Research Aims
Population-based clinical knowledge, causal inference and longitudinal phenotyping
The Sleep-EVAL research program was created to study sleep disorders in the general population with the clinical depth ordinarily available only in specialized settings. Rather than reducing sleep epidemiology to a small number of screening items, the program was designed to preserve detailed symptom characterization, diagnostic criteria, differential diagnosis, comorbidity, treatment exposure, function and longitudinal change.
Sleep-EVAL is the domain-specific knowledge base for sleep disorders. It contains the structured clinical knowledge that allows Ad-Infer to evaluate sleep complaints and sleep disorders within their psychiatric, neurological, medical, pharmacological and functional context.
Ad-Infer: the integrated inferential system
Ad-Infer is the unique inferential system used across EVAL domains. Within Ad-Infer, EVAL-KBS combines domain knowledge, Type-2 causal reasoning, fuzzy evaluation, Hopfield-based recurrent neural computation, learning processes, deep-learning mechanisms and Bayesian belief updating. Diagnostic inference extends beyond recognition of positive criteria to the evaluation of direct and indirect causal relationships, competing explanations, exclusions, comorbid conditions and treatment effects.
Language-model capabilities are integrated within Ad-Infer. They do not function as a separate diagnostic agent. Ad-Infer uses them to extend semantic understanding of the clinical interaction, adapt or reformulate questions, clarify ambiguous information and explore how a response relates to direct or indirect causality. The interaction then feeds new evidence back into the inferential process.
This feedback may strengthen a positive diagnosis when the evidence is concordant, or it may challenge an existing hypothesis, change the current belief state and become part of the differential-diagnosis process. The status of the interaction itself can also be informative: questions may be asked or skipped, understood or misunderstood, accepted or refused, and these events can determine what Ad-Infer needs to explore next.
Architecture of Sleep-EVAL within Ad-Infer
| Sleep-EVAL: sleep knowledge Sleep symptoms and disorders, sleep-wake timing, diagnostic criteria, exclusions, severity and frequency constructs, psychiatric and medical context, treatments, impairment and function. | Other EVAL knowledge bases GERD-EVAL and other domain-specific knowledge bases provide corresponding clinical knowledge while the inferential mechanism remains Ad-Infer. |
| Ad-Infer / EVAL-KBS inferential architecture Type-2 causal reasoning · fuzzy evaluation · Hopfield neural computation · learning processes · deep-learning mechanisms · Bayesian belief updating · temporal and longitudinal reasoning. | |
| Integrated language-model functions and bidirectional feedback Semantic comprehension, adaptive questioning and clarification are driven by current hypotheses and belief states. Returned confirmatory, ambiguous or contradictory evidence is re-evaluated by Ad-Infer and can reinforce the positive diagnosis or redirect differential diagnosis. | |
| Outputs Clinically interpretable diagnoses, degrees of support, differential hypotheses, structured multidimensional phenotypes, comorbidity patterns, longitudinal trajectories and family-based data. | |
AI-driven clinical phenotyping beyond diagnostic labels
A central research aim is to organize symptoms and clinical states into multidimensional phenotypes while preserving their clinical meaning. Sleep-EVAL knowledge and Ad-Infer reasoning can integrate sleep symptoms, psychiatric dimensions, neurological manifestations, medical disorders, medications, substance exposure, function and health-care use in the same inferential framework.
- Sleep initiation, maintenance, timing, breathing-related events, parasomnias, motor phenomena, hypersomnolence and circadian patterns.
- Mood, anxiety, irritability, anhedonia, suicidality, cognition and behavioral features.
- Neurological and medical symptoms relevant to sleep and vigilance.
- Medication and treatment history, including possible causal or modifying effects.
- Functional impairment, disability, quality of life and health-care utilization.
The goal is not only to assign a categorical diagnosis, but to understand how symptoms, causal relationships, comorbidities and trajectories organize disease expression in the community.
Family-based and longitudinal investigation
The same structured framework can be applied to participants and relatives and repeated over time. Family members can be evaluated with comparable definitions and inferential rules, supporting studies of familial aggregation and intergenerational patterns. Longitudinal reassessment allows later symptoms, objective findings, treatment exposures and clinical events to modify earlier interpretations rather than freezing the participant in the diagnostic state observed at one interview.
Core scientific aims
- Characterize sleep disorders in representative general populations using clinically grounded, knowledge-based evaluation.
- Estimate prevalence, incidence, persistence, remission and transitions over time.
- Generate multidimensional sleep, psychiatric, neurological and medical phenotypes.
- Evaluate positive diagnoses together with alternative explanations and exclusion criteria.
- Study direct and indirect causal relationships and the changing strength of clinical hypotheses.
- Analyze comorbidity, treatment pathways, recognition of disorders, function, disability and quality of life.
- Investigate familial aggregation using standardized assessment of relatives.
- Support multilingual and international epidemiology while preserving the same core clinical knowledge and inference principles.