Diagnostic Imaging Interpretation

X-ray (radiography): uses ionizing radiation; structures dense (bone) appear white, structures air-filled (lung) appear black. Fast, cheap, low radiation. Good for bones, pneumonia, foreign bodies. Limitations: 2D, overlapping structures obscure findings. CT (computed tomography): X-ray beam rotates around body, computer reconstructs thin cross-sectional slices; shows soft tissue detail better than X-ray and is sensitive to subtle density differences. Fast, excellent for trauma, cancer, internal organs. Radiation dose higher than X-ray. MRI (magnetic resonance imaging): uses magnetic field and radio waves, no ionizing radiation. Superior soft-tissue contrast (brain, spinal cord, ligaments). Slow, expensive, cannot be used with certain metal implants. Good for musculoskeletal injuries, neuro, pelvic pathology. Ultrasound: uses sound waves (no radiation), real-time, portable, cheap. Excellent for fluid collections, pregnancy, blood-vessel flow. Limited by gas (bone/lung obscure it) and user skill. All modalities are complementary; choice depends on clinical question, urgency, and radiation risk.

DICOM (Digital Imaging and Communications in Medicine) is the standard format for medical images: it stores not just the pixel data but also metadata (patient name, ID, study date, modality, series description). PACS (Picture Archiving and Communication System) is the hospital/clinic system that stores, retrieves, and displays DICOM images. PACS integrates with the RIS (Radiology Information System) that manages orders, reports, and billing. As an imaging technologist or interpreter, you interact with PACS daily: you retrieve prior exams for comparison, window and level images for optimal viewing, measure structures, annotate findings, and route reports to clinicians. PACS also manages image quality: it detects repeat exams (reducing radiation), stores images for legal/regulatory retention periods (7+ years), and enforces access controls (only authorized personnel see images). Understanding PACS workflow and standards is essential for efficient imaging operations and patient safety.

Artifacts are unwanted features that degrade image quality and can mimic or obscure pathology. Common X-ray artifacts: patient motion (blurring), patient rotation (distorted anatomy), underpenetration (image too dark, low kVp), overpenetration (image too light, high kVp), and metallic streak (from pacemakers, dental work). CT artifacts: beam hardening (streak or cupping from metal or dense bone), motion (blurring), windmill or star artifact (from motion of metal), cone-beam artifact at the edge of the field. MRI artifacts: motion (ghosts in the direction of acquisition), susceptibility artifact (from metal), chemical shift (misregistration of fat and water), aliasing (wraparound if field-of-view too small). Ultrasound artifacts: acoustic shadowing (dark area behind bone or stone), reverberation (ghost echoes), mirror image (reflection of anatomy to opposite side). Quality issues include poor positioning (patient not centered, rotated, or limb not fully included), poor inspiration (chest X-ray with patient inhaling shallowly, making lungs appear hyperinflated), improper window/level settings (making pathology invisible), and wrong patient or wrong study (never happened, but checks are critical). Recognizing and flagging artifacts prevents misdiagnosis and may require a repeat exam.

Windowing and leveling (also called window width and window level) are post-processing tools that optimize the display of CT, MRI, or digital X-ray images. The window width determines the range of pixel values (Hounsfield units in CT) displayed: a narrow window makes small differences in density visible (good for soft tissue, lung detail); a wide window shows a broader range (good for bone, metal). The window level (center) shifts the range: a higher level (e.g., +50 HU) makes denser structures visible (bone window); a lower level (e.g., -400 HU) makes less-dense structures visible (lung window). Example: a lung nodule on CT is barely visible in standard soft-tissue window; switch to lung window (narrow width, low level) and the nodule stands out. Mastery of windowing is essential: a radiologist or technologist who doesn't optimize windows will miss findings. Most PACS software has preset windows for common regions (bone, soft tissue, lung, brain, mediastinum); learning when to apply each preset and when to customize is a key skill.

A radiology report must be accurate, clear, and actionable. Structure: (1) indication (why the imaging was ordered), (2) technique (what modality, what was scanned, patient position, contrast used), (3) findings (organized by anatomy: describe normal structures first, then abnormalities in a logical order), (4) impression (summary of key findings and their significance, numbered for clarity), (5) recommendation (whether follow-up imaging is needed, how soon). Findings should use specific measurements (not 'large mass' but '3.2 cm mass in the right upper lobe'), location (anatomical landmarks), and comparison to priors ('new compared to prior exam'). Impression should answer the clinical question ('no acute fracture' or 'findings consistent with pneumonia'). Avoid vague or equivocal language if confident ('mass' not 'lesion-like finding'). Flag critical findings (stroke, PE, cancer) via phone to the ordering clinician immediately; do not wait for the report to be finalized. Communicate clearly with laypeople too: if a patient asks about their imaging, explain in simple terms and note any significant findings they should know about.

AI tools are increasingly deployed in radiology to improve efficiency, consistency, and detection. Common uses: (1) detection/flagging (identify nodules on CT, microcalcifications on mammography, retinal lesions on fundus photography), (2) measurement (automatically measure tumor size, vertebral body height), (3) classification (predict Benign vs. Malignant, Fracture vs. No Fracture), (4) segmentation (delineate organ boundaries or lesion contours for radiation therapy planning). AI can highlight areas of concern, routing urgent studies first and prioritizing worklists. However, AI is not perfect: it can miss rare findings, be fooled by artifacts or unusual anatomy, and can reinforce bias from training data. The gold standard remains human interpretation; AI is a tool to augment, not replace, radiologists and sonographers. Regulatory bodies (FDA, CE marking) are establishing standards for AI validation in medical imaging. Future radiologists must understand AI capabilities and limitations and be able to recognize when AI guidance is appropriate vs. when human judgment must override.

Entry-level: Radiologic Technologist (RT) or Sonographer requires a high school diploma and accredited 2-4 year program (often at community colleges), followed by ARRT or ARDMS certification exams. US: most states do not legally require certification, but hospitals almost always do. Salary: ~$60-70k. Advancement: after 1-2 years, pursue advanced certifications in specialties (CT, MR, Interventional, Nuclear) to increase salary to $75-85k. Some technologists pursue a bachelor's degree to become a Radiologic Scientist or move into QA/management. Doctorate: Radiologist is a physician (MD/DO) who completes 4 years of medical school and 5-6 years of radiology residency (including anatomy, physics, pathology, clinical interpretation, and research). Radiologists interpret images and manage imaging services; average salary $350-450k. Other advanced roles: Physician Assistant or Nurse Practitioner with radiology focus can read certain imaging (ultrasound, plain films) in some settings. International credentials vary; many countries have different regulatory paths (UK: Radiographer, France: Manipulateur). Certification requires continuing education to maintain current knowledge.