A modality-based radiology protocol must state exactly which examination is the index modality, the scanner or equipment class, acquisition and reconstruction parameters, patient preparation, contrast or radiopharmaceutical use where relevant, image interpretation method, reader experience, blinding, and whether the unit of analysis is the patient, organ, lesion, vessel, examination or procedure. These details are not technical decoration: changing the protocol can change the measured value and sometimes the diagnosis.

Quantitative measurements need operational definitions. Apparent diffusion coefficient values depend on sequence and b-value selection; Doppler velocities depend on angle correction and sampling; elastography depends on platform and acquisition technique; computed tomographic attenuation depends on acquisition and reconstruction; and standardised uptake values depend on uptake time, calibration and reconstruction. A locally derived threshold should therefore be described as protocol-specific unless it has been externally validated under comparable conditions.

Diagnostic studies need an independent reference standard. Histopathology, surgery, endoscopy, microbiology, clinical follow-up and another imaging modality answer different questions and should not be substituted for one another without justification. If only imaging-positive patients receive the definitive reference test, verification is partial and apparent diagnostic performance may be biased.

A modality-based radiology synopsis can state the clinical problem, index imaging method, study design and principal outcome briefly. The full protocol must convert those statements into reproducible acquisition, measurement and interpretation rules. For example, “compare computed tomography and magnetic resonance imaging” is not sufficient unless it specifies whether the same patients undergo both examinations, the time interval between tests, how disease change between examinations is handled, and which independent reference determines the final diagnosis.

The protocol should also specify whether readers are blinded to the competing modality and reference standard, how technically inadequate studies are classified, how multiple lesions in one patient are handled, and whether quantitative values are measured once, averaged across regions of interest or independently assessed by more than one reader. These details determine whether the project measures prevalence, agreement, association or true diagnostic performance.

Sample size should be based on the primary endpoint rather than the number of examinations conveniently available. A prevalence study needs an expected proportion and desired precision; a diagnostic-accuracy study needs enough participants with and without the target condition to estimate sensitivity and specificity; an agreement study needs enough paired assessments; and a comparison of modalities should be powered for the expected paired or independent difference according to the actual design.

Paired imaging requires paired analysis. When the same patient undergoes two modalities, the measurements and diagnoses are linked and should not be analysed as though they came from unrelated groups. Conversely, comparing one modality in one patient population with another modality in a different population cannot establish modality superiority unless differences in case mix are addressed appropriately.

Multiple lesions, vessels or organs from one patient are clustered observations and require a predefined index-lesion rule or statistical methods that account for within-patient dependence. Quantitative imaging studies should report measurement reproducibility and confidence intervals, and threshold-finding studies should avoid presenting an internally optimised cut-off as universally applicable without independent validation.

Start with one modality that has dependable case volume, stable acquisition protocols and accessible clinical correlation in the department. Check whether the intended reference standard, reader expertise and required measurements are routinely available. A focused ultrasound, Doppler, computed tomography, magnetic resonance imaging, radiography, fluoroscopy, mammography, interventional or nuclear medicine question is usually more feasible than a project that depends on several modalities being performed in every participant.

Workable designs include observational and analytical cross-sectional studies, comparative studies, retrospective or prospective cohorts, diagnostic-accuracy studies, interobserver agreement studies, imaging-measurement studies and procedure-outcome studies. The design should follow the endpoint. Two radiologists reading the same images generate agreement data, while sensitivity and specificity require an independent reference standard. Outcomes that occur after imaging or intervention require a longitudinal design rather than being labelled cross-sectional.

Settle the index modality, exact patient population, primary endpoint, acquisition protocol, reader method, unit of analysis, reference standard, expected case volume and whether the required equipment will remain consistently available. If two modalities are being compared, decide whether the same patients will undergo both examinations, the acceptable interval between them and how changes in disease status between examinations will be handled.

Such a comparison is possible for descriptive service data, but it is weak evidence of diagnostic superiority because the modality groups may differ in age, disease severity, referral indication and pre-test probability. A stronger comparative diagnostic study applies both modalities to the same participants whenever clinically appropriate, or uses a carefully designed alternative that controls case-mix differences. The protocol should never imply that a difference caused by patient selection is an intrinsic property of the imaging modality.

The synopsis is the concise institutional submission describing the question, objectives and broad methods. The protocol is the operational document that specifies equipment, acquisition and reconstruction, patient preparation, contrast or radiopharmaceutical administration, reader blinding, measurement technique, reference standard, handling of technically inadequate studies, clustering of multiple lesions, missing data and the statistical method corresponding to every objective.

Ethics approval should state whether the research merely analyses standard-care imaging or adds any examination, sequence, contrast administration, radiopharmaceutical, fluoroscopic acquisition, intervention or venepuncture beyond clinical care. Research-only ionising radiation or contrast exposure should not be added simply to strengthen a resident thesis without specific justification, risk assessment and consent. Where children are included, guardian consent and age-appropriate assent from about seven years should be addressed. Retrospective PACS and record-based studies may qualify for a consent waiver under institutional requirements.

Anonymisation should remove identifiers from exported datasets and DICOM headers, and reconstructed facial surfaces from computed tomography or magnetic resonance imaging require particular care because they can remain identifiable after routine header removal. Separate consent is required for identifiable photographs or video. The protocol should also define the escalation route for an unexpected malignancy, aneurysm, acute vascular occlusion, significant infection or other actionable incidental finding so that the reporting radiologist and treating team can act on it promptly.

The sharpest error is claiming that one modality is more accurate or superior when the modalities were tested in different patient groups or against different reference standards. If magnetic resonance imaging is performed mainly in difficult cases, computed tomography in acute cases and ultrasonography in lower-risk referrals, the observed difference reflects case mix as well as technology. The same problem occurs when only positive imaging results proceed to biopsy or surgery. A valid comparison should use comparable patients, a prespecified acquisition protocol and the same independent reference standard wherever possible.

An MD synopsis is usually built around a focused imaging question that can be completed during residency with locally available equipment and patients. A PhD proposal normally requires a broader programme of work, a clearer knowledge gap, stronger methodological novelty and often external validation. Beyond the Indian MD pathway, research requirements differ by programme. The Saudi Board and SCFHS pathway includes a mandatory board research project, while Arab Board of Health Specializations programmes and DHP, DHA, DOH and MOHAP-linked training programmes may require research, audit or other scholarly activity according to the relevant speciality and institution. The applicable board and institutional requirements should therefore be checked before using an Indian MD synopsis unchanged in another training pathway.

Register after the guide has confirmed the question, modality, equipment availability, acquisition protocol, primary endpoint, unit of analysis, reference standard and expected case volume, but before prospective recruitment or research-specific data collection begins. For retrospective work, fix the study period, image-quality criteria, measurement rules and analysis plan before reviewing outcomes so that cases are not selected according to results already known to the investigator.