Shuoqi Chen’s Post

I am thrilled to share that two papers have been accepted to #ICLR2024! - FairSeg is the first fairness segmentation dataset/benchmark and we use a modified segment anything (SAM) model to improve the segmentation fairness. You can find the paper on OpenReview here: https://lnkd.in/dkb3cGK7. The code and dataset have been uploaded here: https://lnkd.in/dTiZqXTw. - AnomalyCLIP introduces an object-agnostic prompt learning method for anomaly detection and segmentation using CLIP. Shows exceptional zero-shot performance on 17 real-world datasets in defect inspection and medical imaging. You can find the paper on OpenReview here: https://lnkd.in/dQ35keec. The code will be available here: https://lnkd.in/dJZ6JmE4.

In a recent paper, LightM-UNet is introduced as an AI model for medical image segmentation. This model is noteworthy for its integration of two architectures: Mamba and UNet. Mamba is a sequence modeling architecture that excels in handling long sequences efficiently. It’s designed to address the computational inefficiencies of Transformer models, particularly in processing long sequences. Mamba achieves this through a linear-time algorithm that selectively propagates or forgets information along the sequence length, depending on the input. UNet, on the other hand, is a deep learning architecture specifically developed for biomedical image segmentation. It features a unique U-shaped design with a contracting path to capture context and an expansive path to enable precise localization. The architecture uses skip connections to combine low-level feature maps with higher-level ones, which helps in precise pixel-level segmentation. The LightM-UNet model leverages the strengths of both Mamba and UNet, aiming to provide a powerful yet efficient tool for medical image analysis. #lightmunet #deeplearning #cnn paper: https://lnkd.in/ghPw5TnX

In recent years, hospitals have faced the challenge of effectively managing both elective and emergency patients while meeting individual patient needs. To address these challenges, #ML algorithms have been increasingly utilized for predictive purposes in medical facilities. In our posts, we have already highlighted studies focused on triage predictions, and forecasting demand within hospitals, including the demand for surgical units. Time-series analysis has been a key approach, with models like autoregressive integrated moving average (#ARIMA) and seasonal ARIMA (#SARIMA) widely used. Yet, these models have certain limitations, leading researchers to explore other algorithms such as support vector regression (#SVR) and artificial neural networks (#ANNs) for demand prediction. Hybrid models combining statistical and non-linear approaches have shown promising results. For instance, combining SARIMA with #MLmodels like SVR or MLP has improved prediction accuracy for surgical unit demand. Despite the progress made, I'd like to recognize the ongoing debate surrounding the effectiveness and generalizability of predictive models in healthcare settings. Questions arise regarding the scalability of these models across different hospital environments and patient populations. Additionally, the interpretability of ML algorithms remains a concern for healthcare practitioners, as the black-box nature of some models may hinder their adoption and trust. https://lnkd.in/eaGUeNji  

The New England Journal of Medicine has launched a new journal, NEJM AI, geared toward ensuring the responsible development of AI in healthcare. Of particular note is the need for new interdisciplinary teams to develop, evaluate and bring to market AI approaches in healthcare. As the NEJM AI editorial puts it: "In our eyes, the most impactful articles will blossom from the fertile ground of multidisciplinary teams, reflecting the vibrance at the intersection of computer science, clinician–patient dynamics, and biomedical research." We're seeing many collaborations of this sort across Carnegie Mellon University, University of Pittsburgh, UPMC, and Allegheny Health Network, with more to come! https://ai.nejm.org/

AI is quickly evolving and changing the landscape of healthcare, both in the academic setting as well as patient care and medical research. Join Ovid and NEJM to look into the expanding influence of AI in medical education and research. https://bit.ly/3TSEMWT #NEJMAI #OVID #HealthcareAI

Glad to announce that the paper titled 'Coupling Neural Networks Between Clusters for Better Personalized Care' I got to contribute to has received the 'Best Paper' award in the 'Information Technology in Healthcare' track in the 57th Hawaii Internation Conference on System Sciences. The abstract is available in the link below for interested readers. The full paper will appear online in due course. Cheers! https://lnkd.in/gCgDp7fd