---

PAPER IV - RECENT ADVANCES: MODEL ANSWERS

---

Q1 (30 Marks) - Triple Cut / Triple Rule-Out CT Coronary Angiography

Introduction

---

Principle

• Scan coverage (coronary CTA does not cover the aortic arch)
• Contrast timing (coronary CTA uses a saline flush to opacify only the left heart; this washes out the right-sided structures needed to evaluate PE)
• Heart rate control (ECG gating for coronary CTA)

---

Technical Protocol

• Beta-blockers given to achieve HR < 65 bpm (for coronary visualization)
• Nitroglycerin sublingual (for coronary vasodilation)
• Breath-hold instructions

---

CT Findings

• Normal: No stenosis, no plaque
• Soft plaque (low attenuation, <130 HU) - lipid-rich, unstable
• Calcified plaque (>130 HU)
• Mixed plaque
• Significant stenosis: >50% luminal narrowing
• Fractional Flow Reserve CT (FFR-CT) can be calculated from the dataset

• Pulmonary embolism: filling defect within pulmonary arterial lumen (saddle embolus, segmental/subsegmental PE)
• Pulmonary infarction: peripheral wedge-shaped opacity

• Dissection: intimal flap, true vs false lumen
• Intramural hematoma: crescentic high-density wall thickening
• Penetrating aortic ulcer
• Aortic aneurysm

---

Clinical Applications

1. Emergency Department triage of low-to-intermediate probability acute chest pain with normal/non-diagnostic ECG and troponin
2. Excludes all three major causes simultaneously - no further testing required in >75% of appropriately selected patients (Fuster's The Heart)
3. Useful when clinical picture is ambiguous (no clear history distinguishing ACS from PE from dissection)
4. Post-aortic valve replacement or TAVI planning that requires combined evaluation

• Higher radiation and contrast dose than dedicated studies
• Heart rate control mandatory for coronary quality
• Reduced coronary image quality compared to dedicated coronary CTA
• Not appropriate for high pre-test probability of a single diagnosis
• Renal impairment is a relative contraindication

---

Q2 (2 × 15 Marks)

---

2a. Elastography - Principle, Technique, and Applications

Principle

Techniques

• Mechanical compression applied manually via probe or by physiological motion (heartbeat, respiration)
• Pre- and post-compression images compared; displacement mapped
• Produces a relative (qualitative) elastogram - strain values are relative to surrounding tissue
• Used in: Breast, thyroid, liver, musculoskeletal, EBUS (endobronchial ultrasound)
• Elastic modulus calculated: E = stress/strain; stress assumed uniform across tissue block
• Semi-quantitative strain ratio: target tissue strain vs. normal reference tissue strain
• Grainger & Allison's Diagnostic Radiology states that in musculoskeletal elastography, strain elastography is the most common technique

• Acoustic Radiation Force Impulse (ARFI) uses focused ultrasound push pulses to generate shear waves in tissue
• Speed of propagating shear waves is measured: faster waves in stiffer tissue
• Provides absolute quantitative values: kPa (Young's modulus) or m/s (wave speed)
• Point SWE (pSWE / ARFI): single measurement
• 2D-SWE: real-time quantitative elasticity map
• More reproducible and operator-independent than strain elastography

• Mechanical vibrations (40-80 Hz) generated by external driver (passive driver on abdomen)
• MRI phase-contrast sequences detect propagating shear waves
• Gold standard non-invasive assessment of liver fibrosis
• Highly accurate even in early fibrosis (stages F1-F4)
• Also applied to brain, spleen, kidney, and breast

• Dedicated device with ultrasound probe combined with vibration source
• Measures liver stiffness in kPa
• 1D technique - not real-time imaging
• Widely used for non-invasive liver fibrosis assessment

Applications

---

2b. Functional MRI (fMRI)

Principle

• Deoxyhemoglobin is paramagnetic - it distorts the local magnetic field and reduces T2* signal
• Oxyhemoglobin is diamagnetic - no field distortion
• During neural activation: increased CBF → increased oxyhemoglobin → decreased deoxyhemoglobin → increased T2 signal*
• This signal difference = the BOLD response

MRI Sequences Used

• Gradient Echo EPI (Echo Planar Imaging): T2*-weighted; captures whole brain in 1-2 seconds; most common fMRI sequence
• Spin Echo EPI: Less susceptible to distortion artefact near air-tissue interfaces; more specific to capillaries
• Typical parameters: TE 30-40ms (3T); TR 1-2s; spatial resolution 2-3mm isotropic

Techniques / Paradigm Designs

• Alternating "ON" (task activation) and "OFF" (rest) blocks
• Statistical subtraction maps brain areas active during task
• Applications: preoperative mapping of eloquent cortex (motor, language, visual areas)

• Individual stimuli presented in randomized sequence
• Models HRF to brief events
• Higher temporal resolution; avoids adaptation effects

• Subject lies at rest; no task performed
• Spontaneous low-frequency BOLD fluctuations (0.01-0.1 Hz) analyzed
• Identifies resting-state networks (default mode network, visual network, motor network) - Rheumatology textbook
• Measures: ALFF (Amplitude of Low-Frequency Fluctuations), ReHo (Regional Homogeneity), functional connectivity

• Alternative to BOLD; magnetically labels inflowing blood water as an endogenous tracer
• Directly measures CBF in absolute units (mL/100g/min)
• No susceptibility artefact; better for longitudinal studies

Clinical and Research Applications

---

Q3 (2 × 10 Marks)

---

3a. FAPI PET vs FDG PET

FDG PET (Fluorodeoxyglucose PET)

• High physiological background in brain (glucose-dependent organ), liver, and GI tract
• Low sensitivity in mucinous tumors, well-differentiated tumors, prostate cancer, hepatocellular carcinoma
• Cannot distinguish post-treatment inflammation from residual tumor
• Low sensitivity for peritoneal carcinomatosis
• Requires fasting; blood glucose control needed

FAPI PET (Fibroblast Activation Protein Inhibitor PET)

Head-to-Head Comparison

Clinical Superiority Summary

• FAPI outperformed FDG in 23/34 studies for primary tumor detection and 24/34 for metastatic lesion detection
• Particularly superior for: gastric cancer, colorectal cancer, cholangiocarcinoma, pancreatic cancer, breast cancer, nasopharyngeal carcinoma
• FDG remains superior for: lymphoma, multiple myeloma, lung cancer (squamous/adenocarcinoma)
• FAPI changed management in 30% of patients

---

3b. K-Space - Definition, Uses, and Techniques in Radiology

What is K-Space?

• Centre of k-space: Contains low-frequency information → determines image contrast and overall brightness
• Periphery of k-space: Contains high-frequency information → determines image edge detail, sharpness, and fine structure

K-Space Filling Strategies and Their Clinical Uses

• Sequential filling from top to bottom (or center-out)
• Baseline technique for spin echo and gradient echo
• Full matrix acquisition

• Exploits the conjugate symmetry (Hermitian symmetry) of k-space
• Acquires only 5/8 or 6/8 of k-space lines; the rest are zero-filled or estimated
• Use: Reduces acquisition time; used in fast SE sequences, EPI
• Tradeoff: Slight SNR reduction

• Undersamples k-space by acquiring only every 2nd or 3rd line
• Uses multiple receiver coil array data to reconstruct missing lines
• Acceleration factor (R): 2-4×
• Use: Faster scan times, reduced motion artefact, cardiac imaging, body MRI
• GRAPPA (GeneRalized Autocalibrating Partial Parallel Acquisition): k-space reconstruction
• SENSE (SENSitivity Encoding): image-domain reconstruction

• Data acquired as radial spokes through centre of k-space
• Each spoke passes through centre → oversamples centre → robust to motion
• Use: Dynamic imaging, free-breathing abdominal MRI, pediatric imaging (motion-robust)
• Basis of PROPELLER/BLADE technique for motion-robust brain imaging

• Data acquired along spiral trajectories from centre outward
• Very fast acquisition
• Use: Cardiac perfusion, fMRI, real-time imaging
• Limitation: Sensitive to off-resonance (chemical shift/susceptibility) artefacts

• Fills entire k-space in a single RF excitation using rapid alternating gradients
• Acquisition time: 20-100ms
• Use: fMRI (BOLD), diffusion-weighted imaging (DWI), perfusion imaging
• Limitation: Susceptibility distortion, N/2 ghost artefact

• Dynamic acquisition: full k-space acquired once; subsequent time-points update only the central portion
• Central k-space = contrast; periphery assumed static
• Use: Dynamic contrast-enhanced MRI (DCE-MRI), MR angiography with high temporal resolution

• Highly undersampled incoherent k-space filling
• Missing data recovered by iterative reconstruction exploiting sparsity
• Use: Cardiac MRI (free-breathing, real-time), pediatric MRI (reduce scan time), 4D flow
• Acceleration: 4-8× over conventional

• Phase encoding in both y and z directions
• Full volume acquisition per TR
• Use: Volumetric brain imaging (MPRAGE), isotropic MRA, knee cartilage imaging

Summary Table

---

Q4 (2 × 10 Marks)

---

4a. Role of Radiology in 3D Printing in Orthopaedic Cases

Overview

Radiologist's Specific Roles

• Thin-slice CT (0.5-1mm slice thickness) with isotropic voxels for high-fidelity 3D reconstruction
• MRI for cartilage, ligamentous, and soft tissue detail
• Dual-energy CT for bone/soft tissue differentiation
• Optimization of dose while maintaining spatial resolution adequate for STL generation

• DICOM data is imported into segmentation software (Mimics, 3D Slicer, OsiriX)
• Radiologist defines region of interest: bone/joint surfaces, tumor margins, vascular anatomy
• Manual and semi-automatic thresholding; editing of artefact-affected regions (metal artefact from existing implants)
• Output: STL (Standard Tessellation Language) file - the digital blueprint for the printer

• Verify anatomical accuracy of the printed model against the original imaging
• Confirm critical dimensions (canal width, joint surface geometry, lesion margins)

Clinical Applications in Orthopaedics

• Physical 3D models of complex fractures (acetabular, tibial plateau, calcaneal, periarticular), tumors, and deformities
• Surgeon rehearses osteotomies, screw trajectories, and reduction maneuvers on the model before surgery
• Evidence: 82% of studies report improved surgical outcomes with 3D-printed models; reduced operative time; reduced blood loss; lower fluoroscopy time
• Particularly valuable in: acetabular fractures, pelvic osteotomies, complex spinal deformities, revision arthroplasty

• Custom-designed prostheses based on the patient's exact bony anatomy
• Applications: Massive bone defects after tumor resection, failed standard implants, complex revision cases
• Materials: Ti-6Al-4V titanium alloy (most common), cobalt-chrome, PEEK
• Porous surfaces (lattice structures) printed to optimize osseointegration
• Printing methods: Selective Laser Sintering (SLS), Electron Beam Melting (EBM), Direct Metal Laser Sintering (DMLS)

• Custom cutting jigs and drilling guides manufactured from the 3D model
• Ensures precise osteotomy angles, screw placement accuracy
• Reduces operating time and intraoperative fluoroscopy
• Used in: Total knee arthroplasty (TKA), total hip arthroplasty (THA), high tibial osteotomy, spine surgery

• 3D models of tumour extent (from MRI/CT) to plan resection margins
• Prefabricated reconstruction implants ready at the time of surgery
• Trial fitting of custom mega-prostheses before actual tumour resection

• Congenital deformities, post-traumatic malunion, correction osteotomies
• Virtual surgical planning (VSP) with simulation of corrective osteotomy
• 3D-printed guides lock onto bone to reproduce planned cut angle exactly

• Patient-specific models for resident education, surgical simulation, patient consent

Printing Technologies Used in Orthopaedics

Workflow Summary

---

4b. AI in Breast Imaging

Overview

Modalities and AI Applications

1. Computer-Aided Detection (CAD) / AI Detection
   • Traditional CAD: rule-based, high false-positive rate
   • Deep Learning CAD: convolutional neural networks (CNN) trained on millions of mammograms
   • Detects masses, microcalcifications, architectural distortion, asymmetries
   • Studies show AI achieves sensitivity and specificity comparable to or exceeding radiologists in standalone mode
   • Reduces miss rate ("second reader" role)
2. Breast Density Assessment
   • Automated AI-based BI-RADS density classification (A/B/C/D)
   • More reproducible than radiologist visual assessment
   • FDA mandated (September 2024) notification of breast density to all mammography patients
   • AI tools: Volpara, Hologic Quantra
3. Risk Stratification / Prediction
   • AI models predict short-term and long-term breast cancer risk from mammogram texture and density
   • Supplements Tyrer-Cuzick and IBIS models with imaging biomarkers
   • Identifies women who need supplemental screening (MRI, US)
4. Workflow Triage
   • AI pre-reads and triages normal mammograms (high NPV) to allow radiologist focus on suspicious cases
   • "AI-first" reading: normal cases safely deferred; reduces radiologist workload by 30-50% in some studies
5. Tomosynthesis (DBT)
   • AI synthesizes 2D equivalent images from 3D DBT volume
   • AI lesion detection across multiple DBT slices (not feasible manually in real-time without AI)

1. Automated Breast Ultrasound (ABUS) with AI
   • AI analyses 3D ABUS volumes for lesion detection
   • Valuable in dense breast supplemental screening
2. AI Characterization of Sonographic Lesions
   • CNN-based BI-RADS categorization of masses
   • Distinguishes benign (cysts, fibroadenoma) from suspicious lesions
   • Decision support for biopsy recommendation (BI-RADS 3 vs 4)
3. Elastography + AI
   • AI quantifies stiffness patterns; reduces interoperator variability

1. DCE-MRI Analysis
   • AI kinetic curve analysis: type I (persistent)/II (plateau)/III (washout)
   • Automated lesion segmentation and volume measurement
   • Treatment response monitoring in neoadjuvant chemotherapy
2. Background Parenchymal Enhancement (BPE) Assessment
   • Automated AI-based BPE quantification
   • High BPE associated with increased cancer risk
3. Contrast-Enhanced Mammography (CEM)
   • AI detection of enhancement patterns analogous to breast MRI

Specific Clinical Tasks Where AI Adds Value

Current Status (2024-2025)

• The EU AI Act (Aug 2024) classifies mammography AI as high-risk, mandating transparency, human oversight, and monitoring
• Multiple FDA-cleared AI tools (iCAD, Hologic Genius, Lunit INSIGHT)
• The AI-as-second-reader model (AI + one radiologist = equivalent to two radiologists) is being implemented in several European screening programs, addressing the radiologist shortage
• EUSOBI guidelines on AI in breast imaging updated; 2025 symposium "Mammography & Beyond: Special Focus on AI"

• Generalizability across different scanner vendors and populations
• Algorithmic bias (underperformance in dense breasts, non-White populations)
• Regulatory and liability frameworks still evolving
• Explainability (black-box models)
• Integration into existing PACS/RIS workflows

---