Understanding cellular responses to stimuli is crucial for biological discovery and drug development. Transcriptomics provides interpretable, gene-level insights, while microscopy imaging offers rich predictive features but is harder to interpret. Weakly paired datasets, where samples share biological states, enable multimodal learning but are scarce, limiting their utility for training and multimodal inference. We propose a framework to enhance transcriptomics by distilling knowledge from microscopy images. Using weakly paired data, our method aligns and binds modalities, enriching gene expression representations with morphological information. To address data scarcity, we introduce (1) Semi-Clipped, an adaptation of CLIP for cross-modal distillation using pretrained foundation models, achieving state-of-the-art results, and (2) PEA (Perturbation Embedding Augmentation), a novel augmentation technique that enhances transcriptomics data while preserving inherent biological information. These strategies improve the predictive power and retain the interpretability of transcriptomics, enabling rich unimodal representations for complex biological tasks.

Unlike color photography images, which are consistently encoded into RGB channels, biological images encompass various modalities, where the type of microscopy and the meaning of each channel varies with each experiment. Im-portantly, the number of channels can range from one to a dozen and their correlation is often comparatively much lower than RGB, as each of them brings specific information content. This aspect is largely overlooked by methods designed out of the bioimage field, and current solutions mostly focus on intra-channel spatial attention, often ignoring the relationship between channels, yet crucial in most biological applications. Importantly, the variable channel type and count prevent the projection of several experiments to a unified representation for large scale pre-training. In this study, we propose ChAda-ViT, a novel Channel Adaptive Vision Transformer architecture employing an Inter-Channel Attention mechanism on images with an arbitrary number, order and type of channels. We also introduce IDR-CellI 00k, a bioimage dataset with a rich set of 79 experi-ments covering 7 microscope modalities, with a multitude of channel types, and counts varying from 1 to 10 per exper-iment. Our architecture, trained in a self-supervised man-ner, outperforms existing approaches in several biologically relevant downstream tasks. Additionally, it can be used to bridge the gap for the first time between assays with differ-ent microscopes, channel numbers or types by embedding various image and experimental modalities into a unified biological image representation. The latter should facilitate interdisciplinary studies and pave the way for better adoption of deep learning in biological image-based analyses.

In this paper, we propose a style-based conditional video generative model. We introduce a novel temporal generator based on a set of learned sinusoidal bases. Our method learns dynamic representations of various actions that are independent of image content and can be transferred between different actors. Beyond the significant enhancement of video quality compared to prevalent methods, we demonstrate that the disentangled dynamic and content permit their independent manipulation, as well as temporal GAN-inversion to retrieve and transfer a video motion from one content or identity to another without further preprocessing such as landmark points.

BACKGROUND: Cell culture on printed micropatterns slides combined with automated fluorescent microscopy allows for extraction of tens of thousands of videos of small isolated growing cell clusters. The analysis of such large dataset in space and time is of great interest to the community in order to identify factors involved in cell growth, cell division or tissue formation by testing multiples conditions. However, cells growing on a micropattern tend to be tightly packed and to overlap with each other. Consequently, image analysis of those large dynamic datasets with no possible human intervention has proven impossible using state of the art automated cell detection methods. RESULTS: Here, we propose a fully automated image analysis approach to estimate the number, the location and the shape of each cell nucleus, in clusters at high throughput. The method is based on a robust fit of Gaussian mixture models with two and three components on each frame followed by an analysis over time of the fitting residual and two other relevant features. We use it to identify with high precision the very first frame containing three cells. This allows in our case to measure a cell division angle on each video and to construct division angle distributions for each tested condition. We demonstrate the accuracy of our method by validating it against manual annotation on about 4000 videos of cell clusters. CONCLUSIONS: The proposed approach enables the high throughput analysis of video sequences of isolated cell clusters obtained using micropatterns. It relies only on two parameters that can be set robustly as they reduce to the average cell size and intensity.

MOTIVATION: High-throughput screening (HTS) is an important method in drug discovery in which the activities of a large number of candidate chemicals or genetic materials are rapidly evaluated. Data are usually obtained by measurements on samples in microwell plates and are often subjected to artefacts that can bias the result selection. We report here a novel edge effect correction algorithm suitable for RNA interference (RNAi) screening, because its normalization does not rely on the entire dataset and takes into account the specificities of such a screening process. The proposed method is able to estimate the edge effects for each assay plate individually using the data from a single control column based on diffusion model, and thus targeting a specific but recurrent well-known HTS artefact. This method was first developed and validated using control plates and was then applied to the correction of experimental data generated during a genome-wide siRNA screen aimed at studying HIV-host interactions. The proposed algorithm was able to correct the edge effect biasing the control data and thus improve assay quality and, consequently, the hit-selection step.

An a posteriori tracing method for broken neurite reconstruction is proposed. It uses local line structure cues from three vector fields; diffusion tensor vector field, Hessian eigenvector field, and diffused gradient vector field. The vector fields are built to give directionalities of line structures with gaps in different configurations. The method uses an iterative search and extend approach to reconstruct possible links between broken neurite segments which for various reasons existing tracing methods fail to extract. Iterative nature of the method enables us to extend traced neurites in directions aligned with underlying line structures. Subsequently, the method connects traced neurites across gaps in presence of other traced neurite segments or disposes extensions otherwise.

A discretized parametric curve can be seen as a sparse graph of vectors where each vertex is linked to two other vertices. Following this observation, we propose to generalize parametric active contours to a larger framework we call active vector graphs. This can be achieved by allowing each vertex of a graph of vectors to be linked to more than two vertices. An active graph does not need to be parameterized and the computation of its energy can be achieved by integrating over all its vertices. The optimization scheme pushes the graph toward the edges and in the direction of the normal which we show can be defined for all vertices. This offers a regularized model which addresses in an elegant and very fast way a certain set of problems such as the segmentation of connected regions. The method is described along with an example.

This paper describes a method to detect fluorescent PSF-like spots in 3D line-scanning confocal microscopy. We suppose that fluorescent spots are small enough to be characterized by the Point Spread function (PSF) of the microscope. To localize spots, we use a derivative expression of the image previously convolved by a Gaussian approximating the PSF. The algorithm validation is made on 3D synthetic images with fluorescent background structures and on a 3D fluorescent biological image.

The automated analysis of images is an active field of research in image processing and pattern recognition. In many applications, the first issue is to face illuminations artifacts that can appear due to bad imaging conditions. These artifacts often have direct consequences on the efficiency of the image analysis algorithms but also on the quantitative measures. This paper presents a fully automated nonuniformity correction based on empirical mode decomposition. The performances are outlined using both synthetic and real data.

We propose a method to detect and track multiple moving biological spot-like particles showing different kinds of dynamics in image sequences acquired through multidimensional fluorescence microscopy. It enables the extraction and analysis of information such as number, position, speed, movement, and diffusion phases of, e.g., endosomal particles. The method consists of several stages. After a detection stage performed by a three-dimensional (3-D) undecimated wavelet transform, we compute, for each detected spot, several predictions of its future state in the next frame. This is accomplished thanks to an interacting multiple model (IMM) algorithm which includes several models corresponding to different biologically realistic movement types. Tracks are constructed, thereafter, by a data association algorithm based on the maximization of the likelihood of each IMM. The last stage consists of updating the IMM filters in order to compute final estimations for the present image and to improve predictions for the next image. The performances of the method are validated on synthetic image data and used to characterize the 3-D movement of endocytic vesicles containing quantum dots.

Bayesian target tracking methods consist of filtering successive measurements coming from a detector. Linear and nonlinear Gaussian Bayesian filters are well adapted to estimate the successive a posteriori state distributions of a single moving target from a sequence of observations. However, when tracking several targets in a cluttered environment, previous techniques must be combined with dedicated procedures for validating and associating the measurements to their predictions. Gating validation techniques are used to increase the reliability of the association technique by retaining only the measurements that could be originated from predicted measurements. In standard techniques, the only constraint imposed on the gate is to contain the correct measurement. However, as the shape of the validation gate is related to the covariance of the transition noise, it is of major importance to estimate it in a reliable manner. We therefore review several methods to update the covariance of transition noise and we propose a new one that enables the validation gate to be adapted both to the smoothly evolving dynamic of a moving target and to an abruptly changing dynamic. All the methods are compared for performance on microscopy image sequences which typically contain objects that abruptly change their behavior.

A method which allows for the first time to perform automatically the detection and the tracking of microscopic objects directly from three dimensional image data is presented. It enables to analyse biological moving objects in three dimensional fluorescence image sequences coming from biological immunomicroscopy experiments, and get quantitative data such as the number of objects, their position, movement phases and speed. After a detection step is performed through the multiscale analysis of images using a shift-invariant wavelet transform, the tracking is achieved using a Kalman filter and an association which enable the position of the moving objects to be predicted, refined and updated. Trajectories are analysed in terms of different parameters relevant for the motility analysis of biological objects.