Training-Free Clinical Reasoning through Medical Ontologies and Cognitive Mapping: A Symbolic-Probabilistic Knowledge Graph Framework

Surajit Das

cs.AISep 28, 2026

Training-Free Clinical Reasoning through Medical Ontologies and Cognitive Mapping: A Symbolic-Probabilistic Knowledge Graph Framework

Authors: Surajit Das

Abstract

Most clinical prediction systems learn patient-variable-outcome associations; we investigate a training-free diagnostic paradigm mapping patient observations to explicit medical knowledge. CKG Reasoner integrates candidate-specific Evidence Feature Nodes, patient-reference matching, a bounded Information Gate, knowledge-weighted evidence accumulation, disease similarity, and decisive clinical rules. Missing-aware normalization and coverage auditing distinguish absent from unavailable evidence. Candidate ranking is separate from outcome-label-independent K-means clustering, which uses four derived evidence coordinates (evidence strength, relative magnitude, directional similarity, and evidence completeness), not raw predictors or targets, to derive cohort-level assignments. Across six retrospective cohorts - four dengue (N = 1000, 1523, 989, 1018), malaria (N = 2190), and influenza (N = 4569) - a uniform, label-free, cohort-fitted K = 2 protocol yielded positive-class F1 scores of 0.996, 0.634, 0.936, 0.917, 0.695, and 0.842, and all-record accuracies of 0.996, 0.558, 0.914, 0.893, 0.707, and 0.906, respectively, with full partition-decision coverage using the frozen package and disease-specific knowledge representations. Neither scoring nor clustering uses outcome labels. Logistic regression provides a supervised baseline. Influenza incorporates confirmatory molecular PCR and is not independent pre-test prediction. Results characterize knowledge-grounded evidence separation, auditability, and sensitivity, not prospective clinical validity or comparative superiority. FOL/LLM-based clinical explanation remains unevaluated.

Figures & tables

Study	Primary representation	Prob.	Rules	Neg. ev.	Patient-specific	Learned	Intrinsic trace	Main distinction from this work
[ 1 ]	Hybrid rule/probabilistic experts	Yes	Yes	–	Yes	Yes	Partial	Combines expert outputs through a neural network; no explicit clinical KG evidence semantics.
[ 6 ]	Ontology + fuzzy rule system	Fuzzy	Yes	Partial	Yes	Partial	Yes	Semantic similarity and fuzzy inference, but no separate support/contradiction channels or decisive rule override layer.
[ 10 ]	Siamese Bayesian networks	Yes	–	Yes	Yes	Yes	Yes	Explicitly addresses negative evidence, but within a learned BN formulation rather than a disease-specific EFN graph.
[ 12 ]	KG + deep reinforcement learning	–	–	–	Yes	Yes	Partial	Learns interactive diagnostic policy over the KG.
[ 16 ]	KG reasoning paths + dynamic weights	Yes	–	Partial	Yes	Partial	Yes	Weighted paths for TCM syndrome reasoning; no hard pathognomonic/exclusionary override semantics.
[ 27 ]	Semantic KG with LR-weighted/conditional edges	Yes	Conditional	Yes	Yes	No/limited	Yes	Very close weighted transparent reasoning; evidence remains primarily edge-weight based rather than explicitly separated clinical attributes.

Table 1 : Representative diagnostic reasoning systems and their relationship to the proposed framework. “Learned” indicates that a substantive diagnostic component is estimated or fine-tuned from data.

Dataset	Target	Package schema	N / outcome	Evidence profile
A	Dengue	C1_Bangladesh_Dengue_1000	1000; 533 positive / 467 negative	Mixed serological, clinical, and routine laboratory evidence
B	Dengue	C2_D4_Dengue_Hematology_1523	1523; 1042 positive / 481 negative	Hematology-dominant evidence
C	Malaria	C3_M1_Malaria_2190	2190; 1068 positive / 1122 negative	Malaria-specific clinical/laboratory evidence
D	Influenza	C4_I1_Thailand_Influenza_4569	4569; 1493 positive / 3076 negative	Symptoms plus molecular PCR and rapid-antigen evidence
E	Dengue	D7 additional evaluation	989; 644 positive / 345 negative	Hematology-focused; four headings mapped to existing Dengue EFNs
F	Dengue	D3 additional evaluation	1018; 697 positive / 321 negative	Clinical symptoms plus platelet and WBC evidence; eight headings mapped to existing Dengue EFNs

Table 2 : Datasets used in the executed CKG Reasoner evaluation.

Original variable	Ontology feature	Original variable	Ontology feature
Gender	Gender	Age	Age
NS1	NS1	IgG	IGG
IgM	IGM	Fever Duration	FEV_dura_dy
Body Temperature	FEV_temperature_c	Platelet Count	MTP_platelet_count
WBC Count	LEU_wbc	Joint Pain	ART_severity
Headache	HDH	Retro-Orbital Pain	ROP

Table 3 : Audited mapping of variables in Dengue Cohort A to ontology features.

Study	Method / Architecture Description	Artifact / Access Status	Technical Reason Precluding Execution on Target Cohorts
[ 1 ]	Hybrid rule-based/probabilistic expert system combining expert outputs via neural network.	Paywalled (IEEE) No public code	Historical 1992 proceeding pre-dating digital code repositories; no source code, inference software, or calibrated parameters available.
[ 6 ]	Ontology-grounded fuzzy decision support system using semantic similarity for diabetes.	Open Access (IEEE Access) No public code	Code unreleased; hospital training dataset is private; transfer to acute febrile cohorts impossible without calibrated fuzzy membership functions.
[ 10 ]	Siamese Bayesian networks incorporating symptom absence as negative evidence over learned BNs.	Paywalled (ACM) Proprietary commercial IP	Developed as proprietary commercial IP for the mFine telemedicine platform; network topologies, learned weights, and code are strictly unreleased.
[ 12 ]	DKDR: Knowledge graph reasoning with deep reinforcement learning for interactive diagnosis.	Paywalled (IEEE) No public code	Closed conference proceeding; no public code repository, pre-trained policy checkpoints, or KG simulation environment released.
[ 16 ]	Multi-hop KG path reasoning with dynamic TF-IDF and Naive Bayes weights for TCM diagnosis.	Open Access (CMC) No public code	Open-access publication, but no source code, reasoning scripts, or path weight tables were publicly deposited.
[ 27 ]	Semantic KG reasoning with likelihood-ratio-weighted edges and conditional patient constraints.	Paywalled (Springer) No public code	Published as a conceptual book chapter in conference proceedings; no software implementation, graph exports, or code packages released.

Table 4 : Reproducibility and empirical execution audit of representative diagnostic reasoning systems surveyed in Table 1 when considered for baseline benchmarking on our target cross-sectional cohorts.

Cohort	Dataset Name	Accuracy	F1 Score	Sensitivity	Specificity	Precision	ROC-AUC	PR-AUC	Test Size (Pos / Neg)
A	Bangladesh mixed evidence (Dengue)	100.00%	100.00%	100.00%	100.00%	100.00%	1.0000	1.0000	200 (107 / 93)
B	Hematology (Dengue)	60.00%	68.23%	62.68%	54.17%	74.86%	0.6513	0.7793	305 (209 / 96)
C	Clinical Data (Bangladesh) (Malaria)	70.09%	69.89%	71.03%	69.20%	68.78%	0.7539	0.7022	438 (214 / 224)
D	Thailand ILI cohort (Influenza)	88.84%	82.59%	80.94%	92.68%	84.32%	0.9268	0.8784	914 (299 / 615)
E	D7 additional evaluation (Dengue)	92.42%	94.30%	96.12%	85.51%	92.54%	0.8968	0.8918	198 (129 / 69)
F	D3 additional evaluation (Dengue)	99.02%	99.29%	99.29%	98.44%	99.29%	0.9941	0.9971	204 (140 / 64)

Table 5 : Cross-cohort supervised logistic regression baseline performance across the six evaluation cohorts. Thailand Influenza and Dengue Cohort A incorporate point-of-care rapid testing alongside clinical and laboratory presentation. Bold highlights values as formatted in the supplied baseline table; cross-cohort values are not directly comparable because cohorts differ.

Dataset	Target	N	Accuracy	Bal. Acc.	F1 +	MCC	Coverage	Silhouette
A: Bangladesh mixed evidence	Dengue	1000	0.996	0.996	0.996	–	1.000	0.724
B: Hematology	Dengue	1523	0.558	0.558	0.634	0.108	1.000	0.635
C: Clinical Data (Bangladesh)	Malaria	2190	0.707	0.706	0.695	0.413	1.000	0.889
D: Thailand ILI cohort	Influenza	4569	0.906	0.871	0.842	0.783	1.000	0.842
E: D7 additional evaluation	Dengue	989	0.914	0.895	0.936	–	1.000	0.870
F: D3 additional evaluation	Dengue	1018	0.893	0.907	0.917	0.776	1.000	0.601

Table 6 : Uniform

K=2

KG partition results. F1 is positive-class F1; all six cohorts have full partition-decision coverage. MCC is omitted where it was unavailable for the uniform protocol.

Dataset	Full	No $P_{\mathrm{evidence}}$	No $r$	No $c$	$P_{\mathrm{evidence}}$ only	Profile only	No CS
B	0.634	0.612	0.660	0.607	0.694	0.612	0.634
C	0.695	0.695	0.695	0.692	0.695	0.695	0.695
D	0.842	0.842	0.843	0.844	1.000	0.842	0.842
F	0.917	0.925	0.928	0.930	0.931	0.925	0.917

Table 7 : Available

K=2

component ablations for Datasets B, C, D, and F relative to the full four-dimensional post-reasoning representation. Values are positive-class F1. The No-CS column is the explicit three-dimensional

[P_{\mathrm{evidence}},r,c]

ablation.

Explore similar work

Oct 5, 2026cs.AI

Proof-Grounded Patient-Specific Clinical Explanations from Knowledge-Graph Reasoning

Clinical decision-support outputs can lack an au- ditable link between patient observations, encoded knowledge, conclusions, and recommendations. We present the CKG Clinical Explanation Engine, a downstream layer for a frozen, training-free clinical knowledge-graph reasoner that converts patient inference states and disease knowledge into typed facts, explicit rule-application traces, provenance-linked conclusions, and policy-licensed recommendations. The design separates measurement availability, representation completeness, and disease-specific activation; consequently, observed zero-activation evience is not treated as missing and partial representation is distinct from unobserved evidence. Optional language generation is restricted to symbolically licensed content. Across five usable workbooks (6,720 patients; 20,160 patient-disease traces; 1,021,440 feature-evidence rows), IG-range validity and knowledge provenance were 100%, numerical cross-sheet fidelity was 100% (120,960/120,960), and exported logical/report trace completeness was 100% (20,160/20,160). Availability representation consistency was 99.7028% (1,018,404/1,021,440); all 3,036 disagreements were confined to three systematic feature-cohort patterns. The corpus contained 86,783 observed zero-activation and 139,949 observed partially represented instances. A separate seeded 25-patient end-to-end audit completed without execution failure and passed all pre-specified trace, licensing, provenance, and state-consistency checks. These results establish structural and implementation auditability, not clinical correctness or utility.

Surajit Das

Biomedical Knowledge Graph

Ask PDF

May 21, 2026cs.CL

ChronoMedKG: A Temporally-Grounded Biomedical Knowledge Graph and Benchmark for Clinical Reasoning

Biomedical knowledge graphs (KGs) treat disease associations as static facts, but temporal information is crucial for clinical reasoning, e.g., a symptom diagnostic of one disease at age 3 may imply a different disease at age 13. Existing KGs such as PrimeKG, Hetionet, and iKraph do not encode when a finding becomes clinically relevant over the course of a disease. This limits their usefulness for longitudinal clinical reasoning and retrieval augmentation. We introduce ChronoMedKG, a temporal biomedical knowledge graph that contains 460,497 evidence-linked triples (filtered from 13M raw extractions) covering 13,431 diseases. Each association is tied to temporal components like onset window or progression stage, which are backed by PMID-traceable evidence and a multi-signal credibility score. The graph is constructed through a disease-autonomous multi-agent pipeline in which multiple frontier LLMs independently extract knowledge from PubMed and PMC literature. Only those relations are kept that are supported by multi-model consensus, survive credibility filtering, as well as ontology alignment. ChronoMedKG scored 92.7% agreement against Orphadata and adds temporal grounding for 6,250 diseases absent from HPOA, Orphadata, and Phenopackets, including 1,657 Orphanet-coded rare diseases. We further introduce ChronoTQA, a benchmark of 3,341 questions across eight task types (six temporal plus two static controls), with a 12-question supplementary probe. Frontier LLMs lose roughly 30 points moving from static to temporal questions; ChronoMedKG retrieval rescues 47-65% of their long-tail failures, against 17-29% for HPOA-RAG. As such, ChronoMedKG provides a crucial temporal axis for retrieval-augmented clinical systems that was previously absent.

Md Shamim Ahmed, Farzaneh Firoozbakht, Lukas Galke Poech +2

University of Southern Denmark · University of Hamburg

Biomedical Knowledge Graph Clinical Reasoning Training Longitudinal Patient Records

Ask PDF

Jun 21, 2026cs.CL

Knowledge-Graph Grounding Helps LLMs Only for Out-of-Training Knowledge: A Controlled Study on Clinical Question Answering

A recent Nature Medicine study reports that general-purpose frontier LLMs outperform specialized retrieval-augmented clinical tools on medical benchmarks, and that retrieval can hurt strong models. We ask the natural follow-up: does structured knowledge-graph (KG) grounding change this, and when does grounding help at all? We contribute two results. First, a reproduction: the study's headline HealthBench score (~88) is the Consensus variant, not full HealthBench, where frontier models and ideal completions both score ~46-47 under a physician-calibrated grader (agreement 82.5%); we reproduce GPT-5.2 Consensus =90.9 and flag a score-deflating grader bug. Second, a knowledge-boundary result. Using a graph+vector engine (samyama-graph) over the public biomedical KG PrimeKG, neither naive triple retrieval nor an agentic natural-language-to-Cypher loop (82% successful queries) improves MedQA across a weak-to-strong model ladder (all |Delta| <= 3.4). On a synthetic counterfactual KG, and on a hybrid benchmark mixing known and novel facts, the identical pipeline lifts out-of-training accuracy from chance to ~100% (+68 to +79) while adding nothing on known facts (a no-LLM arm answers both). Across three regimes (no-knowledge, graph-aided, hybrid), grounding helps only insofar as the decisive fact lies outside the model's training -- public-KG facts are redundant, private and novel data are where it pays -- matching the study's institutional-data caveat.

Madhulatha Mandarapu, Sandeep Kunkunuru

samyama

Contextual Grounding Controlled Study Physical Activities

Ask PDF