Truace
Subscribe
The Index recomputed live from the record

What the evidence says.What the public feels.

The record holds 939 sourced gains and 772 sourced problems, averaging 68 and 66 on the index score. Readers have logged 22 public signals on the Pulse, which is kept apart and never counted as evidence.

939 gains
772 problems
Every sourced claim in the Index, one square each, shaded by the strength of its evidence. High Moderate Emerging

Ranks distinct AI gain and problem claims from the published record. Scores reward impact, independent source strength, scale, confidence, and recency.

1,711 results
Show filters and sorting

Download every matching row, not just this page:Export CSVExport JSON

AI gains · 939

505
HealthStableModerate evidence · 1 source

A convolutional neural network using standard 12-lead ECGs can estimate elevated NT-proBNP levels with strong correlation and good discrimination in internal and external validation.

Investigators built and validated an AI model that estimates serum NT-proBNP levels from routine 12-lead ECGs, training on nearly 85,000 ECG-lab pairs and testing internally in over 8,500 patients and externally in 679 patients at two tertiary centers.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
89

Updated Aug 15, 2026 · TRV-2026-0773

67Index score
506
PolicyStableModerate evidence · 1 source

The review proposes an integrated dual-layer governance model that aligns AI-specific risk identification with ISO 14971 processes and EU AI Act obligations, giving regulators and manufacturers a clearer actionable pathway for lifecycle monitoring.

A peer-reviewed review published August 13, 2026 examined how AI-enabled medical devices challenge traditional safety-risk management. Drawing on 19 academic and regulatory sources, it found ISO 14971, AAMI CR34971 and the EU AI Act each cover parts of device safety and algorithmic governance but remain fragmented in practice.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
89

Updated Aug 15, 2026 · TRV-2026-0771

67Index score
507
ScienceStableModerate evidence · 1 source

A delta-ML graph neural network using GFN2-xTB geometries and DFT single-points as low-fidelity inputs predicted PBE0-D3BJ and PBE-D3BJ level properties of transition metal complexes with higher accuracy and better data efficiency than a conventional benchmark.

Researchers presented a delta-ML approach for transition metal complexes that uses GFN2-xTB geometry optimizations combined with DFT single-point calculations to produce low-fidelity inputs, which are converted to graph representations for a graph neural network trained to predict high-fidelity quantum properties from the tmQMg dataset.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
89

Updated Aug 15, 2026 · TRV-2026-0770

67Index score
508
HealthStableModerate evidence · 1 source

ML-based phenomapping of 106,490 Danish KC patients identified 7 distinct subgroups differentiated by disease burden, comorbidities and socioeconomic status, providing a basis for tailored clinical pathways.

On 2026-08-13, a peer-reviewed study reported machine-learning-based phenomapping of 106,490 keratinocyte carcinoma patients from the Danish Skin Cancer Registry (2014-2022). The model derived seven clusters ranging from young, well-educated, high-income, medically noncomplex females with low-risk BCCs to highly comorbid patients with more SCCs and immunosuppressive drug exposure.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
89

Updated Aug 15, 2026 · TRV-2026-0768

67Index score

AI problems · 772

505
HealthStableModerate evidence · 1 source

Current AI detection systems are not specifically trained for IBD-related lesions, limiting effectiveness for colorectal neoplasia detection in inflammatory bowel disease.

On July 17, 2026, BMC Gastroenterology published the VISION-AI protocol for a pragmatic multicenter non-inferiority randomized trial comparing artificial intelligence-based computer-aided detection guided colonoscopy to pancolonic chromoendoscopy for visible colorectal neoplasia detection in adults with longstanding inflammatory bowel disease. The design targets 752 participants with a 15% anticipated neoplasia rate and 7.5% non-inferiority margin, and reports that enrollment started in April 2026 with 19 participants recruited by June 8, 2026.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
83

Updated Jul 19, 2026 · TRV-2026-0266

67Index score
506
HealthStableModerate evidence · 1 source

Some chatbot models showed lower conciseness & focus and clarity, producing longer less-focused answers that may make it harder for patients to maintain focus and perceive information in a structured manner.

A July 2026 peer-reviewed study compared ChatGPT GPT-5.1, Gemini 2.5 Flash, and Claude Sonnet 4.5 against expert periodontologists on 20 periodontal patient questions. Nine blinded periodontologists rated anonymized answers for scientific accuracy, completeness, conciseness & focus, empathy, and clarity.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
83

Updated Jul 19, 2026 · TRV-2026-0265

67Index score
507
LifestyleStableModerate evidence · 1 source

Opaque AI decision-making in food quality models hinders adoption by inspectors and limits reliability, with explainable AI still underutilized in Food Engineering.

Published April 24, 2026, this peer-reviewed review examines the use of explainable AI in Food Engineering. It describes how AI models using spectral imaging are used to detect contaminants and assess freshness to meet food quality standards, but their complexity creates opacity that hinders adoption by quality control inspectors.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
83

Updated Jul 18, 2026 · TRV-2026-0260

67Index score
508
HealthStableModerate evidence · 1 source

Use of ML and DL for mental health early detection faces challenges with data integration, methodological inconsistency, and ethical issues.

A June 2026 review in Journal of Affective Disorders Reports surveyed ML and DL methods for early identification, diagnosis, and treatment of mental health illnesses, covering medical imaging, genetic and biomarker analysis, behavioral assessments, and longitudinal risk prediction models.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
83

Updated Jul 17, 2026 · TRV-2026-0255

67Index score

Recomputed live from the record · Oct 13, 2026, 3:03 PM