Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “explainable AI”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 37 records · Page 2Linked to original sources
PubMed1988

[The antigen-antibody reaction and its inhibition].

The sera of intact and immune animals contain factors capable of inhibiting the interaction of antibodies with the homologous antigen. These agglutination inhibiting (AI) factors not only block the antigen, thus competing with antibodies, but may induce the dissociation of the antigen-antibody complex. Heating of antisera at 60-63 degrees C leads to an increase in the activity of AI factors. The phenomenon of "prozone" in the agglutination test appears to be explained by the presence of AI factors in sera.

Agglutination Tests↗
PubMed2026

Beyond predictive performance: A systematic review and critical methodological appraisal of AI/ML and conventional modelling strategies in breast, colorectal, and pancreatic Cancer.

BACKGROUND: Predictive modelling for cancer risk, treatment-related complications, and survival is central to precision oncology. Conventional logistic regression (LR) and Cox proportional hazards (CoxPH) regression remain widely used but are limited when modelling nonlinear interactions, high-dimensional imaging features, and multimodal clinical-metabolic predictors. Artificial intelligence (AI) and machine learning (ML) methods offer expanded capability through automated feature extraction, ensemble learning, and flexible survival modelling, but the evidence on when AI/ML adds value over conventional models across cancer sites and predictive tasks remains fragmented. OBJECTIVE: To systematically evaluate the methodological performance, validation strategies, and translational limitations of AI/ML models compared with conventional statistical models in published predictive-modelling studies for breast, colorectal, or pancreatic cancer. METHODS: PubMed, Scopus, and Web of Science were searched for studies published between January 2019 and March 2025. Two reviewers independently conducted title-and-abstract screening, full-text eligibility assessment, and PROBAST risk-of-bias assessment. Sixty-five studies (n = 907,567 participants) were narratively synthesised by cancer site, predictive task, model family, comparator, validation strategy, predictor modality, and calibration or explainability reporting. RESULTS: The 65 studies comprised breast cancer (n = 35), colorectal cancer (n = 21), and pancreatic cancer (n = 9). AI/ML superiority over LR and CoxPH was task- and data-dependent. CNN- and U-Net-based models predominated in imaging and body-composition tasks, tree-based ensembles consistently outperformed LR for tabular perioperative complication prediction, and CoxPH remained competitive, and in the largest pancreatic risk study, superior to XGBoost (C-index 0.802 vs 0.723) in well-structured datasets. PROBAST analysis-domain risk was moderate in 54 of 65 studies (83%), driven by limited external validation, sparse calibration reporting (11/65), and few decision-curve analyses (7/65). CONCLUSION: AI/ML adds the most methodological value in imaging-derived feature extraction and nonlinear perioperative prediction, while conventional regression remains preferable in large, structured datasets with linear predictors. Clinical translation requires standardised body-composition definitions, external validation, calibration assessment, decision-curve analysis, and explainability, in line with TRIPOD+AI and CLAIM standards.

Humans↗
PubMed1991

Two types of asymmetric acetylcholinesterase in chick hindlimb muscle: developmental profiles, in vivo and in cell culture, and recovery after inactivation.

1. We have analyzed the behavior of two types of asymmetric molecular forms (A forms) of acetylcholinesterase (AChE) during development of chick hindlimb muscle, in vivo and in cell culture, and upon irreversible inactivation of peroneal muscle AChE with diisopropylfluorophosphate (DFP) in vivo. 2. In agreement with previous developmental studies on chick muscle, globular forms of AChE (G forms) are predominant in chick hindlimb at early embryonic ages, being gradually replaced by A forms as hatching (and, therefore, onset of locomotion) approaches. Of the two A-form types, AI appears and accumulates significantly earlier than AII, so that A/G and II/I ratios higher than 1 are attained only at about hatching time. 3. Cultures prepared from 11-day chick embryo hindlimb myoblasts express both types of A forms, with a combined activity of 27% of total AChE after 12 days in culture. AI forms appear again earlier and are much more abundant than type II asymmetric species through the life span of cultures. 4. All AChE activity in the peroneal muscle is irreversibly inactivated by injection of DFP in vivo. The recovery of A forms follows the same sequence described for normal development, with a delayed and slower recovery of AII forms as compared with AI. 5. Several hypotheses involving tail polypeptides or tissue target molecules, or posttranslational interconversion, are proposed to help explain the earlier appearance and accumulation of AI forms in chick muscle.

Acetylcholinesterase↗
PubMed2026

Data-centric, robust, and explainable multimodal deep learning for clinical decision support: A systematic review.

PURPOSE: Multimodal deep learning is increasingly proposed for clinical decision support (CDS) under a "data-centric" framing that prioritizes label quality, missing-modality robustness, distribution shift, calibration, and explainability. Prior reviews have examined multimodal medical AI, CDS, and data-centric methods separately, but none address their intersection. We mapped the modalities, fusion strategies, and data-centric and explainability techniques used in this recent literature, quantified how often each is implemented rather than merely mentioned, assessed deployment-relevant evidence (external validation, clinical-outcome measurement, equity), and formally appraised study-level risk of bias. METHODS: Following the PRISMA 2020 statement (PROSPERO CRD420261427815; registered retrospectively), we screened 150 records and included primary, clinical, multimodal studies that applied machine or deep learning to a decision-support task and reported at least one quantitative result. Two reviewers screened and extracted data with consensus adjudication. Each study was coded against pre-specified operational definitions, separating implemented or empirically evaluated techniques from those only mentioned. Study-level risk of bias was assessed with PROBAST + AI. Synthesis was narrative. RESULTS: Thirty-one studies met inclusion; 30 (97%) were published between 2024 and 2026, with a median of three modalities (range 2-6), most commonly structured EHR (71%) and imaging (39%). Data-centric techniques were frequently reported (74-84% across label-noise, distribution-shift, calibration, missing-modality and class-imbalance handling; equity 61%). However, external validation was reported in only 4/31 studies (13%), a clinical or provider outcome in 3/31 (10%), and no study reported routine deployment. Overall risk of bias was high in 27/31 studies (87%), driven by the analysis domain. CONCLUSION: Within this recent, self-selected slice of the field, technical robustness and explainability techniques are widely reported but rarely validated out-of-distribution or against clinical outcomes, and the underlying evidence is at high risk of bias. Progress requires external multi-site validation, clinical-outcome measurement, formal bias appraisal, and adherence to AI reporting standards (e.g., TRIPOD + AI) before deployment can be justified.

Deep Learning↗
PubMed1983

[Lipoproteins, apolipoproteins, lipoprotein lipase, hepatic triglyceride lipase and lecithin cholesterol acyltransferase in patients with nephrotic syndrome].

Chronic renal disease with secondary hyperlipidemia is highly atherogenic. In uremia and patients on chronic hemodialysis there is a high incidence of atherosclerotic complications whereas the incidence of atherosclerotic disease is relatively low in the nephrotic syndrome. This is surprising, as nephrosis produces type-II hyperlipidemia, which is usually highly atherogenic. In this study 10 patients (5 male, 5 female) with a newly diagnosed nephrotic syndrome were compared to 10 controls (5 male, 5 female). As laboratory parameters, lipids, lipoproteins (VLDL, IDL, LDL, HDL2 and HDL3 by rate zonal centrifugation) and the percentage composition of the major apolipoproteins in VLDL, HDL2 and HDL3, as well as lipoprotein lipase (LPL), hepatic lipase (HTGL) and lecithin-cholesterol-acyl-transferase (LCAT) were measured. In nephrotic patients significantly higher plasma levels of cholesterol, triglycerides, phospholipids, VLDL, IDL and LDL were found, whereas HDL-chol, HDL2 and HDL3 were unchanged. LPL and HTGL were both significantly impaired, whereas LCAT was distinctly increased. The percentage composition of apolipoproteins in HDL2 and HDL3 was normal. In nephrotic VLDL, apo-AI was distinctly increased at the expense of a decrease in apo-CII, and increased LCAT was explained by the relative rise of apo-AI in nephrotic VLDL. The increase in apo-AI in VLDL is discussed as a possible reason for the low atherogenic risk of secondary hyperlipidemia in nephrotic syndrome.

Adolescent↗
PubMedDate not supplied

Highly pathogenic virus recovered from chickens infected with mildly pathogenic 1986 isolates of H5N2 avian influenza virus.

A combination of in vitro and in vivo selection procedures was used to examine the possibility that certain mildly pathogenic field isolates of avian influenza (AI) virus may contain minority subpopulations of highly pathogenic virus. Two mildly pathogenic H5N2 isolates, A/chicken/New Jersey/12508/86 (NJ12508) and A/chicken/Florida/27716/86 (FL27716), recovered from chickens epidemiologically associated with urban live-bird markets, were cloned in trypsin-free chicken embryo fibroblast cultures. Selected clones were inoculated intranasally and intratracheally (IN/IT) into specific-pathogen-free laying hens, and virus reisolated from the hens that died was serially passed in hens by IN/IT inoculation. Several highly pathogenic reisolates were recovered from hens infected with the cloned NJ12508 or FL27716 virus. A highly pathogenic NJ12508 reisolate killed 19 of 24 IN/IT-inoculated hens, and a FL27716 reisolate killed all 24 inoculated hens; signs and lesions were typical of fowl plague. In contrast, uncloned NJ12508 stock virus killed 1 of 24 hens and FL27716 stock virus killed 4 of 24 hens, and neither produced the complete spectrum of lesions associated with fowl plague. Recovery of highly pathogenic viruses from these isolates demonstrates the coexistence of pathogenically distinct subpopulations of virus. Competition for dominance among such subpopulations could explain the variable pathogenicity of some AI viruses.

Animals↗
PubMed1989

Microradiographic study of amelogenesis imperfecta.

A material of 22 primary and 4 permanent teeth from 22 children with amelogenesis imperfecta (AI) were examined by microradiographic techniques. The children were part of a patient material earlier examined in genetical and clinical studies. The results were compared with corresponding data two non-affected control groups and correlated with the available clinical and genetical data. Teeth were examined from seven of the eight different variants of AI seen in the clinical study. In most cases both hypoplasias and areas of hypomineralization were observed in the same tooth, indicating that both the secretory and the maturation phases of the amelogenesis are affected in AI. In teeth from children with the same clinical variant but different inheritance patterns, no specific finding could be related to a specific inheritance pattern. The findings in the one boy with AI as an X-linked trait were unique in this material. In all control teeth except one, no hypoplasisas or areas of hypomineralization were found in the enamel. In conclusion, the subclassification of AI into different forms can be questioned. Variations in clinical and histologic characteristics connected with the same inheritance pattern suggest that the genetic defect, in conjuction with a large biological variation, could explain the multiplicity in clinical expressivity that characterizes AI.

Amelogenesis Imperfecta↗
PubMed2026

A Boveri perspective on cancer biomarker testing using artificial intelligence.

Artificial intelligence (AI) can predict genomic alterations from histology, yet its adoption is slowed by a lack of trust. We argue that deliberate morphology (i.e., a cognitive understanding of histological features supported by standardized annotations) creates a bidirectional feedback loop between clinical practice and model outputs.We translate these observations into an actionable hypothesis for clinical and computational teams: that by enhancing explainability, deliberate morphology could facilitate the responsible deployment of AI biomarkers in oncology.

Journal Article↗
PubMedDate not supplied

Indirect evidence for a very fast recovery kinetics of chlorophyll-aII in spinach chloroplasts.

The 690 nm absorption change reflecting the turnover of the system-II-reaction center chlorophyll, Chl-aII (often referred to as P 680), has been investigated under different experimental conditions in spinach chloroplasts. A comparison was made with oxygen evolution and with absorption changes of Chl-aI measured at 703 nm, both indicating the number of electrons produced by system II. It was found: 1. The dependency on actinic flash intensity of the initial amplitudes of the measured 690 nm absorption change, deltaalpha0(Chl-aII) in Tris-washed chloroplasts is similar to that for the total amplutude of the 703 nm absorption change, deltaalpha0(Chl-aI) in normal chloroplasts, and can be described by an exponential function. On the other hand, deltaalphao(chl-aII) in normal chloroplasts exhibits a more complex biphasic dependency and much higher flash intensities are required for saturation. 3. Unver repetitive flash group excitation and in the presence of an ADRY(= acceleration of the deactivation reactions of the water-splitting enzyme system Y)-reagent the initial amplitude of the 690 nm absorption change oscillates in the same characteristic pattern as the oxygen evolution. 4. The initial amplitude of the 690 nm absorption change, deltaalpha0(Chl-aII), IN Tris-washed chloroplasts becomes significantly smaller (more than 50%) by the addition of system-II-electron donors (benzidine, p-phenylendiamine, tetraphenylboron), whereas the total amplitude of the 703 nm absorption change, detalalpha0)Chl-aI) increases 3-4-fold. In order to explain these results, the existance of a very fast reduction kinetics of Chl-aII+ is postulated, which is not detectable by our measuring equipment. The half time of this reaction is less than or equal to mus. Reaction centers with the very gast "undetected" Chl-aII+-reduction are photochemically transformed into slower one by double hit processes with a comparatively low quantum yield. Furthermore, it is inferred, that the dark recovery kinetics of Chl-aII is dependent on be charge accumulation state of the watersplitting enzyme system Y. This phenomenon is shown to explain also the oscillation pattern of delayed fluorescence. On the basis of the present results two alternative reaction schemes for the functional organization of the electron transport on the donor side of system II are discussed.

Chlorophyll↗
PubMed1989

Replenishment of AI-doses with oestrogens in physiological amounts: effect on sow prolificacy in a field trial.

Basing on results about physiological functions of seminal oestrogens in the genital tract of sows, the effects of an oestrogen replenishment to AI-doses were investigated in a field trial. Each ejaculate was split into two halves, which were either diluted to normal AI-doses (controls, n = 353) or diluted and replenished with oestrogens in physiological amounts (n = 384). Insemination by qualified technicians led to an improvement of the pregnancy rate (82.8% vs. 77.1%; p less than 0.05) and the litter size (10.8% vs. 10.3%; p less than 0.05) in favour of the oestrogen replenishment. These results partly explain the known differences in prolificacy between natural mating and AI and thus provide a basis for improvement of pig AI.

Animals↗
PubMed1989

Genetic determination of plasma apolipoprotein AI in a population-based sample.

Apolipoprotein AI (apo AI) is the major protein of high-density lipoprotein (HDL). Using radioimmunoassay, we measured plasma apo AI levels in 1,880 individuals in 283 pedigrees randomly selected from the population with respect to disease status and risk factors for coronary artery disease. Apo AI levels were first adjusted for date of assay (6.8% of apo AI variation) and then adjusted for variability in age and body mass index (an additional 6.6%, 20.4%, and 23.0% of apo AI variations for males, females not using exogenous hormones, and females using exogenous hormones, respectively). A mixture of two normal distributions fit the adjusted data better than did a single normal distribution. Genetic and environmental models that could explain the mixture of normal distributions were investigated using complex segregation analysis. Heterogeneous etiologies for individual differences in adjusted apo AI levels were suggested by the data in the 283 pedigrees. In a subset of 126 pedigrees, there is evidence for the major effect of a nontransmitted environmental factor that explains the mixture of distributions as well as polygenic loci that influence apo AI levels within each distribution. The environmental factor and polygenic loci account for 32% and 65% of the adjusted variation, respectively. In the other 157 pedigrees there is strong support for a single locus with a major effect that accounts for 27% of the adjusted variation. The effect of the polygenic loci is not different from zero in these 157 pedigrees. This is the first study to present evidence for the segregation of a single unmeasured locus with a major effect on levels of apo AI in a population-based sample of pedigrees.

Adolescent↗
PubMed2026

The application of artificial intelligence in healthcare practice: A mapping review of systematic reviews.

Artificial intelligence (AI) is rapidly transforming healthcare practice, with growing evidence supporting its use in diagnosis, prognosis, treatment planning, and operational decision-making. The proliferation of systematic reviews in recent years underscores the need for an updated synthesis of the literature to inform research, policy, and practice. We searched PubMed, Web of Science, Scopus, IEEE Xplore, and CINAHL for systematic reviews and meta-analyses published between 2019 and February 2026. Eligible reviews focused on AI applications in healthcare practice, were peer-reviewed, and written in English. A total of 368 reviews met the inclusion criteria. Publication volume increased steadily, peaking in 2025. AI research was concentrated in high-density domains, such as radiology, oncology, and critical care. Across reviews, diagnostic imaging, electronic health record (EHR) data, and biomarkers/laboratory results accounted for 68% of training data sources, though newer data types, such as wearable device and sensor data, emerged from 2022 onward. Diagnosis, prognosis, and treatment comprised over 80% of AI applications, with novel uses emerging in recent years, such as AI-assisted clinical documentation (e.g., ambient documentation tools) and patient education. Ethical concerns were reported in 78.5% of reviews, with privacy, model accuracy, data and algorithmic bias, and explainability as recurrent themes. The proportion of reviews reporting ethical concerns increased from 2021 to 2025. AI applications in healthcare are expanding in scope, diversifying in data sources, and evolving toward novel clinical and operational uses. The human-centered AI or augmented intelligence paradigm, integrating computational precision with clinical expertise, holds significant promise but will require parallel advances in governance, regulatory frameworks, and ethical oversight to ensure safe adoption.

Artificial Intelligence↗
PubMed1990

Genetic variation at the apolipoprotein gene loci contribute to response of plasma lipids to dietary change.

Dietary intervention studies (from a low polyunsaturated/saturated fatty acid ratio P/S diet to a high P/S diet), carried out on a group of healthy individuals from North Karelia, Eastern Finland between 1981-1984, provided evidence that there may be a genetic component contributing to variation in response to dietary change. We have resampled blood from 107 individuals involved in the original studies and used Restriction Fragment Length Polymorphisms (RFLPs) to study the genetic contribution of variation at a number of candidate gene loci to the response to dietary change. The genes investigated in this study were the apolipoprotein (apo) genes: apo B, apo AII, apo E (protein polymorphism), apo AI-CIII-AIV gene cluster, and the LDL-receptor gene. On the basal diet the major effect of genotype on lipid traits was due to variation at the apo E gene locus; this protein polymorphism explained 14.6% of the phenotypic variance in LDL cholesterol levels and 12.7% of the phenotypic variance in total cholesterol levels. When switched to low fat high P/S diet, these effects of variation at the apo E gene locus on the phenotypic variation of LDL and total cholesterol levels disappeared. The major effect on the response to dietary change, delta, was seen on the difference in apo AI levels mediated by variation at the apo B gene locus (MspI RFLP) explaining 6.3% of the phenotypic variance in apo AI change. For the RFLPs of the apo AI-CII-AIV gene cluster, small but not significant differences on delta were found. Our results indicate that within the limits of the candidate genes studied, the major effects in response to dietary change was on apo AI levels mediated through variation at the apo B gene locus.

Adult↗
PubMed2026

AI-Driven Precision Medicine in Alzheimer's Disease: Drug Repurposing, Digital Therapeutics and Clinical Decision Support.

Alzheimer's Disease (AD) is a neurodegenerative disease that causes significant clinical, social, and economic burden worldwide. Despite improvements in understanding its multifaceted pathogenesis, current treatments are mostly symptomatic and ineffective across varied patient populations. To overcome these constraints, AI-driven precision medicine allows tailored risk assessment, treatment selection, and disease monitoring. This review covers AI's role in AD precision medicine, focusing on drug repurposing, digital therapies and clinical decision support systems. Machine and deep learning models are used to predict medication response, integrate heterogeneous data sources such as genomics, transcriptomics, neuroimaging and electronic health records, and uncover pharmacogenomic treatment success factors. The paper covers AIenabled precision pharmacology, including tailored dosing algorithms, adaptive therapeutic monitoring, and adverse drug reaction prediction. Bioinformatics-based target identification, network pharmacology, graphbased AI models, virtual screening, and real-world and clinical data validation are emphasized in AI-driven medication repurposing. AI-powered digital treatments like personalized cognitive training platforms, wearable- derived digital biomarkers, virtual and mixed reality interventions, adherence monitoring, and digital twins for therapy optimization have been discussed. AI-based clinical decision support systems are also thoroughly assessed for clinical value, accuracy, and explainability in disease subtyping, trajectory prediction, and risk stratification in preclinical and prodromal AD. Despite these promises, data heterogeneity, algorithmic bias, legal barriers, and privacy concerns exist. Federated learning enables safe multi-center collaboration and hybrid AI-human approaches, and it represents the future. AI's ability to alter AD care opens the door to precision medicine paradigms that use repurposed medications, digital tools and intelligent decision-making to improve patient outcomes.

Alzheimer’s disease↗
PubMed1988

Embryonal carcinoma in two cases of androgen insensitivity syndrome: clinical, endocrinological and pathological features.

Embryonal carcinoma in two cases of complete androgen insensitivity syndrome (CAIS) is reported. In both cases gonadectomy carried out for prophylactic purposes led to the discovery of a localized embryonal carcinoma with areas of anaplastic seminoma in one case. In non-neoplastic tissue, gonad morphology in both cases was typical of AIS. Prevalently hypotrophic aspects, especially in the interstitial gland, were found in case 2. This may explain the different endocrine profile in the two cases before gonadectomy. Our study, aside from series of psycho-sexual problems, shows, according to all Authors, that the most serious complication is the high risk of malignancy after puberty in patients with AIS.

Adolescent↗
PubMed1988

Ear-to-ear derivation in short latency brainstem auditory evoked responses.

We utilized an ipsilateral to contralateral earlobe (Ai-Ac) derivation in addition to the scalp to ipsilateral ear (Cz-Ai) and scalp to contralateral ear (Cz-Ac) derivations, in 12 normal hearing community volunteers and 36 patients with a variety of referrals and varying degrees of hearing loss. In normal subjects, the latency of wave I in the Ai-Ac derivation was identical to that in Cz-Ai, but amplitude was consistently smaller (0.21 +/- 0.1 vs. 0.30 +/- 0.13 microV, p less than 0.005). The wave III behaved in a reverse manner. These data can be easily explained based on traditional principles of near- and far-field potentials. The amplitude differences of the wave I in Cz-Ai and Ai-Ac derivations were, however, small and the phase-reversals of that wave between the 2 derivations were striking and consistent in all subjects and patients. This observation indicates that the addition of Ai-Ac derivation to the conventional 1 or 2 channel montage may aid in the identification of wave I.

Adolescent↗
PubMed1976

The HLA system and the clinical response to treatment with chlorpromazine.

A group of 33 schizophrenic patients were typed for HLA-SD antigens and their qualitative clinical responses to chlorpromazine therapy determined. A highly significant positive correlation was found between response to chlorpromazine and HLA-AI positive, while HLA-A2 positive subjects showed a significant negative correlation to chlorpromazine treatment. In a second group of 17 patients the clinical response to chlorpromazine were evaluated quantitatively, by WPRS, in HLA-AI positive and HLA-AI negative patients. There were no pre-treatment differences in the scores. After treatment the scores of positive patients were significantly lower, indicating that they responded to a greater degree. Since the frequency of HLA-AI in hebephrenic patients is higher than that in other schizophrenics this may explain our earlier finding that hebephrenics, as a group, respond better to chlorpromazine than do other schizophrenics.

Anxiety↗
PubMedDate not supplied

Failure of the müllerian regression factor in two patients with complete androgen insensitivity syndrome.

The present paper describes the histological and endocrinologic features of 2 subjects with 46,XY karyotype affected by complete androgen insensitivity syndrome (AIS) with müllerian structures. Both patients had fallopian tubes, but only one had also uterus and presented a seminoma. Serum levels of luteinizing hormone, testosterone and estradiol were high or in the upper part of normal limits, whereas levels of follicle-stimulating hormone were normal. The association between AIS and the presence of müllerian structures observed in these 2 patients might be explained by an impaired synthesis of müllerian regression factor or by a failure in its mechanism of action.

Adolescent↗