Number of Slices

MRIninja Knowledge Base | MRI Parameter Deep Dive Version 1.0 — August 2026

1. Introduction and General Purpose

The number of slices is the parameter that, together with slice thickness and inter-slice gap, determines whether a protocol actually covers the anatomy it was prescribed to cover. It is easy to treat as a purely administrative afterthought — "however many it takes to cover the region" — but it is in fact one of the most consequential parameters in protocol design, because it interacts directly with TR, SAR, acquisition time, and (in 2D acquisitions) the very feasibility of fitting the prescribed coverage into a single acquisition pass.

This page is deliberately distinct from the companion Slice Thickness deep dive. Slice thickness answers "how much tissue does each slice represent?" Number of slices answers a different question: "how many of those slices are required, and can the sequence actually deliver that many within the time and energy constraints of a single TR — or does the protocol need to be split?" The two parameters combine multiplicatively to determine coverage (Section 2), but they diverge sharply in their downstream consequences: slice thickness primarily affects SNR and partial-volume behaviour, while number of slices primarily affects acquisition time, TR feasibility, and SAR accumulation.

Why this parameter is under-appreciated: on most consoles, the number of slices is not typed in directly by the operator in modern practice — it is usually the automatic output of dragging the coverage box over the anatomy on the localiser, with the scanner calculating N_slices = coverage / (thickness + gap) in the background. This makes it easy to lose sight of the fact that a decision has been made, and that this decision has consequences (TR auto-extension, need for concatenations/packages, SAR budget) that are not always surfaced clearly to the operator until after the fact.

Historical evolution: the earliest clinical multi-slice MRI protocols (early–mid 1980s) acquired a handful of thick slices (10–15 mm) per examination, largely because SNR and gradient hardware could not support more. The maturation of 2D multi-slice interleaved acquisition through the late 1980s allowed 15–25 slices per package to become routine. Simultaneous multi-slice (SMS) excitation, introduced experimentally in the mid-2000s and made clinically practical from 2012 onward with the blipped-CAIPI method, fundamentally changed the economics of slice count for EPI-based sequences (DWI, fMRI, DSC) by decoupling scan time from slice number in a way that had never previously been possible [3].

2. Physical Foundations

2.1 The Coverage Equation

For a 2D multi-slice acquisition, anatomical coverage (C, in mm) along the slice-select direction relates to slice thickness (Δz), inter-slice gap (g), and number of slices (N) by:

C = N × (Δz + g), therefore N = C / (Δz + g)

For a 3D acquisition, there is no gap concept in the same sense (Section 2.3 of the companion Slice Thickness page); coverage is instead:

C = N_z × Δz_partition

where N_z is the number of phase-encoded partitions in the slab direction. This is the first fundamental distinction this page must establish: in 2D, N_slices is a geometry-derived output that then constrains timing (Section 2.2); in 3D, N_z (often still labelled "number of slices" or "slices per slab" on the console, per the vendor terminology in Section 3) is itself a Fourier-encoded acquisition dimension that directly and linearly sets scan time (Section 7).

2.2 The 2D Multi-Slice TR Constraint — the "Slice Loop Time"

In a 2D multi-slice acquisition, all slices belonging to one acquisition pass ("package") must be excited and read out within a single TR. The time required to excite and spatially encode one slice is referred to in the technical literature as the slice loop time (T_slice); the maximum number of slices that fit into one TR is [4]:

N_max_per_package = ⌊ TR / T_slice ⌋

This is the single most important equation on this page. It means the number of slices a protocol can acquire in one pass is not a free choice — it is bounded by TR, and TR is itself usually chosen for contrast reasons (short for T1-weighting, long for T2/PD-weighting), not for slice-count reasons. When the anatomically required N (from Section 2.1) exceeds N_max_per_package, the protocol has only two options: increase TR (Section 6), or split the acquisition into multiple sequential passes — a function every vendor implements, under different names (Section 3) — each pass covering a subset of the required slices, with total scan time multiplying accordingly (Section 7).

2.3 Mathematical Foundations

Scan time relationship (2D, single package):

T_acq = TR × N_y × NSA / ETL

Note that N_slices does not appear in this equation as long as N ≤ N_max_per_package — this is the source of the frequently-misunderstood claim that "slices are free" in 2D MRI. They are free only up to the TR ceiling; beyond it, an additional multiplying factor (the number of packages/concatenations, P) must be introduced:

T_acq = TR × N_y × NSA × P / ETL, where P = ⌈ N_required / N_max_per_package ⌉

Scan time relationship (3D):

T_acq = TR × N_y × N_z × NSA / (ETL × R)

Here N_z (the "number of slices"/partitions of the 3D slab) appears explicitly and multiplicatively — every additional partition adds time directly, with no TR-ceiling step function. This is the second fundamental distinction between 2D and 3D behaviour of this parameter, and it is the reason 3D protocols are exquisitely sensitive to over-generous slab coverage (Section 15) in a way that 2D protocols, up to the TR ceiling, are not.

SAR relationship: for a 2D multi-slice acquisition, each of the N slices within a TR requires its own RF excitation pulse. To a first approximation, average power deposition per unit time scales with the number of RF pulses delivered per TR:

SAR ∝ N_slices × (flip angle)² × (RF pulse duty factor) / TR

Clinical meaning: unlike slice thickness — which the companion page describes as having only an indirect SAR relationship via TR — number of slices has a comparatively direct relationship with SAR, because every additional slice literally means another RF pulse fired within the same TR window. This becomes clinically binding at 3T, where SAR headroom is already narrower, and is a common reason a 3T protocol auto-extends TR (with downstream T1-contrast consequences, Section 6) purely because the operator increased slice count.

3. Units, Terminology and Vendor Nomenclature

Number of slices is a dimensionless integer count. The function that splits an over-large slice request into multiple sequential passes, however, carries different names on every platform — a frequent source of cross-vendor protocol-translation confusion:

Vendor terminology confirmed from an actual Philips console (Advanced Parameters, Geometry tab): on Philips systems, slices are organised under a Stack, with the following console fields directly relevant to this page: Stacks → type (e.g. "parallel"), slices (the slice count, e.g. 44), slice gap → gap (mm) ("user defined", e.g. 0.4 mm), slice orientation (e.g. "transverse"), fold-over direction and fat shift direction (governing where wrap and chemical-shift artefacts fall relative to the slice stack, not the slice count itself but co-located on the same console page), and slice scan order (e.g. "HF" — head-to-foot acquisition order). The summary panel on the same Philips screen also displays Act. TR (ms) and TSE factor alongside these fields — a direct, visible illustration of the TR–slice-count coupling described in Section 2.2: the "Act. TR" shown is the value the scanner has already resolved as sufficient to fit the prescribed number of slices and TSE factor into one package.

"Slices" vs "locations": GE terminology historically distinguished "slices" (2D) from "locations" (a more geometry-neutral term also used for 3D and for multi-station acquisitions); both refer to the same underlying count for the purposes of this page.

4. Typical Value Ranges

4.1 Number of Slices by Anatomical Region and Sequence Type

4.2 Context by Field Strength

5. Parameter Interaction Ecosystem

5.1 Parameter Relationships Matrix

6. Effects on Image Appearance

6.1 Increasing Number of Slices (Fixed Thickness)

Increasing N at fixed slice thickness and gap directly extends anatomical coverage — this is usually the entire point of the change, and it does not, by itself, alter the appearance of any individual slice's contrast or resolution. The image-appearance consequences are therefore indirect, arising from whatever TR change or package-splitting the scanner has had to introduce to accommodate the extra slices (Sections 2.2, 6.3).

6.2 Decreasing Number of Slices (Fixed Thickness)

Decreasing N at fixed thickness reduces coverage — the most common cause of an inadvertently incomplete study is a slice-count reduction (often made to save time) that no longer reaches the full extent of the anatomy of clinical interest, an error that is not visible on the images themselves but only on review of coverage against the localiser/scout.

6.3 Indirect Appearance Effects via Forced TR Change

When an increase in N pushes the protocol past N_max_per_package (Section 2.2) and the scanner responds by auto-extending TR rather than adding a package, the contrast weighting of the sequence can shift measurably: a nominally "T1-weighted" TSE with TR auto-extended from 500 ms to 700 ms to accommodate extra slices will show reduced T1-weighting (longer TR allows more T1 recovery in all tissues, narrowing the T1-based signal differences the sequence was designed to display). This is one of the most clinically important — and most frequently unnoticed — indirect effects on this page.

7. Effects on Acquisition Time

7.1 2D Acquisitions — the Step Function

As established in Section 2.3, 2D acquisition time is flat with respect to N up to N_max_per_package, then jumps discretely with each additional package/concatenation required. A practical worked example:

TR = 4000 ms, T_slice ≈ 130 ms (standard TSE) → N_max_per_package = ⌊4000/130⌋ = 30 slices.

• N = 28 required → 1 package → no time penalty beyond the base T_acq
• N = 31 required (just one slice over the ceiling) → 2 packages required → T_acq doubles

This single-slice cliff-edge is the most important practical lesson of this entire page: protocol coverage should be planned with deliberate margin below N_max_per_package, not right up against it, precisely because crossing the ceiling by even one slice is disproportionately costly.

7.2 3D Acquisitions — the Linear Cost

There is no ceiling or step function in 3D — every additional partition adds time linearly, per the T_acq equation in Section 2.3. A 3D MPRAGE at N_z = 176 partitions and one at N_z = 208 partitions differ in scan time by the same proportion as their partition counts differ (all else equal), which is why 3D slab planning (generous enough to avoid wrap, but not wastefully larger than necessary) is a routine, continuous optimisation rather than a discrete threshold decision.

7.3 Breath-Hold-Specific Constraint

For breath-hold body sequences, the practical ceiling on N is frequently set by patient breath-hold tolerance rather than by the TR-slice-loop-time relationship alone: if the achievable N_max_per_package at the desired TR would require a breath-hold longer than the patient can sustain, the protocol must reduce N per breath-hold and split coverage across multiple breath-holds — functionally the same "packages" concept as Section 2.2, but driven by patient physiology rather than by the TR ceiling itself.

8. Effects on SNR and CNR

Number of slices has no direct effect on the SNR of any individual slice — SNR is governed by voxel volume, bandwidth, field strength, coil, and NSA (see the companion Slice Thickness and NEX/NSA pages), none of which N_slices itself alters. The relationship to SNR is entirely indirect, through two pathways: (1) if increasing N forces a package split, the freed time budget could alternatively have been spent on more NSA/higher SNR for the originally-planned coverage — a genuine opportunity-cost trade-off; (2) if increasing N forces TR extension, T1-weighted sequences will show altered CNR (as distinct from SNR) due to the contrast-weighting shift described in Section 6.3, even though voxel-level SNR may be essentially unchanged or even slightly improved by the longer TR.

9. Artefacts Associated with Number of Slices

10. Behaviour Across Sequence Families

Spin Echo (SE) / Turbo Spin Echo (TSE): classic 2D multi-slice behaviour exactly as described in Sections 2–7; T_slice is dominated by the echo train duration, so high-ETL T2-weighted TSE (long TR) tolerates large N per package far more easily than low-ETL, short-TR T1-weighted TSE (Section 17).

Gradient Echo (GRE/FLASH, 2D): short TR by design (often 100–300 ms) sharply limits N_max_per_package for 2D GRE; multi-slice 2D GRE protocols with substantial coverage frequently require multiple concatenations/packages, which is one reason clinical practice has shifted heavily toward 3D GRE (Section below) for large-coverage GRE applications.

Inversion Recovery (STIR, FLAIR, MPRAGE): for 2D STIR/FLAIR, the IR preparation pulse and its TI wait period add directly to T_slice, reducing N_max_per_package at matched TR relative to a non-IR sequence (Section 17, Example 3). For 3D MPRAGE, N_z is a Fourier-encoded partition count exactly as described in Section 7.2 — the single inversion pulse prepares the entire 3D slab, not each partition individually, which is part of why 3D IR-prepared sequences achieve large N_z efficiently.

EPI (DWI, fMRI, DSC): single-shot EPI acquires an entire slice's k-space in one very short readout train (tens of milliseconds), making T_slice intrinsically small; without SMS, N_max_per_package is nonetheless bounded by the TR chosen for diffusion-weighting/T1-recovery reasons (commonly 3000–8000 ms for DWI), which is usually generous enough to fit whole-brain coverage in one package even without acceleration. With SMS/multiband, the effective ceiling multiplies directly by the SMS factor [3], which is the primary reason modern high-resolution fMRI and diffusion protocols can acquire 60–90+ slices at sub-second TR (Section 17, Example 4).

bSSFP (TrueFISP/FIESTA/b-FFE): very short TR (3–6 ms) means the 2D multi-slice N_max_per_package formula becomes impractical for meaningful coverage; bSSFP for volumetric coverage is therefore almost always implemented as 3D (or as 2D single-slice/single-shot real-time acquisitions), sidestepping the 2D slice-loop-time ceiling entirely.

3D sequences generally (SPACE/CUBE/VISTA, MPRAGE/BRAVO/TFE, 3D DCE): governed exclusively by the linear N_z relationship of Section 7.2; parallel imaging (R) and, increasingly, compressed sensing are the primary levers used to keep large-N_z 3D acquisitions clinically practical in time.

11. Field Strength Behaviour

The 2D TR-slice ceiling (Section 2.2) is not itself field-strength-dependent — it is a timing relationship. What changes with field strength is the SAR pressure that pushes protocols toward either a lower achievable N per package or a longer TR at 3T and above (Section 2.3, Section 4.2), and the parallel-imaging/SMS headroom available to offset the linear 3D time cost (Section 7.2), which generally improves at higher field due to higher intrinsic SNR supporting higher acceleration factors before the g-factor penalty becomes limiting.

12. Vendor-Specific Implementation

Siemens: the "Concatenations" parameter is directly exposed and editable; when the requested slice count exceeds what fits in the current TR, the console typically prompts the operator to either increase TR or increase Concatenations. The g-factor-aware SMS (Simultaneous Multi-Slice) implementation is integrated into Siemens EPI-family sequences (DWI, fMRI) as a directly selectable acceleration factor.

GE: the equivalent function is termed "Acquisitions" on the protocol card; GE's HyperBand SMS implementation provides the multiband acceleration for EPI-family sequences.

Philips: termed "Packages" [1]. As confirmed directly from the console screenshots reviewed for this page, Philips organises slice geometry under a Stack (Geometry tab: Stacks → type, slices, slice gap, slice orientation, fold-over direction, fat shift direction, slice scan order), with the resolved "Act. TR (ms)" displayed live in the summary panel once the slice count and other geometry parameters are set — a direct, visible expression of the TR-slice coupling described in Section 2.2. Philips MultiBand is the corresponding SMS implementation for EPI-family sequences.

Canon: multi-slab/multi-acquisition splitting is available under equivalent, differently-labelled controls on the protocol card; Canon's SMS-equivalent acceleration is implemented for EPI-family sequences on compatible platforms.

Hidden coupling — all vendors: on every platform, increasing slice count near the TR ceiling can silently trigger either a TR increase or a prompt to add a package/concatenation; the operator-facing behaviour (silent auto-adjustment vs explicit prompt) differs by vendor and by software version, which is precisely why the "check TR after changing slice count" pearl in Section 16 is universal rather than vendor-specific advice.

13. Practical Optimisation Strategies

13.1 Clinical Optimisation Recipes

Plan coverage from the clinical question backward, not from a default box size forward: define the true anatomical extent required (with margin) first, then let N be the derived output of coverage ÷ (thickness + gap) — rather than starting from a habitual "standard" slice count and hoping it happens to cover the anatomy.

Deliberately stay below the TR ceiling, not right at it: given the step-function cost of crossing N_max_per_package (Section 7.1), a protocol design that requires 29 slices when the ceiling is 30 is fragile — a single additional slice needed at the scanner (patient larger than expected, extra coverage requested) will double the scan time. Building in a small margin, or accepting a marginally longer TR from the outset, is usually the more robust choice.

Use SMS/multiband wherever the sequence family supports it and coverage is large: for EPI-family sequences with substantial slice counts (whole-brain DWI, fMRI, DSC), SMS is close to a "free" multiplier on the achievable N_max_per_package and should be enabled by default where the coil and reconstruction pipeline support it [3].

Prefer 3D over multi-package 2D for large-coverage, non-time-critical structural imaging: if a 2D protocol would require 2 or more packages/concatenations to achieve full coverage, a 3D acquisition of equivalent total scan time frequently provides better SNR efficiency, no cross-talk, and isotropic reformatting — the 2D-vs-3D deep dive develops this trade-off in full.

14. Parameter Extremes

14.1 Minimal Slice Count (Single-Slice / Real-Time)

Single-slice, repeated acquisitions (N=1, acquired many times over) are used for real-time imaging (cardiac real-time bSSFP, interventional MRI guidance, swallowing/speech dynamic studies) where the clinical question is temporal dynamics at one anatomical level rather than volumetric coverage. The TR-slice ceiling of Section 2.2 is not a meaningful constraint here since there is only one slice to fit into each TR.

14.2 Very High Slice Count (Whole-Body and Dense 3D Protocols)

Whole-body multi-station STIR/DWI protocols (Section 4.1) can accumulate 250–350+ total slices across all stations in a single examination; dense 3D structural protocols (whole-brain 1 mm isotropic MPRAGE, whole-breast 3D DCE) routinely exceed 150–200 partitions. At this extreme, total examination time, SAR accumulation across the full protocol (not just per sequence), and patient tolerance become the binding constraints rather than any single-sequence TR ceiling.

15. Common Optimisation Errors

16. MRI Technologist Pearls

Check the TR immediately after any slice-count change — this is the single most important habit this page can instil. If TR has changed, ask whether the resulting contrast weighting is still what was intended, and whether the resulting scan time is still acceptable, before proceeding.

Know your scanner's ceiling behaviour: some platforms silently auto-extend TR when slice count exceeds N_max_per_package; others explicitly prompt for a package/concatenation increase. Knowing which behaviour your specific console defaults to for a given sequence prevents surprises.

A one-slice cliff edge is real: when a protocol sits close to the TR ceiling (Section 7.1), treat the next slice-count increase as a scan-time decision, not a trivial coverage tweak — verify the resulting scan time before committing.

For EPI-family sequences with large coverage, SMS/multiband should be the default, not an occasional extra: whole-brain DWI and fMRI protocols without SMS enabled, on a platform and coil that support it, are very often leaving free scan-time (or achievable slice count) on the table.

For multi-station whole-body protocols, plan station overlap deliberately: a small deliberate overlap of slices between adjacent stations avoids a coverage gap at the station boundary, at a small, predictable slice-count cost per station.

17. Real Clinical Examples — Variation Across Weightings

Example 1: T1-Weighted TSE Brain — Short TR Forces Concatenations

Protocol: axial T1 TSE brain, TR=550 ms (chosen specifically for T1-weighting), T_slice ≈ 110 ms (short ETL) → N_max_per_package = ⌊550/110⌋ = 5 slices.

Problem: whole-brain coverage at 5 mm slices with a small gap requires approximately 24–28 slices — far beyond the single-package ceiling of 5.

Resolution: the scanner automatically splits the acquisition into 5–6 concatenations/packages, each covering roughly 5 slices, executed sequentially. Total scan time is the base single-package time multiplied by the number of packages (Section 7.1) — this is precisely why short-TR T1-weighted 2D sequences with large coverage take noticeably longer, slice-for-slice, than long-TR T2-weighted sequences of the same nominal resolution.

Lesson: this is the weighting where the TR-slice ceiling of Section 2.2 is most aggressively binding, because the TR that gives good T1-weighting is also the TR that most severely limits N_max_per_package.

Example 2: T2/PD-Weighted TSE Spine — Long TR, Single Package

Protocol: sagittal T2 TSE lumbar spine, TR=3500 ms, T_slice ≈ 130 ms → N_max_per_package = ⌊3500/130⌋ = 26 slices.

Coverage need: 11–15 sagittal slices for standard lumbar spine coverage — comfortably within the single-package ceiling.

Lesson: the long TR that T2/PD-weighting requires anyway "happens" to also provide generous slice-count headroom, which is why concatenation/package problems are comparatively rare for standard T2-weighted spine and knee protocols, but common for T1-weighted ones at equivalent coverage.

Example 3: STIR (Limb/Whole-Body) — the Inversion Pulse Tightens the Ceiling

Protocol: coronal STIR of the thigh, TR=4000 ms (long, as is typical for STIR) — but each slice's loop time now includes the 180° inversion pulse and the TI wait (≈150 ms at 1.5T) before the excitation/readout, giving T_slice ≈ 130 + 150 = 280 ms → N_max_per_package = ⌊4000/280⌋ = 14 slices — roughly half the ceiling of a non-IR T2 TSE at the same TR (Example 2's 26-slice ceiling at a comparable TR).

Clinical consequence: whole-thigh STIR coverage requiring 20+ slices at this TR will require 2 packages, even though the TR itself is long — a direct illustration of the Section 5.1 principle that IR preparation shortens the effective slice-count ceiling independent of how generous TR itself is.

Lesson: when planning STIR coverage for large regions (whole-body, long-bone protocols), budget for the inversion-pulse time cost per slice explicitly — do not assume the ceiling will match a same-TR non-IR sequence.

Example 4: Single-Shot EPI (DWI/fMRI) — SMS Effectively Removes the Ceiling

Protocol (no SMS): whole-brain BOLD fMRI, TR=1990 ms, T_slice ≈ 20 ms (single-shot EPI is very fast per slice) → N_max_per_package = ⌊1990/20⌋ ≈ 99 — already generous without any acceleration, since EPI's intrinsically short T_slice means the ceiling is rarely the binding constraint for standard whole-brain coverage; N=96 slices at 1.3 mm isotropic was acquired within this TR in a published protocol using GRAPPA=3 for in-plane acceleration [7].

Protocol (with SMS): the same class of acquisition at a much shorter TR=589 ms and 2.5 mm isotropic voxels achieved N=48 slices using SMS acceleration factor 4 combined with in-plane GRAPPA factor 2 [7] — demonstrating that SMS is used less to overcome a slice-count ceiling that was already generous, and more to push TR itself down (improving temporal resolution for fMRI, or enabling more diffusion directions/b-values per unit time for DWI) while still comfortably covering the same anatomy.

Lesson: EPI-family sequences are the one weighting family on this page where the 2D TR-slice ceiling of Section 2.2 is rarely the limiting factor even without acceleration; SMS's primary value for slice count is enabling short-TR, high-temporal/angular-resolution protocols to retain full anatomical coverage, rather than rescuing an otherwise-infeasible slice count [3].

Example 5: 3D MPRAGE (T1-Weighted) — Linear N_z Cost, No Ceiling

Protocol: sagittal 3D MPRAGE, 1.0 mm isotropic, N_z=176 partitions (the ADNI standard) [6], TR=2300 ms (the 3D slab TR, not a per-slice TR in the 2D sense) — total acquisition time ≈ 4–9 minutes depending on in-plane acceleration.

Contrast with Example 1: this is also a T1-weighted acquisition, but because it is 3D, the short-effective-TR-per-partition constraint of Section 2.2 simply does not apply — N_z is a Fourier-encoded dimension (Section 2.3) that adds time linearly rather than triggering a package-splitting step function. Increasing N_z from 176 to 208 partitions (a 15% increase) increases scan time by approximately 15%, not by a discrete multiplying factor.

Lesson: the same T1-weighting that made Example 1's 2D acquisition so sensitive to the slice-count ceiling is, in 3D, essentially decoupled from that ceiling altogether — the practical reason most large-coverage, high-slice-count T1-weighted structural imaging on modern protocols has moved to 3D rather than multi-package 2D.

18. Visual Educational Material

18.1 The TR–Slice Ceiling Decision Flow

REQUIRED SLICES (from coverage/thickness/gap): N_required
AVAILABLE TR (set by desired contrast): TR
SLICE LOOP TIME (sequence-dependent): T_slice

N_max_per_package = floor(TR / T_slice)

Is N_required <= N_max_per_package?
│
├── YES → single package → base T_acq, no time penalty from slice count
│
└── NO  → choose one:
         │
         ├── Increase TR
         │     └── check: does this still give the desired contrast weighting? (Section 6.3)
         │
         ├── Add package(s) / concatenation(s)
         │     └── T_acq multiplies by number of packages (Section 7.1)
         │
         └── Switch to 3D (if sequence family allows)
               └── N becomes N_z; cost becomes linear, not step-function (Section 7.2)

18.2 2D (Step-Function) vs 3D (Linear) Time Cost of Adding Slices

2D, TR=4000ms, T_slice=130ms → ceiling = 30 slices

N=28: 1 package  → T_acq = T_base
N=30: 1 package  → T_acq = T_base   (still at the ceiling)
N=31: 2 packages → T_acq = 2 x T_base   ← one extra slice DOUBLES scan time

3D, TR=8ms per phase-encode step, N_z varies:
N_z=160 → T_acq = T_base
N_z=176 (+10%) → T_acq = 1.10 x T_base   ← smooth, proportional
N_z=192 (+20%) → T_acq = 1.20 x T_base

18.3 SMS/Multiband Effect on the Effective 2D Ceiling

Without SMS: N_max_per_package = TR / T_slice

With SMS factor M: M slices excited simultaneously per RF pulse
Effective N_max_per_package(SMS) ≈ M x (TR / T_slice)

Example: TR=600ms, T_slice=20ms → base ceiling = 30
  SMS=4  → effective ceiling ≈ 120
  SMS=8  → effective ceiling ≈ 240 (coil/g-factor permitting)

Trade-off: SMS factor is limited in practice by coil channel count
  and acceptable g-factor SNR penalty, not by TR/T_slice alone.

19. Evidence Gaps and Ongoing Debate

Optimal coverage margin below the TR ceiling: while this page recommends deliberate margin below N_max_per_package to avoid the single-slice cliff-edge cost of Section 7.1, no formal, prospectively validated guideline specifies exactly how much margin is appropriate for a given clinical context; current practice (as reflected in this page) is derived from engineering first principles and departmental experience rather than a dedicated comparative study.

Maximum clinically robust SMS factor by coil geometry: the relationship between coil channel count, target g-factor tolerance, and maximum recommended SMS factor is described qualitatively in the foundational SMS literature [3] but has not been comprehensively, prospectively mapped across the full range of clinical coil configurations and field strengths in current use.

Whole-protocol (not per-sequence) SAR accumulation from cumulative slice count: while per-sequence SAR limits are well standardised and enforced at the scanner level, the cumulative thermal/SAR burden across a full multi-sequence, high-total-slice-count examination (e.g. a comprehensive whole-body or multi-contrast brain protocol) is less thoroughly characterised in the prospective literature, particularly for patients scanned back-to-back with short table-time gaps.

Motion-robustness of multi-package (concatenated) 2D acquisitions versus single-pass 3D at matched total time: the anecdotal "slice jump at the package boundary" artefact described in Section 9 is well recognised in clinical practice, but there is limited systematic, prospective comparative data quantifying its frequency and diagnostic impact against an equivalent-duration single-pass 3D alternative across different anatomical regions.

20. Miscellaneous and Future Directions

AI-assisted automatic slice/coverage planning: automated landmark-detection tools (see the console-software discussion on the MRI Parameters Overview master page) increasingly propose not just slice thickness and orientation but full coverage boxes, implicitly setting N. As with automated slice thickness proposals, the radiologist and technologist retain responsibility for verifying that the AI-proposed coverage genuinely reaches the full extent of the structure of clinical interest before acquisition.

Ever-higher SMS factors for ultra-fast whole-organ dynamic imaging: research-stage SMS factors well beyond current clinical defaults are being explored to enable near-real-time whole-organ dynamic acquisitions (e.g. whole-liver or whole-brain sub-second dynamic contrast series); clinical translation depends on continued improvement in coil density and reconstruction robustness to interslice leakage.

Deep-learning-based package/concatenation reduction: emerging reconstruction approaches aim to relax the strict TR-slice-loop-time ceiling of Section 2.2 by reconstructing missing k-space/slice information rather than acquiring every slice within the formal TR constraint — an active but not yet clinically standardised area, conceptually related to, but distinct from, the deep-learning reconstruction techniques already discussed on the MRI Parameters Overview master page.

End of document — Number of Slices — MRIninja v1.0 — August 2026

Parent page: MRI Parameters — Overview and Classification (9501)