A5022944122- Radiomics and Machine Learning in Medical Imaging
- Cancer Genomics and Diagnostics
- AI in cancer detection
- Genetic factors in colorectal cancer
- Colorectal Cancer Screening and Detection
- Statistical Methods and Inference
- Privacy-Preserving Technologies in Data
- Molecular Biology Techniques and Applications
- Renal cell carcinoma treatment
- Cell Image Analysis Techniques
- Pancreatic and Hepatic Oncology Research
- Bladder and Urothelial Cancer Treatments
- Cancer Immunotherapy and Biomarkers
- Machine Learning in Bioinformatics
- MRI in cancer diagnosis
- Lung Cancer Diagnosis and Treatment
- Breast Cancer Treatment Studies
- Genetic and Kidney Cyst Diseases
- Medical Image Segmentation Techniques
- Advanced X-ray and CT Imaging
- Biomedical Text Mining and Ontologies
- Digital Imaging for Blood Diseases
- Insurance, Mortality, Demography, Risk Management
- Artificial Intelligence in Healthcare and Education
- Urinary and Genital Oncology Studies
Deep learning methods for digital pathology analysis are an effective way to address multiple clinical questions, from diagnosis prediction of treatment outcomes. These have also been used predict gene mutations images, but no comprehensive evaluation their potential extracting molecular features histology slides has yet performed. We show that HE2RNA, a model based on the integration data modes, can be trained systematically RNA-Seq profiles whole-slide images alone, without expert...
Standardized and robust risk-stratification systems for patients with hepatocellular carcinoma (HCC) are required to improve therapeutic strategies investigate the benefits of adjuvant systemic therapies after curative resection/ablation.In this study, we used two deep-learning algorithms based on whole-slide digitized histological slides (whole-slide imaging; WSI) build models predicting survival HCC treated by surgical resection. Two independent series were investigated: a discovery set...
Computational pathology is revolutionizing the field of by integrating advanced computer vision and machine learning technologies into diagnostic workflows. It offers unprecedented opportunities for improved efficiency in treatment decisions allowing pathologists to achieve higher precision objectivity disease classification, tumor microenvironment description identification new biomarkers. However, potential computational personalized medicine comes with significant challenges, particularly...
Mismatch Repair Deficiency (dMMR)/Microsatellite Instability (MSI) is a key biomarker in colorectal cancer (CRC). Universal screening of CRC patients for MSI status now recommended, but contributes to increased workload pathologists and delayed therapeutic decisions. Deep learning has the potential ease dMMR/MSI testing accelerate oncologist decision making clinical practice, yet no comprehensive validation clinically approved tool been conducted. We developed MSIntuit, artificial...
Two tumor (Classical/Basal) and stroma (Inactive/active) subtypes of Pancreatic adenocarcinoma (PDAC) with prognostic theragnostic implications have been described. These molecular were defined by RNAseq, a costly technique sensitive to sample quality cellularity, not used in routine practice. To allow rapid PDAC subtyping study heterogeneity, we develop PACpAInt, multi-step deep learning model. PACpAInt is trained on multicentric cohort (n = 202) validated 4 independent cohorts including...
In recent years, the advent of foundation models (FM) for digital pathology has relied heavily on scaling pre-training datasets and model size, yielding large powerful models. While it resulted in improving performance diverse downstream tasks, also introduced increased computational cost inference time. this work, we explore distillation a into smaller one, reducing number parameters by several orders magnitude. Leveraging techniques, our distilled model, H0-mini, achieves nearly comparable...
Microsatellite instability (MSI) is a tumor phenotype whose diagnosis largely impacts patient care in colorectal cancers (CRC), and associated with response to immunotherapy all solid tumors. Deep learning models detecting MSI tumors directly from H&E stained slides have shown promise improving of patients. Prior deep for detection relied on neural networks pretrained ImageNet dataset, which does not contain any medical image. In this study, we leverage recent advances self-supervised by...
Pathogenic activating mutations in the fibroblast growth factor receptor 3 (FGFR3) drive disease maintenance and progression urothelial cancer. 10-15% of muscle-invasive metastatic cancer (MIBC/mUC) are FGFR3-mutant. Selective targeting FGFR3 hotspot with tyrosine kinase inhibitors (e.g., erdafitinib) is approved for mUC requires mutational testing. However, current testing assays (polymerase chain reaction or next-generation sequencing) necessitate high tissue quality, have long turnover...
Building machine learning models from decentralized datasets located in different centers with federated (FL) is a promising approach to circumvent local data scarcity while preserving privacy. However, the prominent Cox proportional hazards (PH) model, used for survival analysis, does not fit FL framework, as its loss function non-separable respect samples. The na\"ive method bypass this non-separability consists calculating losses per center, and minimizing their sum an approximation of...
ABSTRACT Objective Mismatch Repair Deficiency (dMMR) / Microsatellite Instability (MSI) is a key biomarker in colorectal cancer (CRC). Universal screening of CRC patients for dMMR/MSI status now recommended, but contributes to increased workload pathologists and delayed therapeutic decisions. Deep learning has the potential ease testing clinical practice, yet no comprehensive validation clinically approved tool been conducted. Design We developed an MSI pre-screening tool, MSIntuit, that...
Deep learning methods for digital pathology analysis have proved an effective way to address multiple clinical questions, from diagnosis prognosis and even prediction of treatment outcomes. They also recently been used predict gene mutations images, but no comprehensive evaluation their potential extracting molecular features histology slides has yet performed. We propose a novel approach based on the integration data modes, show that our deep model, HE2RNA, can be trained systematically...
Gathering histopathology slides from over 100 publicly available cohorts, we compile a diverse dataset of 460 million pathology tiles covering more than 30 cancer sites. Using this dataset, train large self-supervised vision transformer using DINOv2 and release one iteration model for further experimentation, coined Phikon-v2. While trained on histology slides, Phikon-v2 surpasses our previously released (Phikon) performs par with other foundation models (FM) proprietary data. Our benchmarks...
ABSTRACT Background Correctly classifying early estrogen receptor-positive and HER2-negative (ER+/HER2) breast cancer (EBC) cases allows to propose an adapted adjuvant systemic treatment strategy. We developed a new AI-based tool assess the risk of distant relapse at 5 years for ER+/HER2-EBC patients from pathological slides. Patients Methods The discovery dataset (GrandTMA) included 1429 patients, with long-term follow-up available hematoxylin-eosin saffron (HES) whole slide image (WSI). A...
Abstract Pancreatic ductal adenocarcinoma (PAC) is a highly heterogeneous and plastic tumor with different transcriptomic molecular subtypes that hold great prognostic theranostic values. We developed PACpAInt, multistep approach using deep learning models to determine cell type their phenotype on routine histological preparation at resolution enabling decipher complete intratumor heterogeneity massive scale never achieved before. PACpAInt effectively identified the slide level in three...
Abstract Today, pathology imaging is one of the most common and inexpensive diagnostic/prognostic tools used in oncology, while more sophisticated methods such as next generation sequencing (NGS) remain relatively expensive not routinely a clinical setting. Deep convolutional neural networks (CNNs) have emerged an important image analysis technology enhancing workflow pathologists improving prediction patient prognosis response to treatment. Recently, few attempts been made predict...
4141 Background: Pancreatic adenocarcinoma (PAC) is predicted to be the second cause of death by cancer in 2030 and its prognosis has seen little improvement last decades. PAC a very heterogeneous tumor with preeminent stroma multiple histological aspects. Omic studies confirmed molecular heterogeneity, possibly one main factors explaining failure most clinical trials. Two three transcriptomic subtypes cells respectively, were described major prognostic predictive implications. The subtypes,...
e16580 Background: Gain of function mutations the FGFR3 gene have been reported to be important driver in a subset muscle-invasive bladder urothelial cancer (MIBC). Subsequently, specific kinase inhibitors (e.g., erdafitinib) developed target mutant metastasized MIBC with promising clinical activity. mutational status testing is required for erdafitinib treatment, but are rare metastatic (10-15%). Thus, cheap and fast pre-screening tools such as artificial intelligence (AI) based image...