A5066627589- Privacy-Preserving Technologies in Data
- AI in cancer detection
- Advanced Neural Network Applications
- Radiomics and Machine Learning in Medical Imaging
- Breast Lesions and Carcinomas
- Advanced Image and Video Retrieval Techniques
- Breast Cancer Treatment Studies
- Image and Object Detection Techniques
- Gastric Cancer Management and Outcomes
- Gastrointestinal Tumor Research and Treatment
- Advanced Causal Inference Techniques
- Statistical Methods in Clinical Trials
- Infrared Target Detection Methodologies
- Privacy, Security, and Data Protection
- Remote Sensing and LiDAR Applications
- Video Surveillance and Tracking Methods
- Computational Drug Discovery Methods
- Sarcoma Diagnosis and Treatment
- Health Systems, Economic Evaluations, Quality of Life
- Statistical Methods and Inference
- Medical Imaging and Analysis
- Big Data Technologies and Applications
- Advanced Measurement and Detection Methods
- Ethics in Clinical Research
- MRI in cancer diagnosis
Daikin (United States)
2021
GREYC
2018
Safran Electronics (Canada)
2017-2018
Université de Caen Normandie
2017
École Nationale Supérieure d'Ingénieurs de Caen
2017
Normandie Université
2017
Centre National de la Recherche Scientifique
2017
Object detection-the computer vision task dealing with detecting instances of objects a certain class (e.g., 'car', 'plane', etc.) in images-attracted lot attention from the community during last 5 years. This strong interest can be explained not only by importance this has for many applications but also phenomenal advances area since arrival deep convolutional neural networks (DCNN). article reviews recent literature on object detection CNN, comprehensive way, and provides an in-depth view...
Federated Learning (FL) is a novel approach enabling several clients holding sensitive data to collaboratively train machine learning models, without centralizing data. The cross-silo FL setting corresponds the case of few ($2$--$50$) reliable clients, each medium large datasets, and typically found in applications such as healthcare, finance, or industry. While previous works have proposed representative datasets for cross-device FL, realistic healthcare exist, thereby slowing algorithmic...
Risk assessment of gastrointestinal stromal tumor (GIST) according to the AFIP/Miettinen classification and mutational profiling are major tools for patient management. However, depends heavily on mitotic counts, which is laborious sometimes inconsistent between pathologists. It has also been shown be imperfect in stratifying patients. Molecular testing costly time-consuming, therefore, not systematically performed all countries. New methods improve risk molecular predictions hence crucial...
Detecting small vehicles in aerial images is a difficult job that can be challenging even for humans. Rotating objects, low resolution, inter-class variability and very large comprising complicated backgrounds render the work of photo-interpreters tedious wearisome. Unfortunately best classical detection pipelines like Faster R-CNN cannot used off-the-shelf with good results because they were built to process object centric from day-to-day life multi-scale vertical objects. In this we build...
Since 2014, the NIH funded iDASH (integrating Data for Analysis, Anonymization, SHaring) National Center Biomedical Computing has hosted yearly competitions on topic of private computing genomic data. For one track 2020 iteration this competition, participants were challenged to produce an approach federated learning (FL) training cancer prediction models using differential privacy (DP), with submissions ranked according held-out test accuracy a given set DP budgets. More precisely, in...
This paper addresses the question of detection small targets (vehicles) in ortho-images. differs from general task detecting objects images by several aspects. First, vehicles to be detected are small, typically smaller than 20×20 pixels. Second, due multifarious-ness landscapes earth, pixel structures similar that a vehicle might emerge (roof tops, shadow patterns, rocks, buildings), whereas within class inter-class variability is limited as they all look alike afar. Finally, imbalance...
We consider a cross-silo federated learning (FL) setting where machine model with fully connected first layer is trained between different clients and central server using FedAvg, the aggregation step can be performed secure (SA). present SRATTA an attack relying only on aggregated models which, under realistic assumptions, (i) recovers data samples from clients, (ii) groups coming same client together. While sample recovery has already been explored in FL setting, ability to group per...
The Yeo-Johnson (YJ) transformation is a standard parametrized per-feature unidimensional often used to Gaussianize features in machine learning. In this paper, we investigate the problem of applying YJ cross-silo Federated Learning setting under privacy constraints. For first time, prove that negative log-likelihood fact convex, which allows us optimize it with exponential search. We numerically show resulting algorithm more stable than state-of-the-art approach based on Brent minimization...
<title>Abstract</title> External control arms (ECA) can inform the early clinical development of experimental drugs and provide efficacy evidence for regulatory approval. However, main challenge in implementing ECA lies accessing real-world or historical trials data. Indeed, regulations protecting patients' rights by strictly controlling data processing make pooling from multiple sources a central server often difficult. To address these limitations, we develop new method, 'FedECA' that...
1 Abstract Triple-Negative Breast Cancer (TNBC) is a rare cancer, characterized by high metastatic potential and poor prognosis, has limited treatment options compared to other breast cancers. The current standard of care in non-metastatic settings neoadjuvant chemotherapy (NACT), with the goal breast-conserving surgery for an vivo assessment chemosensitivity. However, efficacy this varies significantly across patients, histological response heterogeneity still poorly understood partly due...
590 Background: Triple-Negative Breast Cancer (TNBC) is characterized by high metastatic potential and poor prognosis with limited treatment options. Neoadjuvant chemotherapy (NACT) the standard of care in non-metastastic setting due to ability assess pathologic responses providing important prognostic information guidance adjuvant therapy decisions. However, histological response heterogeneity still poorly understood. We investigate use Machine Learning (ML) predict from diagnosis...
External control arms (ECA) can inform the early clinical development of experimental drugs and provide efficacy evidence for regulatory approval. However, main challenge in implementing ECA lies accessing real-world or historical trials data. Indeed, regulations protecting patients' rights by strictly controlling data processing make pooling from multiple sources a central server often difficult. To address these limitations, we develop new method, 'FedECA' that leverages federated learning...
While federated learning is a promising approach for training deep models over distributed sensitive datasets, it presents new challenges machine learning, especially when applied in the medical domain where multi-centric data heterogeneity common. Building on previous adaptation works, this paper proposes novel architectures via introduction of local-statistic batch normalization (BN) layers, resulting collaboratively-trained, yet center-specific models. This strategy improves robustness to...
Abstract Background. Most gastrointestinal stromal tumors (GIST) have activating mutations of KIT or PDGFRA genes and imatinib has proven effective in palliative adjuvant situations. Risk assessment according to the AFIP/Miettinen classification mutational profiling are major tools for management patients with GIST. However, depends heavily on mitotic counts, which is laborious sometimes inconsistent between pathologists. It also been shown be imperfect stratifying patients. Molecular...