Optimizing absorption, distribution, metabolism, and excretion (ADME) is an important part of small molecule drug discovery. Many machine learning models have been built to predict ADME properties to facilitate this optimization process, but explaining model predictions is challenging. We propose a new graph neural network architecture with built-in meaningful per-atom attributions. Our model MolLedger outputs predictions that are the sum of per-atom scores. MolLedger's additive framework obtains exact interpretability at no cost to performance because the global context vector gives the additive head enough context to produce good per-atom scores. Furthermore, MolLedger produces attributions that are more faithful to chemical properties than other interpretability methods because the auxiliary loss in MolLedger anchors the atom scores to chemical properties. Our case studies comparing interpretations from multiple methods on molecular pairs reveal that MolLedger is much better at producing sensible explanations for predicted property changes.
Abdullah Al Mamun, Md. Nasif Osman Khansur, Md Ashraful Hossen Akash +2cs.CV
Deep-learning models can achieve strong chest X-ray (CXR) classification performance without establishing whether their predictions predominantly rely on pulmonary image content. This study evaluates pulmonary attribution containment as an anatomy-related reliability property distinct from diagnostic performance. We propose DBCA-SegNet-MGAP, a multi-task anatomy-guided CNN-Transformer framework that combines complementary feature representations through bidirectional cross-backbone attention, predicts a soft lung mask, and incorporates this anatomical prior directly into classification through Mask-Guided Adaptive Global Average Pooling (MGAP). Pulmonary attribution containment is quantified using the Anatomical Local Energy Ratio (ALR) and high-intensity cumulative ALR (cALR@0.9). Experiments were repeated across three training seeds using the COVID-19 Radiography Database for four-class internal testing and a locked Shenzhen-to-Montgomery protocol for zero-shot external tuberculosis testing. On COVID-19, the proposed model achieved a weighted F1 of $0.9615 \pm 0.0015$ and macro ROC-AUC of $0.9906 \pm 0.0007$. In an architecture-matched dual-bridge comparison, replacing conventional GAP with MGAP increased ALR from $0.3878 \pm 0.0098$ to $0.7086 \pm 0.0104$ and cALR@0.9 from $0.5265 \pm 0.0101$ to $0.9905 \pm 0.0018$, while weighted F1 remained essentially unchanged ($0.9618 \pm 0.0015$ vs. $0.9615 \pm 0.0015$). Under locked external transfer to Montgomery, ROC-AUC remained $0.9080 \pm 0.0043$ and pulmonary ALR remained $0.6466 \pm 0.0081$, whereas weighted F1 decreased to $0.7528 \pm 0.0080$ and ECE increased to $0.1683 \pm 0.0055$. These findings show that diagnostic discrimination, calibration, and pulmonary attribution containment are distinct model properties and support their joint evaluation under internal testing and external domain shift.
Electronic Health Record (EHR) prediction models in the intensive care unit must learn from sparse and irregular measurements while preserving the clinical meaning of time and supporting transparent decision-making. We present INTERVenE, a family of Transformer architectures whose input is an interval-based, knowledge-based temporal abstraction (KBTA), a token stream of named clinical concepts (states, trends, events, contexts) drawn from a curated medical ontology, rather than an unnamed bin index or a raw measurement triplet. This naming layer is what we ask KBTA to do: it makes the model's per-token attributions resolve to clinical concepts by construction. INTERVenE offers two complementary variants: an auto-regressive decoder that generates future abstraction trajectories with a per-step risk readout (localizing \emph{when} and \emph{after which events} risk rises), and a bidirectional encoder for single-pass joint risk and time-to-event prediction. Evaluated on 57,078 MIMIC-IV admissions against GRU-D, STraTS, and KarmaLego, INTERVenE-Enc reaches a support-weighted AUPRC$_w$ of 0.672, improving by 0.041 over the strongest neural baseline with non-overlapping 95\% bootstrap CIs, while also taking the best AUROC$_w$ (0.901) and length-of-stay MAE (44.4\,h). INTERVenE-Ar (AUROC$_w$ $0.854$, AUPRC$_w$ $0.587$ under the same evaluation contract - a strictly harder generative readout) provides a complementary token-level risk trajectory. An input-representation ablation confirms the lift transfers across structured discretizations, positioning KBTA-based intervals as the interpretable substrate that makes per-token attributions resolve to meaningful clinical concepts within the deployed model.
Deep learning has a powerful capability of feature extraction. However, the lack of fairness and interpretability in deep neural networks poses limitations to their adoption in the medical domain. This paper proposes a disentangled representation learning (DisenRL) framework, named the Attributes-based Gaussian Estimation for Disentangled Representation (AGEDR), which incorporates Attribute Mapping Embedding (AME) modules designed to map attributes into vectors and align them with a subset of the latent vectors in a Variational AutoEncoder (VAE). This part of the latent vector will be disentangled from the remaining latent vectors by minimizing mutual information. A classifier is then trained using the mean parameters of the latent vectors from the VAE. Extensive experiments demonstrate that AGEDR outperforms both conventional classification models and existing disentangled representation learning methods. The ablation experiments also indicate the disentangling capability and fairness of AGEDR. The source code is publicly available at https://github.com/ZhaoKe1024/DisentangledRepr.
Sparse autoencoders (SAEs), and other interpretability methods could turn AI models in Biology and other fields into engines of scientific discovery by explaining the superhuman capabilities of those models. However, a latent is only useful if we know three things: whether it is coherent, whether it can be described, and whether that description has predictive power. These questions are routinely conflated. We assemble them into a single pipeline and report the practical innovations each stage required. First, cross-seed dictionary stability prioritises which latents are worth spending resources to investigate. Second, an intruder-detection task asks whether a latents activating examples share a recognizable pattern. Third, a separate pass proposes a candidate biological description which we convert into falsifiable predictions which can be tested in silico. Deployed on the Boltz-1 Pairformer trunk, stability prioritisation finds interpretable latents using about 4.4 times fewer latent evaluations each, and at 5.2 times lower measured cost, while recovering over half of them, and the external check shows the surfaced motifs are significantly enriched for their claimed annotations. The results also suggest a possible tension: the cross- seed stability might be selecting for some types of features, like structure-related ones, much more than others, such as function-related features.
Clinical language models can achieve strong in-hospital accuracy yet fail under deployment shifts because they exploit note-specific artifacts (e.g., templates, separators, boilerplate) that do not reflect patient state. We propose CAST (Concept-guided Artifact Suppression Tuning), an SAE-based framework for auditable clinical text classification. CAST uses Sparse Autoencoders to expose sparse, human-auditable features from intermediate Transformer activations, labels SAE latents with an LLM-assisted interpretation pipeline and ICD-10 retrieval constraints, suppresses verified artifact latents via residual subtraction during fine-tuning, and provides post-hoc per-concept attributions for auditing model decisions. On MIMIC-IV discharge-note mortality prediction, CAST improves over its corresponding fine-tuned encoder baselines and remains competitive with strong LLM baselines, while producing a feature-level audit trail of the clinical concepts that support each prediction and the artifact concepts suppressed during training.
Although Multimodal Large Language Models (MLLMs) have demonstrated impressive performance in Medical Visual Question Answering (Med-VQA), their reliance on global image features often lacks precise pixel-level grounding, thereby limiting clinical trustworthiness. To bridge the semantic gap between high-level clinical reasoning and spatial localization, we propose \textsc{\textsc{MedREAL}} (\textbf{Med}ical \textbf{RE}asoning-driven \textbf{A}nswering and \textbf{L}ocalization), a unified framework that seamlessly aligns linguistic reasoning with spatial grounding. Specifically, \textsc{MedREAL} introduces \textbf{S}eg \textbf{A}nchored \textbf{R}easoning \textbf{P}ooling (SARP) to distill task-relevant semantic evidence directly from \texttt{[SEG]} tokens within the MLLM's hidden states. Furthermore, a \textbf{R}easoning-to-\textbf{V}isual (R2V) fusion mechanism is proposed to effectively inject these reasoning-aware features into a segmentation pipeline for accurate mask decoding. To facilitate this paradigm, we construct MedRAVS-13K, a comprehensive dataset comprising 13,824 expertly validated samples across four diverse imaging modalities. Extensive experiments demonstrate that \textsc{MedREAL} significantly outperforms state-of-the-arts, achieving 68.49\% gIoU and 70.47\% cIoU on benchmark evaluations. By generating evidence masks that are strictly consistent with textual diagnoses, \textsc{MedREAL} provides a robust, interpretable framework for reasoning-driven medical image analysis.
Protein language models (pLMs) encode information about protein sequences which enable downstream tasks such as structure prediction, but their internal representations are not well understood. Sparse autoencoders (SAEs) provide a promising tool to disentangle latent pLM representations into interpretable features, but existing annotation pipelines largely rely on protein-level annotations derived from database labels and LLM annotations of top activating sequences. Such annotations can overlook the localized residue-level and geometric patterns encoded by sparse features. We introduce an automated and scalable method for interpreting SAE features in ESM-2 by using geometrically inspired features of the protein $\text{C}_α$ backbone. Across ESM-2 8M layers, an FDR-controlled discovery analysis shows that local geometry is significantly associated with many SAE features, with varying levels of predictive strength, expanding coverage beyond database and sequence-based methods. In particular, geometry can distinguish SAE features sharing the same database annotation, revealing substructure within known biological labels. A significant portion of SAE features activate on unannotated metagenomic protein sequences enabling us to use our SAE annotations to better understand these sequences. In addition, ablation experiments at the level of contact prediction show that removing found geometric features shifts ESM-2's predicted contact maps in the direction of the descriptor. This provides a robust method of annotating proteins activated within SAE neurons at a residue level, providing a bridge between mechanistic interpretability and structural biology.
Spoken-language analysis via prompt-based domain-adaptive models is a promising direction for low-resource, non-invasive dementia screening, but such models remain internally opaque. We study the interpretability of the Domain-Adapted models via Prompt-based Fine-tuning (DAPF) framework, which casts dementia detection as diagnosis-related masked-token prediction. We interpret DAPF and strong baselines using a variety of probing and analysis techniques, finding that DAPF achieved the best overall performance (accuracy=0.83 and macro-F1=0.83) with diagnosis most recoverable from its [MASK] representation. However, this representational advantage did not extend to token-level explanation faithfulness. DAPF attributions primarily reflected language task vocabulary, discourse markers, and transcription artifacts, with perturbation tests showing weak or negative effects. This suggests that its masked-token interface determines diagnosis information without producing faithful token-level explanations.
Alexis Ivan Escamilla-Lopez, Gilberto Ochoa-Ruiz, Salvador Hinojosa +1cs.CV
The interpretation of endoscopic imagery in ulcerative colitis is complex and subjective, with variability in human assessment and subtle mucosal inflammation. Although deep learning has advanced automated analysis, most vision-language models rely on global visual embeddings that overlook the localized and relational nature of pathological evidence, limiting clinical reliability and interpretability. We introduce LUX (Lesion-aware Unified eXplainable captioning), a graph-conditioned vision-language architecture for explainable endoscopic image captioning. LUX constructs a lesion-centric scene graph from Grad-CAM and CBAM activation maps, representing pathological regions as nodes and encoding their spatial and clinical relationships. These graph embeddings are integrated into the cross-attention layers of a T5 decoder, enabling generated words to attend to specific lesion nodes rather than only to global image features. This provides direct alignment between linguistic content and pathological evidence, supporting token-level interpretability and relational reasoning. LUX outperforms strong baseline and state-of-the-art medical captioning models across BLEU, METEOR, ROUGE-L, and CIDEr, with particularly strong gains in CIDEr. It also reduces hallucinated clinical findings and improves lesion-level grounding through stronger correspondence between generated tokens and localized pathological regions.
Dawa Chyophel Lepcha, Aaliya Ali, Sophie A. Martin +3eess.IV cs.AI cs.CV cs.LG
Multimodal neuroimaging combining structural MRI and positron emission tomography (PET) captures complementary structure-function relationships across the Alzheimer's disease (AD) continuum, yet existing artificial intelligence systems produce a single diagnostic label without quantifying which imaging modality drove that decision for a specific patient. We introduce the Modality Contribution Network (MCNet) and the Modality Contribution Score (MCS), the first per-patient attribution framework quantifying the shift in modality dominance from structural atrophy to amyloid and metabolic dysfunction across the cognitively normal to MCI to AD continuum. MCS is normalised to unity per subject via modality ablation (MCS_MRI_i + MCS_PET_i = 1.0 for every subject i), providing an interpretable, clinically actionable score that fluid biomarkers cannot supply. Applied to 327 ADNI-3 participants balanced across cognitively normal, mild cognitive impairment, and AD groups, MCNet achieved competitive three-class staging performance (AUC=0.881). The MCS revealed a statistically significant monotonic gradient (Kruskal-Wallis p<0.0001), with increasing PET dominance from cognitively normal (MCS_PET 0.412+/-0.229) through MCI (0.489+/-0.289) to AD (0.671+/-0.426), validated against amyloid SUVR (r=0.172, p=0.006) and FDG metabolic biomarkers (r=-0.287, p=0.0005) from separate imaging pipelines. External replication in 1,073 independent OASIS-3 subjects confirmed cross-cohort generalisability (H=166.99, p<0.0001, eta^2=0.156). A mechanistic comparison with SHAP demonstrated that ablation-based MCS captures clinically meaningful modality dependence that deviation-based methods cannot. These findings position MCNet as a foundation for personalised imaging decisions, clinical trial stratification, and trustworthy AI in dementia care.
Oral cancer is a leading cause of mortality in low-to-middle-income countries, where a shortage of specialists delays diagnosis. While point-of-care screening via smartphones offers a scalable solution, developing robust AI for resource-constrained settings poses significant challenges, including class imbalance in training data, variable data quality, and computational constraints on edge devices. In this paper, we present the optimisation of lightweight deep learning models for smartphone-based oral cancer screening. Using a diverse, multi-centre retrospective dataset of approximately 30,000 images acquired over a decade, we systematically evaluate state-of-the-art convolutional, transformer, and hybrid architectures. Through rigorous pipeline ablation, we demonstrate that directly optimising hybrid architectures for the edge strictly outperforms computationally heavy paradigms, such as large models or knowledge distillation. Furthermore, interpretability analysis and simulated noise-stress tests revealed that the system anchors on clinical features and remains robust to unstructured sensor noise, despite vulnerabilities to impulse bit errors. In the held-out test set, our optimised MobileViTv2 models achieved an average sensitivity of 83.2 $\pm$ 1.5% and an average specificity of 86.0 $\pm$ 0.8%, with the best model exhibiting 87.4% sensitivity, 86.5% specificity, and a critical negative predictive value of 97.2% with reference to specialist labels. These results confirm that with targeted architectural selection and streamlined optimisation, interpretable and robust lightweight AI models exhibit high potential for edge deployment to enable automated triage in primary care settings.
Multimodal clinical AI is limited by weakly aligned inputs and the absence of domain-specific interpretable representations, particularly when learning from dense video stream, structured time-series, and template-based kinematic text. Here we present ScoliDetect, an explainable framework for adolescent idiopathic scoliosis screening from monocular gait video, built around a kinematic knowledge map (KKM) and complementary template-based kinematic text derived from per-sequence pose statics. KKM is a fixed-index structured representation that encodes gait features across absolute motion, self-skeleton configuration and joint-joint signal correlation, providing anchor-referenced multimodal fusion and factor-level interpretation. We integrate video, KKM, and template-based kinematic text through bidirectional cross-attention with latent-bottleneck aggregation. In a multicenter cohort (n = 1,858 after exclusions), prespecified supervised ablations on an external screening cohort show that KKM-mediated multimodal fusion outperforms unimodal models and late concatenation. Under a staged training protocol, trimodal contrastive pretraining is applied after architecture selection as representation initialization, improving external ROC-AUC from 0.961 to 0.972. Furthermore, the structured nature of the KKM provides inherent, factor-level attributions mapped directly to specific kinematic phases and skeletal indices, offering verifiable interpretability. The results demonstrate that embedding explicit structural topologies into latent spaces significantly enhances both the generalization and explainability of multimodal pattern analysis systems.
Breast ultrasound diagnosis relies on clinically meaningful semantic concepts, yet most deep learning methods adopt end-to-end image-to-label paradigms that lack interpretability and robustness. While concept-based approaches offer a promising alternative, they often assume complete annotations or require multimodal inputs at inference, which significantly limits their real-world applicability. To tackle these issues, we propose Training-time Report-guided and Clinically Ordered Concept Editing (TRACE), a training-time report-guided framework that leverages structured radiology reports as privileged concept supervision while enabling image-only diagnosis at test time. TRACE refines image-derived concepts through a teacher-guided editing mechanism within a malignancy-aware ordered concept space. To address incomplete annotations, we introduce Strategic Concept Missing Training (SCMT) and train an image-only self-editor via edit distillation for autonomous concept refinement. Besides, we introduce BUSC, a concept-enriched benchmark linking images, labels, and structured attributes. Experiments across multiple datasets demonstrate that TRACE achieves superior performance and improved cross-domain robustness compared to existing methods.
Specialist language models are usually understood through endpoint gains: the generalist scores lower, the specialist scores higher, and the difference is treated as evidence of specialization. This leaves the released update itself largely unexamined. We propose a paired weight-delta path audit and apply it to two public, aligned generalist-to-medical-specialist checkpoint pairs: Gemma-3-4B-IT to MedGemma-4B-IT and Qwen2.5-7B-Instruct to HuatuoGPT-o1-7B. In both pairs, the full decoder-side update strongly reconstructs measured medical benchmark movement (0.974 and 1.183 endpoint-normalized retention), making each decoder delta an appropriate substrate for the audit. Yet the movement is not cleanly localized. MLP is the strongest broad component family in both pairs, but mixed off-domain movements, 10-seed matched controls, and endpoint-anchored rollbacks prevent a unique coarse-family explanation. The audit therefore separates update-level reconstruction from component-level explanation. Its claims concern text-only multiple-choice benchmark movement, not clinical validation, repair, or circuit-level mechanism.
Predictive models over structured electronic health records (EHRs) remain central to machine learning for healthcare, but few have jointly emphasized quantitative laboratory information and interpretability with respect to input medical events. We present BERT-LER, a BERT-style model for coded EHR timelines pretrained and fine-tuned from a de-identified EHR dataset of 75 million patients, that encodes laboratory test results as discrete tokens while retaining graded information through percentile-based binning, paired with Integrated Gradients for token-level attributions grounded in the input EHR sequence. We benchmark our approach on the public EHRShot benchmark suite and on an asthma severity progression study based on real-world data. This addresses a methodological gap in EHR foundation-style modeling by unifying laboratory value representation and explainability in a single framework, while assessing whether both predictive performance and explanations generalize beyond standard clinical prediction tasks. Across EHRShot and asthma tasks, BERT-LER achieves predictive performance that is competitive with, and on laboratory-related tasks often exceeds, publicly available benchmark models, and provides attributions that align with clinically known risk factors. Our architecture and explainability approach can be applied to many therapeutic areas and prediction tasks using language models trained on structured EHRs.
Accurate classification of circulating tumor cell (CTC) phenotypes can provide valuable information for assessing metastatic potential. Label free microfluidic devices provide a hydrodynamic obstacle course that transforms subtle biophysical characteristics of CTCs, including size and deformability, into distinct kinematic trajectories. However, the highly nonlinear fluid structure interactions governing these trajectories make the inverse problem of inferring cellular phenotype from trajectory data analytically intractable. While deep neural networks (DNNs) have emerged as a powerful approach for addressing this inverse problem, their effectiveness is constrained by the limited availability of trajectory data and the lack of physical interpretability. To address these challenges, we propose an interpretable and data efficient DNN framework for trajectory based CTC classification. To mitigate the scarcity of data, we develop Subsequence (SubSeq), a targeted augmentation strategy that randomly extracts informative local trajectory segments during training to promote learning from localized patterns. We further apply Gradient Weighted Class Activation Mapping to identify the trajectory features and physical regions of the microfluidic device that drive model predictions. Experimental results demonstrate that SubSeq improves classification accuracy over the evaluated baseline and augmentation methods. Furthermore, interpretability analysis suggests that localized trajectory segments contain substantial biophysical information relevant to accurate classification. This provides justification for SubSeq and also highlights the redundancy of full-length trajectories. More broadly, the proposed framework views microfluidic geometries as physical encoders of cellular mechanical properties, providing mechanistic insights that may inform the future design of diagnostic devices.
A radiologist reading a model's output faces two problems. The model returns a number and no reason, and any system that turns that number into readable prose can quietly add claims the model never made. MIRROR is a research prototype built to separate those failures. It chains a multi-label classifier, a Grad-CAM localizer that turns each positive finding into a named anatomical region, and a report writer that receives the labels, probabilities, and regions but never the image. Because the language layer cannot see pixels, it cannot assert a finding the classifier did not make. We are precise about what that buys: a MIRROR report's findings are auditable against the probability vector, while the sentences framing them are ordinary generated text, and we show one stating a cardiothoracic ratio the system never measured. One registry holds the taxonomy, anatomy, and phrasing for chest X-ray, brain MRI, and head CT, so adding a modality is a data change; all three are routed and tested, one is trained. On ChestMNIST that classifier reaches macro AUROC 0.729 and ranks better than chance on all 14 labels, at 1.6 to 6.8 times the precision a random ranker would get. Yet at the default 0.5 threshold it emits no positive prediction at all for 11 of them, and its excellent-looking Brier score of 0.045 sits beside the 0.047 earned by a predictor that ignores the image. The discrimination is real; the decisions are not. Under the class imbalance normal in radiology, aggregate metrics flatter models that do nothing, and should be reported against that floor.
Functional connectome analysis examines brain-region interactions to understand and identify disorders such as autism spectrum disorder and Alzheimer's disease. Existing methods typically use GNNs and Transformers to model the full functional connectivity matrix. However, processing tens of thousands of connections introduces redundancy and noise, increases computational cost, and limits connection-level interpretability. This raises a central question: do we really need complex interaction modeling, or is identifying a small set of disease-relevant connectivity patterns sufficient? To answer this question, we propose BrainLinear, a lightweight geometry-aware framework for mining disease-discriminative connectome patterns. BrainLinear first maps each functional connectivity matrix to a shared tangent space centered at the Fréchet mean of the training set, capturing subject-specific deviations while respecting matrix geometry. It then scores each ROI-pair tangent direction by its classification contribution and disease--control difference, retaining Top-$K$ directions as a compact representation. Finally, a shallow multilayer perceptron performs classification on the selected representation. Experiments on ABIDE and ADNI show that BrainLinear matches or exceeds strong GNN and Transformer baselines at a fraction of their cost: it improves AUC and ACC over the best baseline for each metric by up to $3.54$ and $1.39$ percentage points, while reducing runtime and peak GPU memory by $84.0\%$ and $68.4\%$ relative to the closest baseline in AUC. The selected directions are directionally consistent with between-group displacements and organized across major functional systems, supporting connection-level interpretation.
Deeksha M Shama, Punnisa Amornsirikul, Archana Venkataramancs.LG
Objective: Foundation models represent the next advancement in AI for EEG analysis; however current explainable AI techniques provide attribution scores in the time-channel input space, which is mismatched to clinical intuition about EEG. Thus, there is a critical need for a universal method that can extend the interpretability of any foundation model to alternative and physiologically relevant domains without modifying or retraining the underlying model. Methods: EEG-PRISM leverages linear transformations and established backpropagation rules to map time-channel attribution scores into alternative domains. We derive mappings to the frequency domain via an invertible DFT and to the source domain via an approximately invertible EEG generative model. We evaluate EEG-PRISM in simulated and real data, assessing recovery of ground-truth phenomena across domains with five foundation models and four AI explainers. Results: In simulation, EEG-PRISM achieves near-perfect spectral recovery and 69.2% spatial accuracy. In epilepsy, EEG-PRISM correctly determines that delta-theta activity is most salient and correctly localizes the seizure onset region with 50% accuracy. In autism, EEG-PRISM localizes the predictive delta-alpha biomarkers to frontal and temporal regions, consistent with prior work. Conclusion: EEG-PRISM is a theoretically-grounded post-hoc attribution method with accurate mapping into the spectral and spatial domains. It supports window-level analysis of transient events (e.g., seizures) and group-level identification of clinically relevant biomarkers (e.g., autism), thus advancing interpretable EEG foundation models. Significance: This work enables physiologically-grounded interpretation of EEG foundation models and supports clinically relevant insights such as event localization and biomarker identification.
Saisha Shetty, Satvik Tripathi, Austin Lin +6cs.AI cs.CL
We present Multi-Agent Reasoning and Coordination (MARC), an open-source framework that replaces monolithic LLM prompting with deterministic multi-agent orchestration for clinical reasoning. MARC coordinates role-specialized agents for extraction, reasoning, answer generation, and evaluation, with explicit context passing and traceable intermediate outputs, enabling stage-wise failure attribution. We additionally introduce a Decomposer module that generates task-specific agent prompts from a plain-language description, eliminating manual prompt engineering. The framework supports both API-based and local CPU-compatible deployments and is entirely configurable via YAML, without code modifications. MARC is designed to be model-agnostic, interpretable, and accessible to clinical domain experts without programming expertise. The full framework is available at https://github.com/Penn-RAIL/MARC-v1.
Concept bottleneck models (CBMs) can improve the transparency of cancer image diagnostic prediction by expressing predictions through radiological concepts. However, their dependence on instance-level concept annotations limits practical applicability. We propose a prior-guided hybrid CBM that integrates limited concept annotations, class-conditional concept distribution matching on unannotated patients, and prior initialization of the concept-to-diagnosis head. We evaluate the method on CBIS-DDSM mammographic masses and calcifications and LIDC-IDRI pulmonary nodules across 0-100% concept annotation. In the clinically relevant 0-20% annotation regime, the hybrid CBM consistently improves mean concept AUC over a matched standard CBM, while maintaining diagnostic performance close to black-box models. At 10% annotation specifically, concept AUC increases from 0.619 to 0.741 for masses, from 0.650 to 0.787 for calcifications, and from 0.597 to 0.642 for pulmonary nodules. Ablation experiments identify prior initialization as the main component contributing to improved concept detection, likely by stabilizing the concept-to-diagnosis head. Zero-shot VLMs remain insufficient for reliable fine-grained tumor-level concept prediction. These findings suggest that structured priors can substantially reduce the annotation burden of interpretable cancer imaging models.
Piotr Jedryszek, Tongmeng Xie, Adam Winnifrith +5q-bio.QM cs.LG
AlphaFold3-class structure predictors pair a representational trunk, which processes sequence and context, with a diffusion module, which generates atomic coordinates. How biological information changes as it crosses this architectural boundary remains poorly understood. We analyze per-residue activations from the Pairformer trunk and diffusion module of Boltz-1 using linear probes, sparse autoencoders (SAEs), and causal interventions. From the trunk, both geometry (secondary structure, disorder) and sequence chemistry (amino-acid identity, signal peptides, disulfide-bond annotations) are linearly decodable. In the diffusion module, the two diverge. Secondary structure transfers essentially unchanged, whereas sequence chemistry is strongly attenuated. We then test whether decodable directions can steer the model, intervening on the final trunk single representation that conditions the diffusion module. Helix and coil directions change predicted structure dose-dependently against matched-norm random controls, but a beta-strand direction that is highly predictive (F1 =0.82) produces no measurable increase in strand content: linear decodability does not imply causal influence at the site we tested. The same probes also score markedly lower against sparse SwissProt annotations than against dense DSSP labels, because unannotated residues that the model gets right are charged as false positives; such scores are therefore lower bounds. Finally, supervised probes outscore single SAE features wherever a label already exists. We release the trained trunk and diffusion SAEs, Boltz-1 per-residue activations, and the analysis code.
Healthcare data, such as Intensive Care Unit (ICU) records, comprise heterogeneous multivariate time series sampled at irregular intervals with pervasive missingness. However, clinical applications demand predictive models that are both accurate and interpretable. We present our Graph Attention-based Relational Learning for Intensive Care (GARLIC) model, a novel neural network architecture that imputes missing data through a learnable exponential-decay encoder, captures inter-sensor dependencies via time-lagged summary graphs, and fuses global patterns with cross-dimensional sequential attention. All attention weights and graph edges are learned end-to-end to serve as built-in observation-, signal-, and edge-level explanations. To reconcile auxiliary reconstruction and primary classification objectives, we developed an alternating decoupled optimization scheme that stabilizes training. On three ICU benchmarks (PhysioNet 2012 & 2019, MIMIC-III), GARLIC sets the new state of the art in outcome prediction, significantly improving AUROC and AUPRC over best-performing baselines at comparable computational cost. Ablation studies confirm the contribution of each module, and feature-removal trials validate the fidelity of importance attribution through a monotonic performance drop (full > top 50% > random 50% > bottom 50%). Real-time case studies demonstrate actionable risk warnings with transparent explanations, marking a significant advance toward accurate, explainable deep learning for irregularly sampled ICU time series data. Moreover, we demonstrated \proposed{}'s superiority in data imputation and classification on various time-series datasets beyond the ICU domain, showing its generalizability and applicability to broader tasks.
Nagur Shareef Shaik, Jeongwoo Park, Yeong-Jin Kim +3cs.CV cs.LG eess.IV eess.SP
Retinal fundus images frequently exhibit multiple co-occurring pathologies, yet standard deep learning classifiers apply static, identical computation to every image regardless of the underlying disease distribution. We propose a novel architecture that resolves this via sparse conditional computation, pairing a Guided Context Gating (GCG) spatial attention front-end with a sparsely-routed Mixture-of-Experts (MoE) block operating over feature tokens. Crucially, this routing yields an interpretable, data-driven decomposition. Expert allocation is significantly disease-dependent (p < 0.001), with the healthy Normal state and morphologically distinct pathologies (e.g., ERM, AMD) isolating to dedicated experts. On a five-class, patient-disjoint 5-fold cross-validation benchmark, our model achieves 0.912 +/- 0.008 macro AUC and 0.653 +/- 0.014 macro F1. Furthermore, Grad-CAM++ and post-MoE t-SNE visualizations confirm that expert routing aligns with localized lesions and geometrically maps co-occurring cases between their constituent clusters, positioning sparse MoE as an interpretable approach to multi-disease retinal screening.
Madhumitha Venkatesh, Shanawaj S Madarkar, Konda Reddy Mopurics.CV
Alzheimer's disease (AD) progression is a longitudinal process with subtle pathological cues in the early stages. Yet, computational constraints have limited most neuroimaging models to either compromise spatial information or limit the number of longitudinal scans. We aim to overcome this bottleneck and fully leverage high-resolution, variable-length T1w structural MRI (4D sMRI) scan sequences. We introduce Parcel2Progression (P2P), a Longitudinal Transformer Framework which tackles this challenge using an Atlas-guided Parcel Encoder that tokenizes 3D scans into a set of richer anatomically grounded representations. A Longitudinal Transformer then integrates irregular, arbitrary-length longitudinal visits with patient age. This synergy delivers two key advantages: (1) parcel-specific interpretability, and (2) computational tractability for long-term analysis, which scales linearly with the number of scans compared to a naive quadratic 4D ViT cost. P2P outperforms prior works and baselines in both MCI (Mild Cognitive Impairment) to AD conversion prediction and AD vs. CN (Cognitively Normal) classification tasks across ADNI, AIBL, and MIRIAD datasets. Leveraging longitudinal scans boosts performance over single-scan baselines by up to 5% and 7% in balanced accuracy for AD classification and MCI conversion prediction tasks, respectively. Interpretability analysis using parcel saliencies and attention rollouts reveals clinically consistent atrophy patterns in AD and MCI subjects. We also demonstrate the frameworks' reliability in anomaly detection using a synthetic dataset, and test the model's generalizability for other neurodegenerative diseases like Frontotemporal Dementia.
Diabetic retinopathy (DR) remains a leading cause of vision loss among working-age adults worldwide, making reliable severity grading clinically important. Despite strong performance, most deep models formulate DR grading as image-level classification and do not explicitly model clinically grounded evidence, such as lesion types and spatial relations. In this paper, we propose PARAGraph, a Pathology-Anatomy-Aware Hierarchical Graph framework for DR grading. PARAGraph represents each image as a three-level hierarchical graph with lesion-level nodes, intermediate category and region nodes, and global anatomical and semantic nodes. To incorporate medical priors into nodes, we construct an optic disc-fovea-anchored coordinate frame that provides a scale- and rotation-normalized retinal reference system. Within this frame, lesion nodes are encoded with category, normalized area, and anatomical coordinates. To mitigate noisy lesion segmentation, PARAGraph uses a dual-fusion strategy that introduces global visual context into a graph semantic node and a decision-level prediction branch, improving robustness when lesion evidence is unreliable. Extensive experiments on Messidor-2, APTOS, and DDR show that PARAGraph achieves consistent DR grading performance over state-of-the-art methods. Interpretability and robustness analyses further demonstrate that its predictions are clinically grounded, closely associated with lesion evidence and robust to lesion segmentation noise.
Fakrul Islam Tushar, Stephen Adamo, Geoffrey D. Rubincs.CV cs.AI
Foundation models provide transferable CT representations, but predictions based directly on these embeddings are difficult to interpret. We developed concept bottleneck models that map two frozen CT foundation-model representations to eight radiologist-defined pulmonary-nodule attributes and predict malignancy from the estimated concepts and nodule size. The models included CT-FM, a whole-CT self-supervised encoder using a 96^3-voxel nodule-centered patch, and FMCIB, a nodule-focused contrastive encoder using a 50-mm crop. Eight ridge-regression concept heads were trained on 2,610 LIDC-IDRI nodules. Malignancy models were trained on LUNA25 and evaluated on a held-out internal test set and the external DLCS cohort. Concept fidelity was assessed using five-fold cross-validated R^2, and malignancy discrimination was assessed using AUROC with 95% confidence intervals estimated by patient-grouped bootstrap resampling. Concept fidelity was modest but higher for FMCIB than CT-FM for subtlety (R2, 0.24 vs. 0.11), spiculation (0.17 vs. 0.08), texture (0.17 vs. 0.07), and lobulation (0.15 vs. 0.05). Internally, the CT-FM and FMCIB concept+size models achieved AUROCs of 0.86 (95% CI, 0.80-0.92) and 0.86 (0.79-0.92), respectively. Externally, AUROCs were 0.72 (0.68-0.75) and 0.73 (0.70-0.76), compared with 0.73 for nodule size alone and 0.60 and 0.67 for the corresponding embedding only probes. Additive predictions could be decomposed into feature-level contributions and modified through controlled concept interventions. Concept bottlenecks provided transparent malignancy predictions with discrimination similar to nodule size alone, while differences in concept fidelity suggest that concept recovery depends on the underlying foundation-model representation.
In a feature-tokenized transformer (arXiv:2106.11959) such as BiomeGPT (doi:10.64898/2026.01.05.697599), each input token is built by fusing a fixed identity with a sample-specific measurement: a fixed species and a variable abundance, T = S + A. To interpret downstream classification in such models, prior work inspects the attention weights of the special [CLS] token (arXiv:2106.11959, arXiv:1810.04805, BiomeGPT) to rank sample tokens by importance. These weights have two critical limitations: they are nonnegative, so they cannot separate disease-supporting from health-supporting evidence (arXiv:2201.12114), and they act after token fusion, obscuring how the input sources S and A each affect the output. To address this we use Integrated Gradients (arXiv:1703.01365), a signed, fusion-aware attribution method, and propose a source-derived baseline T' = S + A_0 for feature-tokenized models such as BiomeGPT, which preserves species identity as a fixed biological coordinate while isolating the effect of abundance variation. Applied to a disease-versus-health decision margin, it yields polarity that explicitly separates pathogenic from protective microbial signals. We show that this gradient-based approach uncovers species-abundance directional relationships and sensitivity diagnostics entirely obscured by unsigned [CLS] attention weights. We further recommend second-order Integrated Hessians (arXiv:2002.04138) to expose microbiome community interaction rules: how a perturbation in one member alters the model's sensitivity to another, and which other species drive ambiguous cases toward disease or health at a given abundance level. This provides a principled approach to explainability in BiomeGPT that generalizes to other smooth and differentiable feature-tokenized transformers. Code is available at https://github.com/nohren/token-source-attribution
The emergence of medical deepfakes, i.e., medical images manipulated by deep generative models, poses a significant threat to clinical workflows. However, existing detectors suffer from two critical limitations: poor generalization to unseen generative architectures for manipulation detection and lack of interpretability. In this context, we present HexMIL (Hierarchical EXplainable Multiple Instance Learning), a mask-free medical deepfake detector that simultaneously addresses both limitations using only binary volume-level supervision. HexMIL decomposes each CT volume into a two-level hierarchy of patches and slices, aggregated via independent Gated Attention modules whose weights are directly combined into a full-resolution 3D attention volume that localizes the manipulated sub-region without any pixel-level annotation. Unlike post-hoc methods such as Grad-CAM, HexMIL's attention weights constitute the exact forward computation driving the classification decision, providing ante-hoc and structurally faithful spatial attribution. We evaluate HexMIL on M3DSynth and CT-GAN datasets under a rigorous cross-generator generalization protocol, training on a single generative architecture and testing on unseen ones. HexMIL outperforms all baselines by $+9.1$ AUC and $+9.4$ F1 in out-of-domain classification, and achieves the best average IoU and Pointing Game score in localization. Project page: opontorno.github.io/hexmil.