0000000000220960
AUTHOR
Stéphanie Bricq
Leveraging Uncertainty Estimates to Improve Segmentation Performance in Cardiac MR
In medical image segmentation, several studies have used Bayesian neural networks to segment and quantify the uncertainty of the images. These studies show that there might be an increased epistemic uncertainty in areas where there are semantically and visually challenging pixels. The uncertain areas of the image can be of a great interest as they can possibly indicate the regions of incorrect segmentation. To leverage the uncertainty information, we propose a segmentation model that incorporates the uncertainty into its learning process. Firstly, we generate the uncertainty estimate (sample variance) using Monte-Carlo dropout during training. Then we incorporate it into the loss function t…
Leveraging Uncertainty Estimates to Improve Segmentation Performance in Cardiac MR
International audience; In medical image segmentation, several studies have used Bayesian neural networks to segment and quantify the uncertainty of the images. These studies show that there might be an increased epistemic uncertainty in areas where there are semantically and visually challenging pixels. The uncertain areas of the image can be of a great interest as they can possibly indicate the regions of incorrect segmentation. To leverage the uncertainty information, we propose a segmentation model that incorporates the uncertainty into its learning process. Firstly, we generate the uncertainty estimate (sample variance) using Monte-Carlo dropout during training. Then we incorporate it …
Automatic classification of tissues on pelvic MRI based on relaxation times and support vector machine
International audience; Tissue segmentation and classification in MRI is a challenging task due to a lack of signal intensity standardization. MRI signal is dependent on the acquisition protocol, the coil profile, the scanner type, etc. While we can compute quantitative physical tissue properties independent of the hardware and the sequence parameters, it is still difficult to leverage these physical properties to segment and classify pelvic tissues. The proposed method integrates quantitative MRI values (T1 and T2 relaxation times and pure synthetic weighted images) and machine learning (Support Vector Machine (SVM)) to segment and classify tissues in the pelvic region, i.e.: fat, muscle, …
Automatic Quality Assessment of Cardiac MR Images with Motion Artefacts using Multi-task Learning and K-Space Motion Artefact Augmentation
The movement of patients and respiratory motion during MRI acquisition produce image artefacts that reduce the image quality and its diagnostic value. Quality assessment of the images is essential to minimize segmentation errors and avoid wrong clinical decisions in the downstream tasks. In this paper, we propose automatic multi-task learning (MTL) based classification model to detect cardiac MR images with different levels of motion artefact. We also develop an automatic segmentation model that leverages k-space based motion artefact augmentation (MAA) and a novel compound loss that utilizes Dice loss with a polynomial version of cross-entropy loss (PolyLoss) to robustly segment cardiac st…
Using Polynomial Loss and Uncertainty Information for Robust Left Atrial and Scar Quantification and Segmentation
Automatic and accurate segmentation of the left atrial (LA) cavity and scar can be helpful for the diagnosis and prognosis of patients with atrial fibrillation. However, automating the segmentation can be difficult due to the poor image quality, variable LA shapes, and small discrete regions of LA scars. In this paper, we proposed a fully-automatic method to segment LA cavity and scar from Late Gadolinium Enhancement (LGE) MRIs. For the loss functions, we propose two different losses for each task. To enhance the segmentation of LA cavity from the multicenter dataset, we present a hybrid loss that leverages Dice loss with a polynomial version of cross-entropy loss (PolyCE). We also utilize …
Multi-modal image fusion for small animal studies in in-line PET /3T MRI
Congrès sous l’égide de la Société Française de Génie Biologique et Médical (SFGBM).; National audience; In the framework of small animal multi-modal imaging, the current progression of the IMAPPI project is illustrated by the design of an in-line PET/MRI prototype, coupled to a dedicated multi-resolution registration method allowing the robust fusion of data coming from both modalities. The first results show a good alignment of the data from tumor imaging at the level of the abdomen.
Integrated PET/MRI in preclinical studies State of the art
International audience; The exquisite tissue contrast of magnetic resonance imaging (MRI), the absence of ionising radiation and the opportunity to obtain new molecular and functional data have strengthened the enthusiasm for coupling MRI rather than computed tomography (CT) to positron emission tomography (PET). When reviewing the current literature one might be surprised by the almost unlimited diversity of what is placed under the name of PET/MRI in the articles. The magnetic field is varying from 0.3 Tesla (T) to 9.4 T, the size of the bore varies also from the wide bore of clinical scanners to volumes limited to a few tens of mL. Many preclinical studies are performed using separate PE…
First steps toward the generation of PET/MR attenuation map in the case of prostate cancer
Congrès sous l’égide de la Société Française de Génie Biologique et Médical (SFGBM).; National audience; A new methodology providing the first step towards the generation of attenuation maps for PET/MR systems based solely on MR information is presented in this paper. From T1-and T2-weighted MR data set and anatomical-based knowledge, our method segments and classifies the attenuation-differing regions of the patient's pelvis using a robust implementation of the weighted fuzzy C-means algorithm. Providing no signal, particular process is performed for the bones. We have demonstrated the feasibility of this approach by correctly segmenting and classifying six attenuation-differing regions on…
Evaluation automatique du remodelage ventriculaire gauche à partir d'imagerie multimodale en pré-clinique et clinique.
Automatic evaluation of the left ventricular remodelling from MRI in pre-clinical and clinical practice.
Automatic classification of tissues using T1 and T2 relaxation times from prostate MRI: a step toward generation of PET/MR attenuation map
This paper presents a new methodology providing the first step towards generating attenuation maps for PET/MR systems based solely on MR information. The new method segments and classifies the attenuation-differing regions of the patient's pelvis based on acquired T 1 - and T 2 -weighted MR data sets and anatomical-based knowledge by computing the tissue specific T 1 and T 2 relaxation times, using a robust implementation of the weighted fuzzy C-means algorithm and applying a novel process to detect bones. We have demonstrated the feasibility of this approach by correctly segmenting and classifying six differing regions of structural and anatomical importance: fat, muscle, prostate, air, ba…
A Survey on Microaneurysms Detection in Color Fundus Images
Early Detection of Microaneurysms (MA) plays a vital role in preventing the blindness caused by diabetic retinopathy (DR). DR is preventable yet a serious diabetic problem. Treatment at an earlier stage reduces the risk of blindness. Microaneurysm is the first sign of DR found in fundus images while doing screening. Detection of MA is a challenging task mainly because of its size. MA appears as a tiny red spot ranging from 15µm to 60µm size. The most common way to detect the MA from a colour fundus image is by classification/segmentation through machine learning and deep learning approaches. The FROC-based performance evaluation shows that the existing methods can reach only up to 80% of se…
NCprocessing: a software to determine non-compacted and compacted masses from MRI
International audience
Semiautomatic detection of myocardial contours in order to investigate normal values of the left ventricular trabeculated mass using MRI
Purpose To propose, assess, and validate a semiautomatic method allowing rapid and reproducible measurement of trabeculated and compacted left ventricular (LV) masses from cardiac magnetic resonance imaging (MRI). Materials and Methods We developed a method to automatically detect noncompacted, endocardial, and epicardial contours. Papillary muscles were segmented using semiautomatic thresholding and were included in the compacted mass. Blood was removed from trabeculae using the same threshold tool. Trabeculated, compacted masses and ratio of noncompacted to compacted (NC:C) masses were computed. Preclinical validation was performed on four transgenic mice with hypertrabeculation of the LV…