In agriculture, explainable deep neural networks (DNNs) can be used to pinpoint the discriminative part of weeds for an imagery classification task, albeit at a low resolution, to control the weed population. This paper proposes the use of a multi-layer attention procedure based on a transformer combined with a fusion rule to present an interpretation of the DNN decision through a high-resolution attention map. The fusion rule is a weighted average method that is used to combine attention maps from different layers based on saliency. Attention maps with an explanation for why a weed is or is not classified as a certain class help agronomists to shape the high-resolution weed identification keys (WIK) that the model perceives. The model is trained and evaluated on two agricultural datasets that contain plants grown under different conditions: the Plant Seedlings Dataset (PSD) and the Open Plant Phenotyping Dataset (OPPD). The model represents attention maps with highlighted requirements and information about misclassification to enable cross-dataset evaluations. State-of-the-art comparisons represent classification developments after applying attention maps. Average accuracies of 95.42% and 96% are gained for the negative and positive explanations of the PSD test sets, respectively. In OPPD evaluations, accuracies of 97.78% and 97.83% are obtained for negative and positive explanations, respectively. The visual comparison between attention maps also shows high-resolution information.

Reliable detection of white matter hyperintensities (WMH) is crucial for studying the impact of diffuse white-matter pathology on brain health and monitoring changes in WMH load over time. However, manual annotation of 3D high-dimensional neuroimages is laborious and can be prone to biases and errors in the annotation procedure. In this study, we evaluate the performance of deep learning (DL) segmentation tools and propose a novel volumetric segmentation model incorporating self-attention via a transformer-based architecture. Ultimately, we aim to evaluate diverse factors that influence WMH segmentation, aiming for a comprehensive analysis of the state-of-the-art algorithms in a broader context. We trained state-of-the-art DL algorithms, and incorporated advanced attention mechanisms, using structural fluid-attenuated inversion recovery (FLAIR) image acquisitions. The anatomical MRI data utilized for model training was obtained from healthy individuals aged 62-70 years in the Live active Successful Aging (LISA) project. Given the potential sparsity of lesion volume among healthy aging individuals, we explored the impact of incorporating a weighted loss function and ensemble models. To assess the generalizability of the studied DL models, we applied the trained algorithm to an independent subset of data sourced from the MICCAI WMH challenge (MWSC). Notably, this subset had vastly different acquisition parameters compared to the LISA dataset used for training. Consistently, DL approaches exhibited commendable segmentation performance, achieving the level of inter-rater agreement comparable to expert performance, ensuring superior quality segmentation outcomes. On the out of sample dataset, the ensemble models exhibited the most outstanding performance. DL methods generally surpassed conventional approaches in our study. While all DL methods performed comparably, incorporating attention mechanisms could prove advantageous in future applications with a wider availability of training data. As expected, our experiments indicate that the use of ensemble-based models enables the superior generalization in out-of-distribution settings. We believe that introducing DL methods in the WHM annotation workflow in heathy aging cohorts is promising, not only for reducing the annotation time required, but also for eventually improving accuracy and robustness via incorporating the automatic segmentations in the evaluation procedure.

The automated quantification of complex animal behavior is fundamental to neuroscience and pharmacology, yet converting high-dimensional pose data into reproducible and interpretable behavioral measures remains challenging. Here, we present GESTURE, an unsupervised, graph-based deep generative framework that discovers and quantifies behavioral structure from pose dynamics without genotype labels or manual behavioral annotations. Applied to mice with motor dysfunction and wild-type controls, GESTURE identified a shared vocabulary of behavioral motifs. Differential use of these motifs yielded distinct “behavioral fingerprints” that reliably separated genotypes. By modeling behavior as a sequence rather than a static partition, GESTURE also quantified its temporal organization. Affected animals maintained motifs for longer and transitioned between them more predictably, indicating a slowing and stereotyping of behavioral sequences rather than simply reduced activity. GESTURE’s graph-based representation enables training across multiple recordings and embeds behavior in a shared latent space, supporting consistent cross-animal comparisons and future cross-experiment alignment. Automatically derived measures of behavioral divergence tracked the temporal dynamics of expert-annotated disability scores, reaching agreement comparable to that of independent human raters. Node- and edge-level explainability analyses further indicated that the latent representations emphasize the animal’s core motor scaffold in a biologically plausible manner. These findings support GESTURE as an interpretable, scalable framework for automated behavioral phenotyping that links genetic perturbation to quantitative behavioral phenotypes. Author summary: How can behavioral changes caused by a disease-associated mutation be measured automatically and consistently? Manual assessment is slow, difficult to scale, and can vary among observers. We studied mice carrying a mutation in a calcium-channel gene whose human counterpart is associated with episodic ataxia and related movement disorders. We developed GESTURE, which learns recurring movement patterns from the tracked coordinates of a freely moving mouse without genotype labels or manual behavioral annotations. By representing the body as a set of connected landmarks, GESTURE tracks how their configuration changes over time and discovers a shared vocabulary of movements. Affected and healthy mice used this vocabulary differently. Each animal’s pattern of use formed a distinctive “behavioral fingerprint”, and these fingerprints correctly identified every animal carrying the mutation. The analysis also quantified temporal features beyond overall activity: affected mice held each movement for longer and transitioned between movements more predictably. An automatically derived severity score agreed with expert judgment as closely as two trained raters agreed with each other and yielded the same result on every run.