Matthew Baugh
10 Papers
Matthew Baugh is an academic researcher. The author has contributed to research in topics: Computer science & Engineering. The author has an hindex of 1, co-authored 7 publications.
Chat about Author
Papers
DISYRE: Diffusion-Inspired SYnthetic REstoration for Unsupervised Anomaly Detection
Sergio Naval Marimont,Matthew Baugh,Vasilis Siomos,Christos Tzelepis,Bernhard Kainz,Giacomo Tarroni +5 more
TL;DR: This work addresses the problem of how to learn a score function relevant for UAD and proposes DISYRE: Diffusion-Inspired SYnthetic REstoration, which retains the diffusion-like pipeline but replaces the Gaussian noise corruption with a gradual, synthetic anomaly corruption so the learned score function generalizes to medical, naturally occurring anomalies.
3
Adnexal Mass Segmentation with Ultrasound Data Synthesis
Clara Lebbos,J. Barcroft,Jeremy Tan,J. Müller,Matthew Baugh,Athanasios Vlontzos,Srdjan Saso,B. Kainz +7 more
- 25 Sep 2022
TL;DR: In this article , a novel pathology-specific data synthesizer was used to segment adnexal masses using Poisson image editing to integrate less common masses into other samples, achieving an improvement of up to 8% when compared with nnU-Net baseline approaches.
MoCoSR: Respiratory Motion Correction and Super-Resolution for 3D Abdominal MRI
Weitong Zhang,Doga Basaran,Qingjie Meng,Matthew Baugh,Jonathan K Stelter,Phillip Lung,Uday Patel,Wenjia Bai,Dimitrios C. Karampinos,Bernhard Kainz +9 more
TL;DR: This paper proposes MoCoSR, a deep adversarial super-resolution approach for 3D abdominal MRI, addressing respiratory motion and organ motion through a staged reconstruction model, leveraging a low-resolution latent space for motion correction and super-resolution reconstruction.
1
Pay Attention: Accuracy Versus Interpretability Trade-off in Fine-tuned Diffusion Models
TL;DR: The authors showed that fine-tuning text-to-image models with learnable text encoder leads to a lack of interpretability of diffusion models, and demonstrated the interpretability by keeping the language encoder frozen, enables diffusion models to achieve state-of-the-art phrase grounding performance on certain diseases for a challenging multi-label segmentation task, without any additional training.
nnOOD: A Framework for Benchmarking Self-supervised Anomaly Localisation Methods
TL;DR: In this article , a comparison of self-supervised anomaly localisation methods is made by isolating the synthetic task from the rest of the training process and making the workflow for evaluating over a given dataset quick and easy.
1