arXiv subjects

Robert Graf

Publications and source records attributed to Robert Graf.

23 records · Page 2Linked to original sources

arXiv2023

Denoising diffusion-based MRI to CT image translation enables automated spinal segmentation

Background: Automated segmentation of spinal MR images plays a vital role both scientifically and clinically. However, accurately delineating posterior spine structures presents challenges. Methods: This retrospective study, approved by the ethical committee, involved translating T1w and T2w MR image series into CT images in a total of n=263 pairs of CT/MR series. Landmark-based registration was performed to align image pairs. We compared 2D paired (Pix2Pix, denoising diffusion implicit models (DDIM) image mode, DDIM noise mode) and unpaired (contrastive unpaired translation, SynDiff) image-to-image translation using "peak signal to noise ratio" (PSNR) as quality measure. A publicly available segmentation network segmented the synthesized CT datasets, and Dice scores were evaluated on in-house test sets and the "MRSpineSeg Challenge" volumes. The 2D findings were extended to 3D Pix2Pix and DDIM. Results: 2D paired methods and SynDiff exhibited similar translation performance and Dice scores on paired data. DDIM image mode achieved the highest image quality. SynDiff, Pix2Pix, and DDIM image mode demonstrated similar Dice scores (0.77). For craniocaudal axis rotations, at least two landmarks per vertebra were required for registration. The 3D translation outperformed the 2D approach, resulting in improved Dice scores (0.80) and anatomically accurate segmentations in a higher resolution than the original MR image. Conclusion: Two landmarks per vertebra registration enabled paired image-to-image translation from MR to CT and outperformed all unpaired approaches. The 3D techniques provided anatomically correct segmentations, avoiding underprediction of small structures like the spinous process.

eess.IV↗

arXiv2023

3D Arterial Segmentation via Single 2D Projections and Depth Supervision in Contrast-Enhanced CT Images

Automated segmentation of the blood vessels in 3D volumes is an essential step for the quantitative diagnosis and treatment of many vascular diseases. 3D vessel segmentation is being actively investigated in existing works, mostly in deep learning approaches. However, training 3D deep networks requires large amounts of manual 3D annotations from experts, which are laborious to obtain. This is especially the case for 3D vessel segmentation, as vessels are sparse yet spread out over many slices and disconnected when visualized in 2D slices. In this work, we propose a novel method to segment the 3D peripancreatic arteries solely from one annotated 2D projection per training image with depth supervision. We perform extensive experiments on the segmentation of peripancreatic arteries on 3D contrast-enhanced CT images and demonstrate how well we capture the rich depth information from 2D projections. We demonstrate that by annotating a single, randomly chosen projection for each training sample, we obtain comparable performance to annotating multiple 2D projections, thereby reducing the annotation effort. Furthermore, by mapping the 2D labels to the 3D space using depth information and incorporating this into training, we almost close the performance gap between 3D supervision and 2D supervision. Our code is available at: https://github.com/alinafdima/3Dseg-mip-depth.

cs.CV↗

arXiv2023

Attention-based Saliency Maps Improve Interpretability of Pneumothorax Classification

Purpose: To investigate chest radiograph (CXR) classification performance of vision transformers (ViT) and interpretability of attention-based saliency using the example of pneumothorax classification. Materials and Methods: In this retrospective study, ViTs were fine-tuned for lung disease classification using four public data sets: CheXpert, Chest X-Ray 14, MIMIC CXR, and VinBigData. Saliency maps were generated using transformer multimodal explainability and gradient-weighted class activation mapping (GradCAM). Classification performance was evaluated on the Chest X-Ray 14, VinBigData, and SIIM-ACR data sets using the area under the receiver operating characteristic curve analysis (AUC) and compared with convolutional neural networks (CNNs). The explainability methods were evaluated with positive/negative perturbation, sensitivity-n, effective heat ratio, intra-architecture repeatability and interarchitecture reproducibility. In the user study, three radiologists classified 160 CXRs with/without saliency maps for pneumothorax and rated their usefulness. Results: ViTs had comparable CXR classification AUCs compared with state-of-the-art CNNs 0.95 (95% CI: 0.943, 0.950) versus 0.83 (95%, CI 0.826, 0.842) on Chest X-Ray 14, 0.84 (95% CI: 0.769, 0.912) versus 0.83 (95% CI: 0.760, 0.895) on VinBigData, and 0.85 (95% CI: 0.847, 0.861) versus 0.87 (95% CI: 0.868, 0.882) on SIIM ACR. Both saliency map methods unveiled a strong bias toward pneumothorax tubes in the models. Radiologists found 47% of the attention-based saliency maps useful and 39% of GradCAM. The attention-based methods outperformed GradCAM on all metrics. Conclusion: ViTs performed similarly to CNNs in CXR classification, and their attention-based saliency maps were more useful to radiologists and outperformed GradCAM.

eess.IV↗

arXiv2014

Critical heights of destruction for a forest-fire model on the half-plane

Consider the following forest-fire model on the upper half-plane of the triangular lattice: Each site can be "vacant" or "occupied by a tree". At time 0 all sites are vacant. Then the process is governed by the following random dynamics: Trees grow at rate 1, independently for all sites. If an occupied cluster reaches the boundary of the upper half-plane, the cluster is instantaneously destroyed, i.e. all of its sites turn vacant. At the critical time $t_c := \log 2$ the process is stopped. Now choose an arbitrary infinite cone in the half-plane whose apex lies on the boundary of the half-plane and whose boundary lines are non-horizontal. We prove that in the final configuration a.s. only finitely many sites in the cone have been affected by destruction.

math.PR↗

arXiv2014

Self-destructive percolation as a limit of forest-fire models on regular rooted trees

Let $T$ be a regular rooted tree. For every natural number $n$, let $B_n$ be the finite subtree of vertices with graph distance at most $n$ from the root. Consider the following forest-fire model on $B_n$: Each vertex can be "vacant" or "occupied". At time $0$ all vertices are vacant. Then the process is governed by two opposing mechanisms: Vertices become occupied at rate $1$, independently for all vertices. Independently thereof and independently for all vertices, "lightning" hits vertices at rate $λ(n) > 0$. When a vertex is hit by lightning, its occupied cluster instantaneously becomes vacant. Now suppose that $λ(n)$ decays exponentially in $n$ but much more slowly than $1/|B_n|$. We show that then there exist a supercritical time $τ$ and $ε> 0$ such that the forest-fire model on $B_n$ between time $0$ and time $τ+ ε$ tends to the following process on $T$ as $n$ goes to infinity: At time $0$ all vertices are vacant. Between time $0$ and time $τ$ vertices become occupied at rate $1$, independently for all vertices. At time $τ$ all infinite occupied clusters become vacant. Between time $τ$ and time $τ+ ε$ vertices again become occupied at rate $1$, independently for all vertices. At time $τ+ ε$ all occupied clusters are finite. This process is a dynamic version of self-destructive percolation.

math.PR↗