sampling method. To evaluate our proposed method, comprehensive and comparative experiments have been conducted on three datasets. Experimental results show that the proposed method improves the adversarial robustness.

Hong Joo Lee, Seong Tae Kim, Hakmin Lee et al.

Recent studies have shown that ensemble approaches could not only improve accuracy and but also estimate model uncertainty in deep learning. However, it requires a large number of parameters according to the increase of ensemble models for better prediction and uncertainty estimation. To address this issue, a generic and efficient segmentation framework to construct ensemble segmentation models is devised in this paper. In the proposed method, ensemble models can be efficiently generated by using the stochastic layer selection method. The ensemble models are trained to estimate uncertainty through Bayesian approximation. Moreover, to overcome its limitation from uncertain instances, we devise a new pixel-wise uncertainty loss, which improves the predictive performance. To evaluate our method, comprehensive and comparative experiments have been conducted on two datasets. Experimental results show that the proposed method could provide useful uncertainty information by Bayesian approximation with the efficient ensemble model generation and improve the predictive performance.

Stefan Denner, Ashkan Khakzar, Moiz Sajid et al.

Segmentation of Multiple Sclerosis (MS) lesions in longitudinal brain MR scans is performed for monitoring the progression of MS lesions. We hypothesize that the spatio-temporal cues in longitudinal data can aid the segmentation algorithm. Therefore, we propose a multi-task learning approach by defining an auxiliary self-supervised task of deformable registration between two time-points to guide the neural network toward learning from spatio-temporal changes. We show the efficacy of our method on a clinical dataset comprised of 70 patients with one follow-up study for each patient. Our results show that spatio-temporal information in longitudinal data is a beneficial cue for improving segmentation. We improve the result of current state-of-the-art by 2.6% in terms of overall score (p<0.05). Code is publicly available.

Seong Tae Kim, Farrukh Mushtaq, Nassir Navab

Recent advances in deep learning have resulted in great successes in various applications. Although semi-supervised or unsupervised learning methods have been widely investigated, the performance of deep neural networks highly depends on the annotated data. The problem is that the budget for annotation is usually limited due to the annotation time and expensive annotation cost in medical data. Active learning is one of the solutions to this problem where an active learner is designed to indicate which samples need to be annotated to effectively train a target model. In this paper, we propose a novel active learning method, confident coreset, which considers both uncertainty and distribution for effectively selecting informative samples. By comparative experiments on two medical image analysis tasks, we show that our method outperforms other active learning methods.

Tobias Czempiel, Magdalini Paschali, Matthias Keicher et al.

Automatic surgical phase recognition is a challenging and crucial task with the potential to improve patient safety and become an integral part of intra-operative decision-support systems. In this paper, we propose, for the first time in workflow analysis, a Multi-Stage Temporal Convolutional Network (MS-TCN) that performs hierarchical prediction refinement for surgical phase recognition. Causal, dilated convolutions allow for a large receptive field and online inference with smooth predictions even during ambiguous transitions. Our method is thoroughly evaluated on two datasets of laparoscopic cholecystectomy videos with and without the use of additional surgical tool information. Outperforming various state-of-the-art LSTM approaches, we verify the suitability of the proposed causal MS-TCN for surgical phase recognition.

Maria Tirindelli, Maria Victorova, Javier Esteban et al.

Spine injections are commonly performed in several clinical procedures. The localization of the target vertebral level (i.e. the position of a vertebra in a spine) is typically done by back palpation or under X-ray guidance, yielding either higher chances of procedure failure or exposure to ionizing radiation. Preliminary studies have been conducted in the literature, suggesting that ultrasound imaging may be a precise and safe alternative to X-ray for spine level detection. However, ultrasound data are noisy and complicated to interpret. In this study, a robotic-ultrasound approach for automatic vertebral level detection is introduced. The method relies on the fusion of ultrasound and force data, thus providing both "tactile" and visual feedback during the procedure, which results in higher performances in presence of data corruption. A robotic arm automatically scans the volunteer's back along the spine by using force-ultrasound data to locate vertebral levels. The occurrences of vertebral levels are visible on the force trace as peaks, which are enhanced by properly controlling the force applied by the robot on the patient back. Ultrasound data are processed with a Deep Learning method to extract a 1D signal modelling the probabilities of having a vertebra at each location along the spine. Processed force and ultrasound data are fused using a 1D Convolutional Network to compute the location of the vertebral levels. The method is compared to pure image and pure force-based methods for vertebral level counting, showing improved performance. In particular, the fusion method is able to correctly classify 100% of the vertebral levels in the test set, while pure image and pure force-based method could only classify 80% and 90% vertebrae, respectively. The potential of the proposed method is evaluated in an exemplary simulated clinical application.

Ashkan Khakzar, Soroosh Baselizadeh, Saurabh Khanduja et al.

Attributing the output of a neural network to the contribution of given input elements is a way of shedding light on the black-box nature of neural networks. Due to the complexity of current network architectures, current gradient-based attribution methods provide very noisy or coarse results. We propose to prune a neural network for a given single input to keep only neurons that highly contribute to the prediction. We show that by input-specific pruning, network gradients change from reflecting local (noisy) importance information to global importance. Our proposed method is efficient and generates fine-grained attribution maps. We further provide a theoretical justification of the pruning approach relating it to perturbations and validate it through a novel experimental setup. Our method is evaluated by multiple benchmarks: sanity checks, pixel perturbation, and Remove-and-Retrain (ROAR). These benchmarks evaluate the method from different perspectives and our method performs better than other methods across all evaluations.

Hyebin Lee, Seong Tae Kim, Yong Man Ro

The ambiguity of the decision-making process has been pointed out as the main obstacle to applying the deep learning-based method in a practical way in spite of its outstanding performance. Interpretability could guarantee the confidence of deep learning system, therefore it is particularly important in the medical field. In this study, a novel deep network is proposed to explain the diagnostic decision with visual pointing map and diagnostic sentence justifying result simultaneously. For the purpose of increasing the accuracy of sentence generation, a visual word constraint model is devised in training justification generator. To verify the proposed method, comparative experiments were conducted on the problem of the diagnosis of breast masses. Experimental results demonstrated that the proposed deep network could explain diagnosis more accurately with various textual justifications.

Jae-Hyeok Lee, Seong Tae Kim, Hakmin Lee et al.

This paper deals with a method for generating realistic labeled masses. Recently, there have been many attempts to apply deep learning to various bio-image computing fields including computer-aided detection and diagnosis. In order to learn deep network model to be well-behaved in bio-image computing fields, a lot of labeled data is required. However, in many bioimaging fields, the large-size of labeled dataset is scarcely available. Although a few researches have been dedicated to solving this problem through generative model, there are some problems as follows: 1) The generated bio-image does not seem realistic; 2) the variation of generated bio-image is limited; and 3) additional label annotation task is needed. In this study, we propose a realistic labeled bio-image generation method through visual feature processing in latent space. Experimental results have shown that mass images generated by the proposed method were realistic and had wide expression range of targeted mass characteristics.

Seong Tae Kim, Hakmin Lee, Hak Gu Kim et al.

In this study, a novel computer aided diagnosis (CADx) framework is devised to investigate interpretability for classifying breast masses. Recently, a deep learning technology has been successfully applied to medical image analysis including CADx. Existing deep learning based CADx approaches, however, have a limitation in explaining the diagnostic decision. In real clinical practice, clinical decisions could be made with reasonable explanation. So current deep learning approaches in CADx are limited in real world deployment. In this paper, we investigate interpretability in CADx with the proposed interpretable CADx (ICADx) framework. The proposed framework is devised with a generative adversarial network, which consists of interpretable diagnosis network and synthetic lesion generative network to learn the relationship between malignancy and a standardized description (BI-RADS). The lesion generative network and the interpretable diagnosis network compete in an adversarial learning so that the two networks are improved. The effectiveness of the proposed method was validated on public mammogram database. Experimental results showed that the proposed ICADx framework could provide the interpretability of mass as well as mass classification. It was mainly attributed to the fact that the proposed method was effectively trained to find the relationship between malignancy and interpretations via the adversarial learning. These results imply that the proposed ICADx framework could be a promising approach to develop the CADx system.

Seong Tae Kim, Yong Man Ro

Human face analysis is an important task in computer vision. According to cognitive-psychological studies, facial dynamics could provide crucial cues for face analysis. The motion of a facial local region in facial expression is related to the motion of other facial local regions. In this paper, a novel deep learning approach, named facial dynamics interpreter network, has been proposed to interpret the important relations between local dynamics for estimating facial traits from expression sequence. The facial dynamics interpreter network is designed to be able to encode a relational importance, which is used for interpreting the relation between facial local dynamics and estimating facial traits. By comparative experiments, the effectiveness of the proposed method has been verified. The important relations between facial local dynamics are investigated by the proposed facial dynamics interpreter network in gender classification and age estimation. Moreover, experimental results show that the proposed method outperforms the state-of-the-art methods in gender classification and age estimation.

Geonmo Gu, Seong Tae Kim, Kihyun Kim et al.

In face-related applications with a public available dataset, synthesizing non-linear facial variations (e.g., facial expression, head-pose, illumination, etc.) through a generative model is helpful in addressing the lack of training data. In reality, however, there is insufficient data to even train the generative model for face synthesis. In this paper, we propose Differential Generative Adversarial Networks (D-GAN) that can perform photo-realistic face synthesis even when training data is small. Two discriminators are devised to ensure the generator to approximate a face manifold, which can express face changes as it wants. Experimental results demonstrate that the proposed method is robust to the amount of training data and synthesized images are useful to improve the performance of a face expression classifier.