Background: Accurate dietary-intake assessment underpins nutrition research and clinical dietetics, yet self-report methods remain vulnerable to recall, portion-size and reporting error. Objective: This structured narrative review evaluates AI-enabled and sensor-supported dietary-assessment software, focusing on technical performance, measurement validity, user engagement, food-composition databases, cultural applicability and clinical readiness. Methods: Evidence was identified through searches of five electronic sources—PubMed (including MEDLINE-indexed records), Scopus, Web of Science Core Collection, IEEE Xplore and ScienceDirect—covering January 2020 to March 2025, and a complete rerun of the same five sources on 7 August 2026, supplemented by backward and forward citation tracking. The retained primary screening record comprised 1620 electronic-source records and 42 additional records (1662 before deduplication); 478 duplicates were removed, 1184 records were screened, 212 full texts were assessed and 15 sources were retained for the primary-phase qualitative synthesis. These figures are retained aggregate review-log totals and should not be interpreted as a fully reconstructable systematic-review audit trail. The 2026 update incorporated 18 publications: 10 from the January 2020–March 2025 historical interval through separate recovery/reassessment, and eight published after March 2025 through the 497-record update-specific screening stream, with retrieval phase reported separately from evidence role. The higher, moderate and limited/emerging within-map convergence descriptors summarise patterns in the purposively selected literature and do not represent exhaustive retrieval, formal certainty of evidence or reproducible screening. Results: Food recognition is most mature for visually distinct foods under controlled conditions. Portion-size estimation, nutrient quantification, mixed dishes, sustained engagement and culturally diverse foods remain less consistent. Selected systems show reasonable group-level energy and macronutrient agreement under defined conditions, whereas validation in free-living and clinical settings remains variable; a 2026 doubly labelled water study reported substantial underestimation and poor individual agreement in females with obesity. Conclusions: Current tools may support self-monitoring, research screening and dietitian-assisted assessment, but stand-alone use for high-stakes therapeutic decisions requires independent, nutrient-specific validation in the target population.

Background/Objectives: Automated dermoscopic image analysis can assist research on early skin cancer detection, but deep learning models may learn acquisition-related shortcuts, including terminal hairs, illumination gradients, gel bubbles, shadows and measurement markings. This study evaluates a causal uncertainty-decomposed ensemble (CUDE) for confounding-aware skin-lesion classification within the ISIC 2019 benchmark setting. Methods: CUDE uses a structured variational autoencoder (SVAE) to impose separate lesion-relevant and acquisition-related nuisance latent heads. Three stochastic latent-space experts are trained on complementary streams and are integrated by a Dirichlet-based fusion module trained with nuisance sampling. The causal graph is used as a modeling prior rather than proof of causal identification. Results: Using a stratified internal split of the labeled ISIC 2019 training collection, CUDE achieved a balanced accuracy of 89.7% and an AUROC of 0.972. Under the predefined synthetic artifact protocol, CUDE showed a relative performance drop of 5.0%, compared with 9.6% for the deep ensemble. Deferring the 10% most uncertain cases increased non-deferred accuracy from 89.7% to 93.2% and reduced false-negative rates for melanoma, BCC and SCC in the non-deferred subset. Conclusions: Within the evaluated ISIC 2019 split and controlled synthetic corruption protocol, structured latent separation, stochastic expert diversity and Dirichlet fusion were associated with improved benchmark robustness and calibration. These results should not be interpreted as evidence of broad real-world robustness or clinical deployment readiness; external multi-center, device-diverse and skin-tone-diverse validation remains required.