🎗 Tomosynthesis 3D Mammography Detection Simulator
This simulation focuses on the detection of breast cancer using tomosynthesis (3D mammography). It helps in understanding how this advanced imaging technique improves the accuracy and sensitivity of breast cancer screening.
Conventional 2D Mammography — When Flattened Tissue Hides or Fakes a Lesion
Standard full-field digital mammography (FFDM) compresses the breast between two plates and captures a single 2D X-ray projection per view (craniocaudal and mediolateral oblique). This projection sums the attenuation of every tissue layer along the beam path into one flat image. In dense breasts, this superposition is the single largest limitation of screening mammography: normal fibroglandular tissue can stack up to create a dense white region that looks exactly like a mass, or conversely, a real tumor can be hidden behind or between layers of overlapping dense tissue.
- ~78%: 2D mammography sensitivity (drops further in dense breasts)
- ~10–12%: Recall rate (2D screening) (US screening population average)
- ~40%: Dense breast prevalence (of women screened in the US)
- up to 2×: Missed cancer rate (dense) (vs. fatty-replaced breasts)
Why flattening a 3D organ into 2D causes diagnostic ambiguity
A mammogram is a projection image: X-rays pass through the entire compressed breast thickness (typically 4–8 cm) and every voxel of tissue along that path contributes to the final pixel value. This is mathematically equivalent to summing a 3D density volume along one axis.
Two consequences follow directly:
• False-negative masking: a small cancer (5–10mm) embedded within or behind a region of dense fibroglandular tissue can be radiographically invisible — its subtle density difference is swamped by the surrounding normal tissue summed into the same projection.
• False-positive summation artifact: unrelated strands of normal fibroglandular tissue at different depths can coincidentally overlap in the 2D projection, creating a focal density that mimics the appearance of a spiculated mass or architectural distortion. This triggers a diagnostic workup (additional views, ultrasound, sometimes biopsy) for tissue that, when viewed at true depth, is entirely normal.
Radiologists refer to this dual failure mode simply as "tissue overlap" or "superimposition artifact" — it is the dominant reason 2D mammography underperforms in dense breast tissue (BI-RADS category C/D), which affects roughly 40% of women undergoing screening.
Because breast density itself is also an independent risk factor for developing cancer, the population most likely to have cancer masked by tissue overlap is also the population at elevated baseline risk — making this a compounding clinical problem, not just a technical nuisance.
Tomosynthesis Acquisition — Sweeping the X-Ray Source Through an Arc
Digital breast tomosynthesis (DBT) solves the projection-summation problem at the acquisition stage. Instead of firing a single X-ray exposure straight down through the compressed breast, the X-ray tube moves through a limited arc — typically 15° to 50° depending on the system — pausing at discrete angles (commonly 9 to 25 positions) to capture a low-dose projection image at each one. The detector and the compressed breast remain stationary throughout the sweep.
- 15°–50°: Typical sweep arc (vendor-dependent (narrow vs. wide angle))
- 9–25: Projection images per exam (discrete angular positions)
- very low: Per-projection dose (summed total ≈ one 2D exam)
- ~4 seconds: Acquisition time (full arc sweep, single breath-hold)
How the angular sweep is engineered and dosed
During a single compression and breath-hold (about 4 seconds), the X-ray tube head travels along a motorized arc above the patient while the detector panel below stays fixed. At each angular stop, a brief low-dose pulse is fired and the digital detector records one projection radiograph.
Because each individual exposure carries only a small fraction of a conventional 2D dose, the summed radiation dose across all projections in a tomosynthesis sweep is engineered to remain close to — and in combined-mode exams only modestly above — a single standard 2D mammogram, despite acquiring far more angular information.
Narrow-angle systems (~15°) prioritize in-plane spatial resolution and produce sharper slices but with more limited depth resolution; wide-angle systems (~40–50°) improve the ability to resolve structures separated in depth (better z-axis resolution) at some cost to in-plane sharpness. Detector readout must be fast enough to capture each projection before the tube moves to the next angle, and the whole sequence must complete before the patient needs to release compression or breathe.
Why multiple viewing angles recover depth information
A single 2D projection collapses depth entirely — there is no way to tell whether two dense structures seen in the image are actually touching or separated by centimeters. Multiple projections from different angles change this fundamentally: a structure near the detector shifts very little in the image as the tube angle changes (small parallax), while a structure near the X-ray source shifts a great deal (large parallax) — exactly like closing one eye and then the other to judge distance.
By capturing this angle-dependent parallax shift across 9–25 views, tomosynthesis acquires enough independent information to mathematically infer the depth (z-axis position) of every tissue structure in the breast — the raw material needed for slice reconstruction in the next stage.
Thin-Slice Reconstruction — Rebuilding the Breast in Depth
Once the full set of angled projection images has been acquired, reconstruction algorithms — most commonly filtered back-projection or iterative statistical methods — combine them into a stack of thin slices, typically about 1mm thick, running through the entire depth of the compressed breast. Instead of one flat image, the radiologist now has dozens of individual cross-sectional images that can be scrolled through like pages in a book.
- ~1 mm: Reconstructed slice thickness (typical in-depth resolution)
- 40–70: Slices per exam (per view) (depending on breast thickness)
- FBP / iterative: Reconstruction methods (filtered back-projection, SART, etc.)
- +30–50%: Reader time increase (more images to review per case)
From projections to a depth-resolved slice stack
Reconstruction is the computational inverse of acquisition: given a set of 2D projections taken from known angles, the algorithm estimates the 3D attenuation map of the breast that would have produced those exact projections, then renders that volume as a series of thin parallel slices.
Filtered back-projection (FBP) is the classic, fast approach: each projection is "smeared" back through the reconstructed volume along its acquisition angle, and the combined smears are filtered to sharpen structures and suppress blurring — structures present at the same depth in every projection reinforce each other, while structures only present in some projections (because they were shifted by parallax) partially cancel out, effectively "unstacking" the tissue that was superimposed in a single 2D view.
Iterative statistical reconstruction methods refine this further, modeling detector noise and X-ray physics more precisely at the cost of greater computation time, generally yielding sharper, lower-noise slices — increasingly standard on modern systems.
The outcome is a slice stack: instead of the tissue at every depth being summed into one pixel value, each ~1mm-thick slice shows primarily the structures that actually sit at that depth, with out-of-plane structures blurred into the background rather than sharply superimposed.
Because each slice only strongly represents tissue actually located near that depth, a lesion that was invisible in a flattened 2D projection — because it was buried inside a dense overlapping mass of normal tissue — can become clearly visible on the one or two slices where it actually sits, isolated from the confounding tissue above and below it.
Reduced Tissue Overlap — Better Detection, Fewer False Recalls
The direct clinical payoff of slice-by-slice viewing is a measurable reduction in the tissue-overlap problem described in Stage 1. Radiologists reading tomosynthesis stacks can visually "page through" depth and mentally exclude structures that only appeared suspicious because they were superimposed in a flat 2D view — while simultaneously being more likely to spot a genuine lesion that a 2D projection would have hidden.
- +1 to +2.7: Cancer detection rate change (additional cancers per 1,000 screened)
- −15% to −40%: Recall rate reduction (relative reduction vs. 2D alone)
- improved: Invasive cancer detection (larger relative gain than for DCIS)
- largest: Effect size in dense breasts (where overlap risk is highest at baseline)
Two independent benefits from the same mechanism
Reducing tissue overlap produces two distinct, clinically important effects that both trace back to the same underlying mechanism — depth separation:
1. Improved true-positive detection: a real lesion that was previously camouflaged by overlapping normal tissue in a 2D projection becomes visible as a discrete, well-defined finding on the slice(s) where it is actually located. This is most pronounced for architectural distortions and small masses, and especially valuable in dense breast tissue where 2D sensitivity is weakest.
2. Reduced false-positive recalls: apparent densities or "masses" that were purely an artifact of multiple unrelated tissue strands coincidentally overlapping in the 2D projection tend to resolve — that is, disappear or clearly reveal themselves as normal overlapping structures at different depths — when viewed slice by slice. Because these summation pseudo-masses were never real, tomosynthesis prevents many of the additional imaging visits, ultrasounds, and occasionally biopsies that 2D mammography alone would have triggered.
The net clinical effect reported across large screening studies is an increase in cancer detection rate combined with a simultaneous decrease in recall rate — a rare combination, since most changes that increase sensitivity (like lowering a diagnostic threshold) usually increase false positives too. Tomosynthesis avoids that trade-off because it is not changing a detection threshold — it is removing a structural source of ambiguity in the image itself.
Clinical Adoption — Tomosynthesis Combined with 2D Mammography
In routine practice, tomosynthesis is rarely used entirely on its own. Most protocols pair the 3D slice stack with a standard 2D image — either an additional low-dose 2D exposure acquired in the same compression, or a "synthesized 2D" image computationally generated directly from the tomosynthesis slice data, avoiding extra radiation. This combined-mode workflow has become the basis for a large and growing body of screening evidence.
- majority: Combined-mode adoption (US) (of screening facilities offer DBT)
- no extra dose: Synthesized 2D approach (derived from the same slice stack)
- ≈ up to 2×: Combined dose (DBT + true 2D) (a 2D-only exam, still low absolute dose)
- ≈ 1×: Combined dose (DBT + synthetic 2D) (comparable to 2D-only exam)
Why 2D and 3D are used together rather than 3D alone
The thin slice stack excels at resolving fine local structure and eliminating overlap ambiguity, but it can make it harder to appreciate the overall symmetry and large-scale architecture of the breast at a glance — something a single summary 2D image conveys efficiently. Combined-mode protocols keep both: radiologists typically scan the 2D image first for overall gestalt and obvious findings, then scroll the tomosynthesis stack to resolve ambiguous areas and confirm or exclude subtle findings in depth.
Two ways to obtain the 2D component:
• True acquired 2D: a separate standard 2D exposure taken during the same compression, at the cost of some additional radiation dose beyond the tomosynthesis sweep alone.
• Synthesized 2D (s2D): a 2D-like image computationally generated from the already-acquired tomosynthesis projections/slices, requiring no additional X-ray exposure. Modern combined-mode screening increasingly favors DBT + synthesized 2D specifically because it delivers dose comparable to conventional 2D-only mammography while retaining the full depth-resolved slice stack.
Growing evidence base and practical trade-offs
Multiple large prospective and retrospective screening studies have reported that combined DBT + 2D protocols detect more cancers per 1,000 women screened and recall fewer women for additional imaging compared with 2D mammography alone — with the largest relative gains observed in women with dense breast tissue, the group most affected by the tissue-overlap problem described in Stage 1.
Trade-offs that come with adoption include: longer radiologist interpretation time per case (more images to review), higher upfront equipment cost, larger data storage and archiving requirements, and — for true acquired 2D protocols — a modest increase in radiation dose versus 2D alone (largely avoided with synthesized 2D). None of these trade-offs have been large enough to offset the detection and recall-rate benefits in the populations studied, and tomosynthesis combined with 2D (increasingly synthesized 2D) has become a standard or preferred screening option at a large and growing share of breast imaging centers.
The consistent pattern across the evidence base — more cancers found, fewer women recalled unnecessarily — is precisely the outcome predicted by the underlying mechanism: separating tissue by depth removes a structural source of both missed cancers and false alarms at the same time, rather than simply trading one for the other.
This simulation focuses on the detection of breast cancer using tomosynthesis (3D mammography). It helps in understanding how this advanced imaging technique improves the accuracy and sensitivity of breast cancer screening.
2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install