手の骨折検出のためのマルチモダルの大型言語モデルの診断精度と安定性:平面放射線写真での複数の実行評価

Ibrahim Güler^1,2, Gerrit Grieb^3,4, Armin Kraus¹

¹Department of Plastic, Aesthetic and Hand Surgery, Otto-von-Guericke University, 39120 Magdeburg, Germany.

Diagnostics (Basel, Switzerland)

|February 13, 2026

まとめ

マルチモダル大言語モデル (MLLMs) は,骨折検出の有望性を示しているが,診断の安定性は欠けている. GPT-5 Proは最高の精度と一貫性を提供しましたが,MLLMは実験的なツールです.

科学分野:

放射線学放射線学
人工知能 (AI) とは,人工知能 (AI) のことです.
メディカルイマージング (医学イメージング)

背景:

マルチモダル大言語モデル (MLLMs) は,自動骨折検出のために調査されています.
繰り返し推論されたMLLMの診断の安定性と一貫性は十分に理解されていません.
本研究では,四つの主要なMLLMの性能を評価し,普通のX線写真で手の骨折を検出しました.

研究の目的:

手の骨折を特定するための4つのMLLMの診断の正確性と安定性を評価する.
GPT-5 Pro,Gemini 2.5 Pro,Claude Sonnet 4.5およびMistral Medium 3.1.のイントラモデル一貫性と信頼性を比較するために
異なる骨折タイプにおけるMLLMのパフォーマンスを評価し,人口統計的推論の能力を評価する.

主な方法:

4つのMLLM (GPT-5 Pro,Gemini 2.5 Pro,Claude Sonnet 4.5,Mistral Medium 3.1) は,65人の患者の手によるX線画像を分析した.
各画像は,同一のゼロショットプロンプトを使用して,モデルごとに5回推論され,合計1300の推論が行われました.
診断の正確性,インターラン信頼性 (Fleiss' κ) と人口統計的推論能力が評価されました.

主要な成果:

GPT-5 Proは最も高い精度 (64.3%) と一貫性 (κ = 0.71) を達成し,次にGemini 2.5 Pro (56.9%, κ = 0.57) が続いた.
Mistral Medium 3.1は高い合意 (κ = 0.88) を示したが,低精度 (38.5%) を示し",自信のある幻覚"を示した.
クラウド・ソネット4.5は低精度 (33.8%) と一貫性 (κ = 0.33) を示し,不安定性を示した;骨折はすべてのモデルに挑戦的であった;人口統計的推論は貧しかった.

結論:

診断の正確性と一貫性は,別々のパフォーマンスメトリックであり,高い合意が正しさを保証するものではありません.
GPT-5 Proは,評価されたMLLMの中で,精度と安定性の最良のバランスを示しました.
現在のMLLMは実験的な診断推論システムであり,臨床骨折検出のための信頼できるスタンドアロンツールではありません.

キーワード:

さらに関連する動画

04:13

Less-Invasive Technique for Non-stabilized Mandibular Fracture in Mouse Models

Published on: September 27, 2024

891

03:57

Transverse Fracture of the Mouse Femur with Stabilizing Pin

Published on: December 29, 2021

4.2K

関連する概念動画

03:37

Uncertainty in Measurement: Accuracy and Precision

105.2K

Scientists typically make repeated measurements of a quantity to ensure the quality of their findings and to evaluate both the precision and the accuracy of their results. Measurements are said to be precise if they yield very similar results when repeated in the same manner. A measurement is considered accurate if it yields a result that is very close to the true or the accepted value. Precise values agree with each other; accurate values agree with a true value.

105.2K

01:16

Language

921

Language is a unique communication system that uses words and systematic rules to organize and transmit information. Unlike other forms of communication, which may involve postures, movements, odors, or vocalizations, language relies on symbols and grammar. This makes human communication distinct from that of other species, who also communicate but do not use language in the same way humans do.
Corballis and Suddendorf (2007) and Tomasello and Rakoczy (2003) highlight the role of language in...

921

02:07

Improving Translational Accuracy

15.0K

Base complementarity between the three base pairs of mRNA codon and the tRNA anticodon is not a failsafe mechanism. Inaccuracies can range from a single mismatch to no correct base pairing at all. The free energy difference between the correct and nearly correct base pairs can be as small as 3 kcal/ mol. With complementarity being the only proofreading step, the estimated error frequency would be one wrong amino acid in every 100 amino acids incorporated. However, error frequencies observed in...

15.0K

01:30

Heart Failure IV: Classification and Diagnostic Evaluation

429

Heart failure can be classified in various ways, with the most common classifications based on physical activity limitations, disease progression, severity, and treatment strategies.The Functional Classification of Heart Failure divides patients into four categories based on physical activity limitation due to symptom burden.Class I: Patients in this class have cardiac disease but no physical activity limitations. Ordinary activities like walking, climbing stairs, or routine tasks do not cause...

429

01:21

Peripheral Arterial Disease II: Clinical Manifestations and Diagnostic Evaluation

458

Clinical manifestationsPeripheral Arterial Disease (PAD) manifests through a range of symptoms, from the characteristic intermittent claudication to atypical presentations and severe complications in advanced stages. Intermittent claudication, a hallmark symptom of PAD, presents as exercise-induced muscle pain that typically resolves within minutes of rest. This pain is reproducible and stems from inadequate blood flow, leading to the accumulation of lactic acid produced during anaerobic...

458

01:30

Irritable Bowel Syndrome II: Clinical Features and Diagnostic Evaluation

900

Irritable Bowel Syndrome II: Clinical Features and Diagnostic Evaluation
Irritable Bowel Syndrome (IBS) is classified into subtypes based on the predominant bowel habits as determined by the Bristol Stool Form Scale (BSFS). The subtypes are:

900

関連する実験動画