A5086241341- Biometric Identification and Security
- Face recognition and analysis
- Forensic and Genetic Research
- Advanced Neural Network Applications
- Visual Attention and Saliency Detection
- Radiomics and Machine Learning in Medical Imaging
- Forensic Fingerprint Detection Methods
- Forensic Anthropology and Bioarchaeology Studies
- Glioma Diagnosis and Treatment
- Brain Tumor Detection and Classification
- Face Recognition and Perception
- Medical Imaging and Analysis
- Generative Adversarial Networks and Image Synthesis
- User Authentication and Security Systems
- Medical Imaging Techniques and Applications
- Adversarial Robustness in Machine Learning
- Artificial Intelligence in Healthcare and Education
- Face and Expression Recognition
- Domain Adaptation and Few-Shot Learning
- Explainable Artificial Intelligence (XAI)
- 3D Surveying and Cultural Heritage
- Seed Germination and Physiology
- Industrial Vision Systems and Defect Detection
- Video Analysis and Summarization
- Neural Networks and Applications
Brigham and Women's Hospital
2023-2024
Boston Children's Hospital
2023-2024
Mass General Brigham
2023-2024
Harvard University
2023-2024
Dana-Farber Cancer Institute
2023-2024
Dana-Farber Brigham Cancer Center
2023-2024
Intel (United States)
2023-2024
University of Notre Dame
2019-2023
Purpose To develop and externally test a scan-to-prediction deep learning pipeline for noninvasive, MRI-based
Can deep learning models achieve greater generalization if their training is guided by reference to human perceptual abilities? And how can we implement this in a practical manner? This paper proposes strategy ConveY Brain Oversight Raise Generalization (CYBORG). new approach incorporates human-annotated saliency maps into loss function that guides the model's focus on image regions humans deem salient for task. The Class Activation Mapping (CAM) mechanism used probe current each batch,...
Purpose To develop, externally test, and evaluate clinical acceptability of a deep learning pediatric brain tumor segmentation model using stepwise transfer learning. Materials Methods In this retrospective study, the authors leveraged two T2-weighted MRI datasets (May 2001 through December 2015) from national consortium (n = 184; median age, 7 years [range, 1-23 years]; 94 male patients) cancer center 100; 8 1-19 47 to develop neural networks for low-grade glioma approach maximize...
Modern deep learning techniques can be employed to generate effective feature extractors for the task of iris recognition. The question arises: should we train such structures from scratch on a relatively large image dataset, or it is better fine-tune existing models adapt them new domain? In this work explore five different sets weights popular ResNet-50 architecture find out whether iris-specific perform than trained non-iris tasks. Features are extracted each convolutional layer and...
Launched in 2013, LivDet-Iris is an international competition series open to academia and industry with the aim assess report advances iris Presentation Attack Detection (PAD). This paper presents results from fourth of series: 2020. year's introduced several novel elements: (a) incorporated new types attacks (samples displayed on a screen, cadaver eyes prosthetic eyes), (b) initiated as on-going effort, testing protocol available now everyone via Biometrics Evaluation Testing (BEAT)*...
Abstract Background Postoperative recurrence risk for pediatric low-grade gliomas (pLGGs) is challenging to predict by conventional clinical, radiographic, and genomic factors. We investigated if deep learning (DL) of magnetic resonance imaging (MRI) tumor features could improve postoperative pLGG stratification. Methods used a pretrained DL tool designed segmentation extract from preoperative T2-weighted MRI patients who underwent surgery (DL-MRI features). Patients were pooled 2...
The automatic generation of representative natural language descriptions for observable patterns in time series data enhances interpretability, simplifies analysis and increases cross-domain utility temporal data. While pre-trained foundation models have made considerable progress processing (NLP) computer vision (CV), their application to has been hindered by scarcity. Although several large model (LLM)-based methods proposed forecasting, captioning is under-explored the context LLMs. In...
Deep learning has driven remarkable accuracy increases in many computer vision problems. One ongoing challenge is how to achieve the greatest cases where training data limited. A second that trained models oftentimes do not generalize well even new subjectively similar set. We address these challenges a novel way, with first-ever (to our knowledge) exploration of encoding human judgement about salient regions images into data. compare and generalization state-of-the-art deep algorithm for...
ABSTRACT Purpose Artificial intelligence (AI)-automated tumor delineation for pediatric gliomas would enable real-time volumetric evaluation to support diagnosis, treatment response assessment, and clinical decision-making. Auto-segmentation algorithms tumors are rare, due limited data availability, have yet demonstrate translation. Methods We leveraged two datasets from a national brain consortium (n=184) cancer center (n=100) develop, externally validate, clinically benchmark deep learning...
Post-mortem biometrics entails utilizing the biometric data of a deceased individual for determining or verifying human identity. Due to fundamental biological changes that occur in person's traits after death, post-mortem can be significantly different from ante-mortem data, introducing new challenges sensors, feature extractors and matchers. This paper surveys research date on problem using iris images acquired death automated recognition. A comprehensive review existing literature is...
Face image synthesis has progressed beyond the point at which humans can effectively distinguish authentic faces from synthetically-generated ones. Recently developed synthetic face detectors boast ``better-than-human'' discriminative ability, especially those guided by human perceptual intelligence during model's training process. In this paper, we investigate whether these human-guided assist non-expert operators in task of detection when compared to models trained without human-guidance....
Forensic iris recognition, as opposed to live is an emerging research area that leverages the discriminative power of biometrics aid human examiners in their efforts identify deceased persons. As a machine learning-based technique predominantly human-controlled task, forensic recognition serves "back-up" expertise task post-mortem identification. such, learning model must be (a) interpretable, and (b) post-mortem-specific, account for changes decaying eye tissue. In this work, we propose...
Research in presentation attack detection (PAD) for iris recognition has largely moved beyond evaluation "closed-set" scenarios, to emphasize ability generalize types not present the training data. This paper offers multiple contributions understand and extend state-of-the-art open-set PAD. First, it describes most authoritative date of We have curated largest publicly-available image dataset this problem, drawing from 26 benchmarks previously released by various groups, adding 150,000...
Iris recognition of living individuals is a mature biometric modality that has been adopted globally from governmental ID programs, border crossing, voter registration and de-duplication, to unlocking mobile phones. On the other hand, possibility recognizing deceased subjects with their iris patterns emerged recently. In this paper, we present an end-to-end deep learning-based method for postmortem segmentation special visualization technique intended support forensic human examiners in...
The performance of convolutional neural networks has continued to improve over the last decade. At same time, as model complexity grows, it becomes increasingly more difficult explain decisions. Such explanations may be critical importance for reliable operation human-machine pairing setups, or selection when "best'" among many equally-accurate models must established. Saliency maps represent one popular way explaining decisions by highlighting image regions deem important making a...
This paper describes the results of 2023 edition "LivDet" series iris presentation attack detection (PAD) competitions. New elements in this fifth competition include (1) GAN-generated images as a category instruments (PAI), and (2) an evaluation human accuracy at detecting PAI reference benchmark. Clarkson University Notre Dame contributed image datasets for competition, composed samples representing seven different categories, well baseline PAD algorithms. Fraunhofer IGD, Beijing Civil...
Can deep learning models achieve greater generalization if their training is guided by reference to human perceptual abilities? And how can we implement this in a practical manner? This paper proposes strategy ConveY Brain Oversight Raise Generalization (CYBORG). new approach incorporates human-annotated saliency maps into loss function that guides the model's focus on image regions humans deem salient for task. The Class Activation Mapping (CAM) mechanism used probe current each batch,...
To develop and externally validate a scan-to-prediction deep-learning pipeline for noninvasive, MRI-based BRAF mutational status classification pLGG.
Abstract Pediatric low-grade gliomas (pLGGs) have heterogeneous outcomes, making individualized clinical decision-making challenging. Imaging-based biomarkers for progression and/or recurrence may better risk-stratify pLGG patients and guide management. We developed externally validated a deep learning (DL) algorithm to predict event-free survival (EFS) in using multi-institutional databases. collected linked, T2-weighted diagnostic Magnetic Resonance Images (MRI) data with WHO grade 1-2...
2066 Background: Pediatric low-grade gliomas (pLGGs) have heterogeneous clinical presentations and prognoses. Given the morbidity of treatment, some suspected pLGGs, especially those found incidentally, are surveilled without though natural histories these tumors yet to be systematically studied. We leveraged deep learning multi-institutional data methodically analyze longitudinal volumetric trajectories pLGGs on surveillance, yielding insights into their growth implications. Methods:...
Abstract BACKGROUND Pediatric low-grade gliomas (pLGGs) have heterogeneous clinical presentations and prognoses. Given the morbidity of treatment, suspected pLGGs are surveilled without though natural histories these tumors yet to be systematically studied. METHODS We conducted a pooled, retrospective study pLGG patients diagnosed between 1992 2020 from two sources (Dana-Farber Cancer Institute/Boston Children’s Hospital Brain Tumor Network), who were untreated for at least one-year...