Radiomics and anaesthetic planning: Quantitative imaging as a new frontier in preoperative risk assessment

Table 1 Search strategy used for the narrative review

Database	Search string/field tags	Date range	Limits/filters
PubMed	[Radiomic* (tiab) OR radiomics (tiab) OR “quantitative imaging”(tiab) OR “texture analysis”(tiab)] AND [CT (tiab) OR “computed tomography”(tiab) OR MRI (tiab) OR “magnetic resonance imaging”(tiab) OR PET (tiab)] AND [body composition (tiab) OR sarcopenia(tiab) OR myosteatosis (tiab) OR visceral adiposity (tiab) OR liver (tiab) OR cardiac (tiab) OR pulmonary (tiab) OR renal (tiab) OR vascular (tiab) OR airway (tiab)] AND [perioperative (tiab) OR postoperative (tiab) OR surgical outcome (tiab) OR anaesthe* (tiab) OR anesthe* (tiab) OR “risk prediction” (tiab) OR “machine learning” (tiab)]	2012/01/01-2026/02/28	English language; humans
Scopus	TITLE-ABS-KEY (radiomic* OR “quantitative imaging”) AND TITLE-ABS-KEY (CT OR MRI OR PET) AND TITLE-ABS-KEY (“body composition” OR sarcopenia OR airway OR vascular OR cardiac OR pulmonary OR liver OR renal) AND TITLE-ABS-KEY (perioperative OR postoperative OR anaesthesia OR anesthesia OR “machine learning”)	2012-2026	English; article, review
Web of Science	TS = (radiomic* OR “quantitative imaging”) AND TS = (CT OR MRI OR PET) AND TS = (“body composition” OR sarcopenia OR airway OR vascular OR cardiac OR pulmonary OR liver OR renal) AND TS = (perioperative OR postoperative OR anaesthesia OR anesthesia OR “machine learning”)	2012-2026	English; core collection
Google scholar	Radiomics OR “quantitative imaging” AND perioperative OR anaesthesia OR anesthesia AND “body composition” OR sarcopenia OR airway OR vascular	2012-2026	Title/keyword search; first 200 results screened by relevance

Table 2 Classification of radiomic features with definitions and perioperative relevance

Feature category	Description	Key features	Perioperative relevance
First-order (histogram)	Statistical measures of voxel intensity distribution without spatial consideration	Mean, median, entropy, skewness, kurtosis, energy, uniformity	Tissue density characterization; hepatic steatosis grading; muscle quality assessment (myosteatosis)
Shape-based	Geometric properties of the segmented region	Volume, surface area, sphericity, compactness, elongation	Organ volumetry; tumour burden estimation; airway dimensional analysis
GLCM	Second-order texture features capturing spatial arrangements of pixel intensity pairs	Contrast, correlation, homogeneity, energy, entropy	Tissue heterogeneity; myocardial fibrosis detection; pulmonary parenchymal characterization
GLRLM	Texture features based on consecutive pixels of identical intensity	Run length non-uniformity, short/long run emphasis, gray-level non-uniformity	Vascular calcification pattern analysis; skeletal muscle architecture
GLSZM	Features characterizing zones of homogeneous intensity	Zone entropy, large/small zone emphasis, zone non-uniformity	Emphysema quantification; adipose tissue distribution
GLDM	Features describing pixel-pair intensity dependencies	Dependence entropy, dependence non-uniformity	Coronary plaque vulnerability assessment
NGTDM	Intensity differences between adjacent pixels	Coarseness, busyness, complexity, contrast	Tissue microstructure characterization; perivascular adipose tissue analysis

GLCM/GLRLM/GLSZM/GLDM/NGTDM: Texture feature matrices.

Table 3 Computed tomography-derived body composition parameters and their anaesthetic implications

Parameter	Measurement method	Diagnostic threshold	Anaesthetic/perioperative significance
Skeletal Muscle Index	L3 axial CT area/height2	< 52.4 cm2/m2 (M), < 38.5 cm2/m2 (F)	Reduced physiologic reserve; prolonged recovery; increased ICU stay; altered drug metabolism
Skeletal muscle density	Mean HU of skeletal muscle at L3	< 41 HU (BMI < 25), < 33 HU (BMI ≥ 25)	Myosteatosis; worse perioperative morbidity and mortality than sarcopenia alone
Visceral adipose tissue	Area within peritoneal cavity at L3 (HU: -150 to -50)	≥ 100 cm2	Increased volume of distribution for lipophilic anaesthetics; technical difficulty with neuraxial/regional blocks
Subcutaneous adipose tissue	Area outside muscle fascia at L3	Context-dependent	Needle length considerations for regional anaesthesia; injection site planning
V/S ratio	VAT/SAT	> 0.4 (metabolic risk)	Metabolic syndrome; increased cardiovascular perioperative risk; insulin resistance
Psoas Muscle Index	Bilateral psoas area/height2	Population-specific	Frailty surrogate; postoperative mortality predictor (HR 2.15)

CT: Computed tomography; HU: Hounsfield units; BMI: Body mass index; ICU: Intensive care unit; M: Male; F: Female; L3: Third lumbar vertebra; VAT: Visceral adipose tissue; SAT: Subcutaneous adipose tissue; HR: Hazard ratio.

Table 4 Cardiac radiomics applications relevant to perioperative risk assessment

Application	Imaging modality	Key radiomic features	Diagnostic performance	Perioperative relevance
Coronary plaque vulnerability	CCTA	Texture features (GLCM, GLRLM)	AUC 0.73 vs 0.65 (visual)	Identification of high-risk plaques; perioperative MI risk
Acute MI prediction	Cardiac CT	Kurtosis, short-run high gray-level emphasis	AUC 0.90	Perioperative cardiac event prediction
Myocardial pathology classification	Cardiac CT	Multiple texture features	Sensitivity 86%, specificity 81%	Subclinical cardiomyopathy detection
Cardiac amyloidosis detection	Cardiac CT	Radiomic signature	AUC 0.91 (external validation)	Restrictive physiology identification; fluid management
POAF prediction (CABG)	Non-contrast CT (EAT)	EAT texture and shape	Improved combined model	Postoperative arrhythmia risk; anticoagulation planning
POAF prediction (aortic stenosis)	CT (EAT)	Maximum grey-level, heterogeneity	AUC 0.80	Rate/rhythm control prophylaxis
MACE prediction	CCTA (PVAT)	Fat radiomic profile	ΔC-statistic 0.126	Long-term cardiovascular risk stratification
Acute MI vs stable CAD	CCTA (PCAT)	Pericoronary fat radiomics	AUC 0.87 vs 0.77 (attenuation)	Coronary inflammation assessment

CCTA: Coronary computed tomography angiography; AUC: Area under the curve; MI: Myocardial infarction; EAT: Epicardial adipose tissue; POAF: Postoperative atrial fibrillation; CABG: Coronary artery bypass grafting; PVAT: Perivascular adipose tissue; PCAT: Pericoronary adipose tissue; MACE: Major adverse cardiovascular events; CAD: Coronary artery disease.

Table 5 Machine learning algorithms for radiomics-based perioperative risk prediction

Algorithm	Mechanism	Strengths	Perioperative application examples	Typical AUC range
Logistic regression	Linear decision boundary for binary classification	Interpretable; established statistical framework	Surgical site infection prediction; transfusion requirement	0.70-0.82
Random forest	Ensemble of multiple decision trees	Handles high-dimensional data; resistant to overfitting	Sarcopenia detection; postoperative complication prediction	0.80-0.90
Support vector machine	Optimal hyperplane separation in feature space	Effective in high-dimensional spaces; robust with small samples	Myocardial pathology classification; plaque vulnerability	0.78-0.90
LightGBM/XGBoost	Gradient boosting ensemble methods	High accuracy; fast training; handles missing data	Postoperative gastric cancer complications; pancreatic fistula	0.84-0.93
Deep neural networks	Multi-layer non-linear feature learning	Automatic feature extraction; captures complex patterns	Airway difficulty prediction; cardiac event risk	0.80-0.95
Convolutional neural networks	Spatial feature learning from image data	Direct image input; no manual feature engineering	Airway segmentation; organ volumetry; body composition	0.85-0.96

Area under the curve ranges are indicative values compiled across heterogeneous studies and do not represent single validated estimates. AUC: Area under the curve; LightGBM: Light gradient boosting machine; XGBoost: Extreme Gradient Boosting.

Table 6 Comparison of traditional and artificial intelligence/radiomics-based airway assessment approaches

Assessment method	Sensitivity (%)	Specificity (%)	AUC	Advantages	Limitations
Mallampati score	20-62	82-97	0.55-0.65	Simple; bedside; no equipment needed	Low sensitivity; subjective; operator-dependent
Thyromental distance	25-50	80-95	0.60-0.70	Quick measurement; objective distance	Poor positive predictive value; single parameter
CT-based airway analysis	71-91	85-95	0.80-0.91	Objective 3D assessment; quantitative measurements	Requires CT; radiation exposure; not routine
Facial image deep learning	80-82	84-90	0.81-0.86	Non-invasive; rapid; smartphone-compatible	Requires validation across ethnicities; lighting dependency
Combined ML models	80-90	90-100	> 0.80	Integrates multiple parameters; outperforms individual tests	Limited emergency applicability; dataset-dependent

Reported performance ranges are pooled across heterogeneous studies and do not represent single prospectively validated estimates. CT: Computed tomography; AUC: Area under the curve; ML: Machine learning; 3D: Three-dimensional.

Table 7 Image Biomarker Standardisation Initiative standardization framework: Key phases and achievements

Phase	Timeline	Key achievements	Impact on reproducibility
IBSI phase 1	2016-2020	Standardized definitions for > 170 features; digital phantoms; benchmark values across multiple software platforms	Established common mathematical language; enabled cross-platform comparison
IBSI phase 2	Completed February 2024	Formalized preprocessing workflow (discretization, interpolation, filtering); expanded test phantoms for CT, MRI, PET pipelines	Entropy coefficient of variation decreased from 34% to 7%
Reporting guidelines	Ongoing	Transparent documentation of discretization, filtering, ROI handling, feature selection rationale	Improved study reproducibility and meta-analytic comparability
Compliance tools	Available	Online validation portal; excel-based reference sheets; benchmark feature values	Software platforms (PyRadiomics, LIFEx, SERA, RaCaT) can verify compliance

IBSI: Image Biomarker Standardisation Initiative; CT: Computed tomography; MRI: Magnetic resonance imaging; PET: Positron emission tomography; ROI: Region of interest.