arXiv subjects

Carlos Fernandez-Granda

Publications and source records attributed to Carlos Fernandez-Granda.

At least 19 recordsLinked to original sources

arXiv2026

Comparative Study of Anatomical and Learned Features in AI Models for Structural Brain MRI

In this work, we comprehensively evaluate three popular feature-extraction paradigms in AI-based neuroimaging modeling: (1) computation of anatomical surfaces and volumes, (2) supervised learning with convolutional neural networks (CNNs), and (3) unsupervised pretraining of vision transformer (ViT) foundation models, followed by supervised finetuning. Our study is based on 18 publicly available datasets containing 3D structural T1-weighted MRI scans from approximately 80,000 participants across seven distinct clinical tasks. We observe that a linear model based on anatomical features matches the diagnostic performance of complex nonlinear features learned by sophisticated AI frameworks, including foundation models trained on thousands of scans. Conversely, CNNs and pretrained ViTs learn features that implicitly capture relevant anatomical information, bypassing the need for explicit feature extraction. Building upon these insights, we propose Anatomy Segmentation Pretraining (ASP), a novel method to incorporate anatomical information during foundation-model pretraining, which outperforms existing models in biological age estimation.

cs.CV↗

arXiv2026

Pitfalls of Administrative Censoring in Survival Models with Time-Indexed Inputs

Survival models can model time-to-event outcomes using partially observed data. They are widely used in clinical prediction, including cancer risk, disease progression, treatment response, and mortality. Recent models often rely on rich inputs collected at a specific clinical encounter, such as medical images, laboratory tests, electronic health record snapshots, or sensor measurements. In large retrospective datasets, these inputs are usually collected over many calendar years. As a result, they may contain clues about when they were acquired through changes in devices, protocols, documentation, patient mix, or clinical practice. This creates a potential failure mode when outcomes are observed only up to a fixed study end date. More recent records necessarily have less potential follow-up than older records. A model that can infer the record date from the input may therefore learn to predict how much follow-up was available rather than the patient's true risk of experiencing the event. We call this failure mode administrative-cutoff leakage. In this paper, we characterize when this leakage can occur, distinguish it from classical informative censoring and genuine temporal changes in risk, and propose practical ways to detect it. In simulations, we show that administrative-cutoff leakage can inflate fixed-horizon AUC and can also affect Harrell's C-index under realistic follow-up patterns. We then demonstrate the same behavior in a real mammography cohort. These results motivate a simple design principle for survival prediction: for an n-year prediction task, the dataset should provide at least n years of potential follow-up after the latest input date. Otherwise, the models may be subject to bias induced by administrative-cutoff leakage.

cs.LG↗

arXiv2026

Samudra 2: Scaling Ocean Emulators across Resolutions

Ocean general circulation models (OGCMs) are essential to climate science but computationally expensive, limiting ensemble size and forcing scenarios. Neural emulators promise orders-of-magnitude speedups, yet existing ocean emulators have not combined fine spatial resolution with multi-year autoregressive rollouts. Samudra, the first autoregressive neural ocean emulator to produce multi-decade global rollouts, is limited to $1^\circ$ resolution and exhibits two long-horizon failure modes: \emph{variance collapse}, the loss of temporal variability, and \emph{imprinting artifacts}, in which velocity patterns leak into deep-ocean fields. We present Samudra 2, which introduces a wider U-Net backbone with modified ConvNeXt-style blocks and a reduced block-internal expansion factor, together with a dynamic loss that reweights output channels according to their prediction errors, strengthening gradients for slow-evolving deep-ocean fields. At $1^\circ$, Samudra 2 increases upper-ocean global-mean temperature $R^2$ from 0.56 to 0.87 and reduces deep-ocean temperature error by roughly sevenfold. The same architecture scales to $1/2^\circ$ and $1/4^\circ$ over approximately 8-year autoregressive rollouts, recovering mesoscale eddies and sharp western boundary currents. Running on a single GPU, Samudra 2 enables larger ensembles for sea-level projections, ocean heat uptake, and climate variability studies. All artifacts are publicly available: project page, code, checkpoints, documentation.

cs.CE↗

arXiv2026

SamudrACE: Fast and Accurate Coupled Climate Modeling with 3D Ocean and Atmosphere Emulators

Traditional numerical global climate models simulate the full Earth system by exchanging boundary conditions between separate simulators of the atmosphere, ocean, sea ice, land surface, and other geophysical processes. This paradigm allows for distributed development of individual components within a common framework, unified by a coupler that handles translation between realms via spatial or temporal alignment and flux exchange. Following a similar approach adapted for machine learning-based emulators, we present SamudrACE: a coupled global climate model emulator which produces centuries-long simulations at 1-degree horizontal, 6-hourly atmospheric, and 5-daily oceanic resolution, with 145 2D fields spanning 8 atmospheric and 19 oceanic vertical levels, plus sea ice, surface, and top-of-atmosphere variables. SamudrACE is highly stable and has low climate biases comparable to those of its components with prescribed boundary forcing, with realistic variability in coupled climate phenomena such as ENSO that is not possible to simulate in uncoupled mode.

physics.ao-ph↗

arXiv2026

Atomic Depth Estimation From Noisy Electron Microscopy Data Via Deep Learning

We present a novel approach for extracting 3D atomic-level information from transmission electron microscopy (TEM) images affected by significant noise. The approach is based on formulating depth estimation as a semantic segmentation problem. We address the resulting segmentation problem by training a deep convolutional neural network to generate pixel-wise depth segmentation maps using simulated data corrupted by synthetic noise. The proposed method was applied to estimate the depth of atomic columns in CeO2 nanoparticles from simulated images and real-world TEM data. Our experiments show that the resulting depth estimates are accurate, calibrated and robust to noise.

cs.CV↗

arXiv2025

The Potential and Limitations of Vision-Language Models for Human Motion Understanding: A Case Study in Data-Driven Stroke Rehabilitation

Vision-language models (VLMs) have demonstrated remarkable performance across a wide range of computer-vision tasks, sparking interest in their potential for digital health applications. Here, we apply VLMs to two fundamental challenges in data-driven stroke rehabilitation: automatic quantification of rehabilitation dose and impairment from videos. We formulate these problems as motion-identification tasks, which can be addressed using VLMs. We evaluate our proposed framework on a cohort of 29 healthy controls and 51 stroke survivors. Our results show that current VLMs lack the fine-grained motion understanding required for precise quantification: dose estimates are comparable to a baseline that excludes visual information, and impairment scores cannot be reliably predicted. Nevertheless, several findings suggest future promise. With optimized prompting and post-processing, VLMs can classify high-level activities from a few frames, detect motion and grasp with moderate accuracy, and approximate dose counts within 25% of ground truth for mildly impaired and healthy participants, all without task-specific training or finetuning. These results highlight both the current limitations and emerging opportunities of VLMs for data-driven stroke rehabilitation and broader clinical video analysis.

cs.CV↗

arXiv2025

A Framework for Hybrid Physics-AI Coupled Ocean Models

Climate simulations, at all grid resolutions, rely on approximations that encapsulate the forcing due to unresolved processes on resolved variables, known as parameterizations. Parameterizations often lead to inaccuracies in climate models, with significant biases in the physics of key climate phenomena. Advances in artificial intelligence (AI) are now directly enabling the learning of unresolved processes from data to improve the physics of climate simulations. Here, we introduce a flexible framework for developing and implementing physics- and scale-aware machine learning parameterizations within climate models. We focus on the ocean and sea-ice components of a state-of-the-art climate model by implementing a spectrum of data-driven parameterizations, ranging from complex deep learning models to more interpretable equation-based models. Our results showcase the viability of AI-driven parameterizations in operational models, advancing the capabilities of a new generation of hybrid simulations, and include prototypes of fully coupled atmosphere-ocean-sea-ice hybrid simulations. The tools developed are open source, accessible, and available to all.

physics.ao-ph↗

arXiv2025

Leveraging Fine-Tuned Large Language Models for Interpretable Pancreatic Cystic Lesion Feature Extraction and Risk Categorization

Background: Manual extraction of pancreatic cystic lesion (PCL) features from radiology reports is labor-intensive, limiting large-scale studies needed to advance PCL research. Purpose: To develop and evaluate large language models (LLMs) that automatically extract PCL features from MRI/CT reports and assign risk categories based on guidelines. Materials and Methods: We curated a training dataset of 6,000 abdominal MRI/CT reports (2005-2024) from 5,134 patients that described PCLs. Labels were generated by GPT-4o using chain-of-thought (CoT) prompting to extract PCL and main pancreatic duct features. Two open-source LLMs were fine-tuned using QLoRA on GPT-4o-generated CoT data. Features were mapped to risk categories per institutional guideline based on the 2017 ACR White Paper. Evaluation was performed on 285 held-out human-annotated reports. Model outputs for 100 cases were independently reviewed by three radiologists. Feature extraction was evaluated using exact match accuracy, risk categorization with macro-averaged F1 score, and radiologist-model agreement with Fleiss' Kappa. Results: CoT fine-tuning improved feature extraction accuracy for LLaMA (80% to 97%) and DeepSeek (79% to 98%), matching GPT-4o (97%). Risk categorization F1 scores also improved (LLaMA: 0.95; DeepSeek: 0.94), closely matching GPT-4o (0.97), with no statistically significant differences. Radiologist inter-reader agreement was high (Fleiss' Kappa = 0.888) and showed no statistically significant difference with the addition of DeepSeek-FT-CoT (Fleiss' Kappa = 0.893) or GPT-CoT (Fleiss' Kappa = 0.897), indicating that both models achieved agreement levels on par with radiologists. Conclusion: Fine-tuned open-source LLMs with CoT supervision enable accurate, interpretable, and efficient phenotyping for large-scale PCL research, achieving performance comparable to GPT-4o.

cs.AI↗

arXiv2025

Samudra: An AI Global Ocean Emulator for Climate

AI emulators for forecasting have emerged as powerful tools that can outperform conventional numerical predictions. The next frontier is to build emulators for long climate simulations with skill across a range of spatiotemporal scales, a particularly important goal for the ocean. Our work builds a skillful global emulator of the ocean component of a state-of-the-art climate model. We emulate key ocean variables, sea surface height, horizontal velocities, temperature, and salinity, across their full depth. We use a modified ConvNeXt UNet architecture trained on multi-depth levels of ocean data. We show that the ocean emulator - Samudra - which exhibits no drift relative to the truth, can reproduce the depth structure of ocean variables and their interannual variability. Samudra is stable for centuries and 150 times faster than the original ocean model. Samudra struggles to capture the correct magnitude of the forcing trends and simultaneously remain stable, requiring further work.

physics.ao-ph↗

arXiv2025

Understanding differences in applying DETR to natural and medical images

Transformer-based detectors have shown success in computer vision tasks with natural images. These models, exemplified by the Deformable DETR, are optimized through complex engineering strategies tailored to the typical characteristics of natural scenes. However, medical imaging data presents unique challenges such as extremely large image sizes, fewer and smaller regions of interest, and object classes which can be differentiated only through subtle differences. This study evaluates the applicability of these transformer-based design choices when applied to a screening mammography dataset that represents these distinct medical imaging data characteristics. Our analysis reveals that common design choices from the natural image domain, such as complex encoder architectures, multi-scale feature fusion, query initialization, and iterative bounding box refinement, do not improve and sometimes even impair object detection performance in medical imaging. In contrast, simpler and shallower architectures often achieve equal or superior results. This finding suggests that the adaptation of transformer models for medical imaging data requires a reevaluation of standard practices, potentially leading to more efficient and specialized frameworks for medical diagnosis.

cs.CV↗

arXiv2025

Revealing Information from Weak Signal in Electron Energy-Loss Spectroscopy with a Deep Denoiser

Electron energy-loss spectroscopy (EELS) coupled with scanning transmission electron microscopy (STEM) is a powerful technique to determine materials composition and bonding with high spatial resolution. Noise is often a limitation especially with the increasing sophistication of EELS experiments. The signal characteristics from direct electron detectors provide a new opportunity to design superior denoisers. We have developed a CNN based denoiser, the unsupervised deep video denoiser (UDVD), which can be applied to EELS datasets acquired with direct electron detectors. We described UDVD and explained how to adapt the denoiser to energy-loss spectral series. To benchmark the performance of the denoiser on EELS datasets, we generated and denoised a set of simulated spectra. We demonstrate the charge spreading effect associated with pixel interfaces on direct electron detectors, which leads to artifacts after denoising. To suppress such artifacts, we propose some adjustments. We demonstrate the effectiveness of the denoiser using two challenging real data examples: mapping Gd dopants in $CeO_2$ nanoparticles and characterizing vibrational modes in hexagonal boron nitride (h-BN) with atomic resolution.

physics.ins-det↗

arXiv2025

Black Box Causal Inference: Effect Estimation via Meta Prediction

Causal inference and the estimation of causal effects plays a central role in decision-making across many areas, including healthcare and economics. Estimating causal effects typically requires an estimator that is tailored to each problem of interest. But developing estimators can take significant effort for even a single causal inference setting. For example, algorithms for regression-based estimators, propensity score methods, and doubly robust methods were designed across several decades to handle causal estimation with observed confounders. Similarly, several estimators have been developed to exploit instrumental variables (IVs), including two-stage least-squares (TSLS), control functions, and the method-of-moments. In this work, we instead frame causal inference as a dataset-level prediction problem, offloading algorithm design to the learning process. The approach we introduce, called black box causal inference (BBCI), builds estimators in a black-box manner by learning to predict causal effects from sampled dataset-effect pairs. We demonstrate accurate estimation of average treatment effects (ATEs) and conditional average treatment effects (CATEs) with BBCI across several causal inference problems with known identification, including problems with less developed estimators.

cs.LG↗

arXiv2025

Multi-modal AI for comprehensive breast cancer prognostication

Treatment selection in breast cancer is guided by molecular subtypes and clinical characteristics. However, current tools including genomic assays lack the accuracy required for optimal clinical decision-making. We developed a novel artificial intelligence (AI)-based approach that integrates digital pathology images with clinical data, providing a more robust and effective method for predicting the risk of cancer recurrence in breast cancer patients. Specifically, we utilized a vision transformer pan-cancer foundation model trained with self-supervised learning to extract features from digitized H&E-stained slides. These features were integrated with clinical data to form a multi-modal AI test predicting cancer recurrence and death. The test was developed and evaluated using data from a total of 8,161 female breast cancer patients across 15 cohorts originating from seven countries. Of these, 3,502 patients from five cohorts were used exclusively for evaluation, while the remaining patients were used for training. Our test accurately predicted our primary endpoint, disease-free interval, in the five evaluation cohorts (C-index: 0.71 [0.68-0.75], HR: 3.63 [3.02-4.37, p<0.001]). In a direct comparison (n=858), the AI test was more accurate than Oncotype DX, the standard-of-care 21-gene assay, achieving a C-index of 0.67 [0.61-0.74] versus 0.61 [0.49-0.73], respectively. Additionally, the AI test added independent prognostic information to Oncotype DX in a multivariate analysis (HR: 3.11 [1.91-5.09, p<0.001)]). The test demonstrated robust accuracy across major molecular breast cancer subtypes, including TNBC (C-index: 0.71 [0.62-0.81], HR: 3.81 [2.35-6.17, p=0.02]), where no diagnostic tools are currently recommended by clinical guidelines. These results suggest that our AI test improves upon the accuracy of existing prognostic tests, while being applicable to a wider range of patients.

cs.AI↗

arXiv2024

Active learning for efficient discovery of optimal gene combinations in the combinatorial perturbation space

The advancement of novel combinatorial CRISPR screening technologies enables the identification of synergistic gene combinations on a large scale. This is crucial for developing novel and effective combination therapies, but the combinatorial space makes exhaustive experimentation infeasible. We introduce NAIAD, an active learning framework that efficiently discovers optimal gene pairs capable of driving cells toward desired cellular phenotypes. NAIAD leverages single-gene perturbation effects and adaptive gene embeddings that scale with the training data size, mitigating overfitting in small-sample learning while capturing complex gene interactions as more data is collected. Evaluated on four CRISPR combinatorial perturbation datasets totaling over 350,000 genetic interactions, NAIAD, trained on small datasets, outperforms existing models by up to 40\% relative to the second-best. NAIAD's recommendation system prioritizes gene pairs with the maximum predicted effects, resulting in the highest marginal gain in each AI-experiment round and accelerating discovery with fewer CRISPR experimental iterations. Our NAIAD framework (https://github.com/NeptuneBio/NAIAD) improves the identification of novel, effective gene combinations, enabling more efficient CRISPR library design and offering promising applications in genomics research and therapeutic development.

q-bio.GN↗

arXiv2024

An Analysis of Deep Learning Parameterizations for Ocean Subgrid Eddy Forcing

Due to computational constraints, climate simulations cannot resolve a range of small-scale physical processes, which have a significant impact on the large-scale evolution of the climate system. Parameterization is an approach to capture the effect of these processes, without resolving them explicitly. In recent years, data-driven parameterizations based on convolutional neural networks have obtained promising results. In this work, we provide an in-depth analysis of these parameterizations developed using data from ocean simulations. The parametrizations account for the effect of mesoscale eddies toward improving simulations of momentum, heat, and mass exchange in the ocean. Our results provide several insights into the properties of data-driven parameterizations based on neural networks. First, their performance can be substantially improved by increasing the geographic extent of the training data. Second, they learn nonlinear structure, since they are able to outperform a linear baseline. Third, they generalize robustly across different CO2 forcings, but not necessarily across different ocean depths. Fourth, they exploit a relatively small region of their input to generate their output. Our results will guide the further development of ocean mesoscale eddy parameterizations, and multiscale modeling more generally.

physics.ao-ph↗

arXiv2024

A Monte Carlo Framework for Calibrated Uncertainty Estimation in Sequence Prediction

Probabilistic prediction of sequences from images and other high-dimensional data is a key challenge, particularly in risk-sensitive applications. In these settings, it is often desirable to quantify the uncertainty associated with the prediction (instead of just determining the most likely sequence, as in language modeling). In this paper, we propose a Monte Carlo framework to estimate probabilities and confidence intervals associated with the distribution of a discrete sequence. Our framework uses a Monte Carlo simulator, implemented as an autoregressively trained neural network, to sample sequences conditioned on an image input. We then use these samples to estimate the probabilities and confidence intervals. Experiments on synthetic and real data show that the framework produces accurate discriminative predictions, but can suffer from miscalibration. In order to address this shortcoming, we propose a time-dependent regularization method, which is shown to produce calibrated predictions.

cs.LG↗

arXiv2024

A stable implementation of a data-driven scale-aware mesoscale parameterization

Ocean mesoscale eddies are often poorly represented in climate models, and therefore, their effects on the large scale circulation must be parameterized. Traditional parameterizations, which represent the bulk effect of the unresolved eddies, can be improved with new subgrid models learned directly from data. Zanna and Bolton 2020 (ZB20) applied an equation-discovery algorithm to reveal an interpretable expression parameterizing the subgrid momentum fluxes by mesoscale eddies through the components of the velocity-gradient tensor. In this work, we implement the ZB20 parameterization into the primitive-equation GFDL MOM6 ocean model and test it in two idealized configurations with significantly different dynamical regimes and topography. The original parameterization was found to generate excessive numerical noise near the grid scale. We propose two filtering approaches to avoid the numerical issues and additionally enhance the strength of large-scale energy backscatter. The filtered ZB20 parameterizations led to improved climatological mean state and energy distributions, compared to the current state-of-the-art energy backscatter parameterizations. The filtered ZB20 parameterizations are scale-aware and, consequently, can be used with a single value of the non-dimensional scaling coefficient for a range of resolutions. The successful application of the filtered ZB20 parameterizations to parameterize mesoscale eddies in two idealized configurations offers a promising opportunity to reduce long-standing biases in global ocean simulations in future studies.

physics.ao-ph↗

arXiv2024

Transfer Learning for Emulating Ocean Climate Variability across $CO_2$ forcing

With the success of machine learning (ML) applied to climate reaching further every day, emulators have begun to show promise not only for weather but for multi-year time scales in the atmosphere. Similar work for the ocean remains nascent, with state-of-the-art limited to models running for shorter time scales or only for regions of the globe. In this work, we demonstrate high-skill global emulation for surface ocean fields over 5-8 years of model rollout, accurately representing modes of variability for two different ML architectures (ConvNext and Transformers). In addition, we address the outstanding question of generalization, an essential consideration if the end-use of emulation is to model warming scenarios outside of the model training data. We show that 1) generalization is not an intrinsic feature of a data-driven emulator, 2) fine-tuning the emulator on only small amounts of additional data from a distribution similar to the test set can enable the emulator to perform well in a warmed climate, and 3) the forced emulators are robust to noise in the forcing.

physics.ao-ph↗