Robust Identification of Contrasted Frames in Fluoroscopic Images

Hoffmann, Matthias; Müller, Simone; Kurzidim, Klaus; Strobel, Norbert; Hornegger, Joachim

doi:10.1007/978-3-662-46224-9_6

Cited by 4 publications

(8 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The table also lists the number of sequences with a small prediction error (3 frames, being about 0.2 seconds) and a large error (>10 frames). The method of Condurache et al has mis-detection on 7 sequences (the first entry in the last two columns in Table 1), which was also reported in [9]. While the two proposed methods both have 55 sequences with a small error ( 3 frames) out of 80, the CNN-based approach has the smallest numbers of sequences with a large error (> 10 frames) among the three methods.…”

Section: Methodsmentioning

confidence: 65%

“…[7] uses a heuristic approach to detect the first contrast-enhanced frame from X-ray sequences of left atrium (LA) used for electrophysiology (EP) ablation procedures. [9] developed a learning-based framework on X-ray images of LA for EP procedures. The method used a SVM classifer with the heuristic features introduced in [3] and [7].…”

Section: Introductionmentioning

confidence: 99%

“…To the best of our knowledge, this is the first work that applies deep learning for contrast inflow detection in X-ray images. To validate the detection, the position of the beginning contrast frame (BCF) [8,9] in a sequence (where contrast starts being visible) was used in the experiment.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Fast Prospective Detection of Contrast Inflow in X-ray Angiograms with Convolutional Neural Network and Recurrent Neural Network

Ma¹,

Ambrosini²,

Walsum³

2017

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Automatic detection of contrast inflow in X-ray angiographic sequences can facilitate image guidance in computer-assisted cardiac interventions. In this paper, we propose two different approaches for prospective contrast inflow detection. The methods were developed and evaluated to detect contrast frames from X-ray sequences. The first approach trains a convolutional neural network (CNN) to distinguish whether a frame has contrast agent or not. The second method extracts contrast features from images with enhanced vessel structures; the contrast frames are then detected based on changes in the feature curve using long short-term memory (LSTM), a recurrent neural network architecture. Our experiments show that both approaches achieve good performance on detection of the beginning contrast frame from X-ray sequences and are more robust than a state-of-the-art method. As the proposed methods work in prospective settings and run fast, they have the potential of being used in clinical practice.

show abstract

Section: Methodsmentioning

confidence: 65%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Fast Prospective Detection of Contrast Inflow in X-ray Angiograms with Convolutional Neural Network and Recurrent Neural Network

Ma¹,

Ambrosini²,

Walsum³

2017

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…The contrasted area is found based on a difference image (DSA) I DSA = I u − I c , I ∈ ℝ m × n involving a frame I c that contains contrast agent and an uncontrasted frame I u . To distinguish between contrasted and uncontrasted frames, either manual annotation or a learning based method can be used, for example, the method described in [ 23 ]. Depending on the chosen contrasted frame I c , I DSA may contain artifacts, for example, due to motion of the diaphragm or from catheters, if they are at different positions in I u and I c .…”

Section: Registration Methodsmentioning

confidence: 99%

Contrast-Based 3D/2D Registration of the Left Atrium: Fast versus Consistent

Hoffmann

Kowalewski

Maier

et al. 2016

International Journal of Biomedical Imaging

Self Cite

View full text Add to dashboard Cite

For augmented fluoroscopy during cardiac ablation, a preoperatively acquired 3D model of a patient's left atrium (LA) can be registered to X-ray images recorded during a contrast agent (CA) injection. An automatic registration method that works also for small amounts of CA is desired. We propose two similarity measures: The first focuses on edges of the patient anatomy. The second computes a contrast agent distribution estimate (CADE) inside the 3D model and rates its consistency with the CA as seen in biplane fluoroscopic images. Moreover, temporal filtering on the obtained registration results of a sequence is applied using a Markov chain framework. Evaluation was performed on 11 well-contrasted clinical angiographic sequences and 10 additional sequences with less CA. For well-contrasted sequences, the error for all 73 frames was 7.9 ± 6.3 mm and it dropped to 4.6 ± 4.0 mm when registering to an automatically selected, well enhanced frame in each sequence. Temporal filtering reduced the error for all frames from 7.9 ± 6.3 mm to 5.7 ± 4.6 mm. The error was typically higher if less CA was used. A combination of both similarity measures outperforms a previously proposed similarity measure. The mean accuracy for well contrasted sequences is in the range of other proposed manual registration methods.

show abstract

“…The ECG of the patient is recorded along with the images. Standard algorithm as [4] can then be used to identify the subset of the images where the coronary images are well opacified with the contrast agent. In this subset, a reference sequence of about 10 to 15 images is then selected that covers a full cardiac cycle and includes best opacified images.…”

Section: Introductionmentioning

confidence: 99%

VOIDD: Automatic Vessel-of-Intervention Dynamic Detection in PCI Procedures

Bacchuwar

Cousty

Vaillant

et al. 2017

Lecture Notes in Computer Science

View full text Add to dashboard Cite

In this article, we present the work towards improving the overall workflow of the Percutaneous Coronary Interventions (PCI) procedures by capacitating the imaging instruments to precisely monitor the steps of the procedure. In the long term, such capabilities can be used to optimize the image acquisition to reduce the amount of dose or contrast media employed during the procedure. We present the automatic VOIDD algorithm to detect the vessel of intervention which is going to be treated during the procedure by combining information from the vessel image with contrast agent injection and images acquired during guidewire tip navigation. Due to the robust guidewire tip segmentation method, this algorithm is also able to automatically detect the sequence corresponding to guidewire navigation. We present an evaluation methodology which characterizes the correctness of the guide wire tip detection and correct identification of the vessel navigated during the procedure. On a dataset of 2213 images from 8 sequences of 4 patients, VOIDD identifies vesselof-intervention with accuracy in the range of 88% or above and absence of tip with accuracy in range of 98% or above depending on the test case.

show abstract

Robust Identification of Contrasted Frames in Fluoroscopic Images

Cited by 4 publications

References 7 publications

Fast Prospective Detection of Contrast Inflow in X-ray Angiograms with Convolutional Neural Network and Recurrent Neural Network

Fast Prospective Detection of Contrast Inflow in X-ray Angiograms with Convolutional Neural Network and Recurrent Neural Network

Contrast-Based 3D/2D Registration of the Left Atrium: Fast versus Consistent

VOIDD: Automatic Vessel-of-Intervention Dynamic Detection in PCI Procedures

Contact Info

Product

Resources

About