Biology · Observation, Imaging & Experimental Methods
Microscopy, imaging, resolution and artefacts
Two nearby bright objects can make one broad-looking patch in a recorded image. Enlarging that patch makes it easier to view without supplying a new observation of what is inside it. A biological image is evidence produced through a measurement system, and interpreting it begins with that distinction.
Research Design, Data & Scientific Practice introduces questions, comparisons, uncertainty and independent units. Cells, Membranes & Organisation supplies the cellular boundaries and compartments behind localisation and counting. Gene Expression: RNA & Proteins explains why DNA and RNA can act as templates and why a measured reporter product is distinct from its input. Ordinary fractions, averages and the distance between planar points support the worked cases. Follow each supplied inventory and comparison first; the formulas summarise that reasoning rather than replacing the biological conditions.
Back to the Biology mapBy the end, you can
Distinguish a specimen from its recorded image, compare detection, localisation, resolution and enlargement, and reason about background and lost information.
Follow the path from specimen to recorded values
A microscope forms an image from the way a specimen interacts with light or electrons. A detector then records selected information. The visible result can depend on the specimen’s preparation, thickness, optical conditions, detector and processing, as well as on its biology. A dark region can mean a weak signal, an unlabelled structure or information the method does not collect; darkness alone is not a universal absence test.
A digital image contains pixels, each with values representing a finite sampled area or volume. Display colour can be assigned to those values rather than being an organism’s natural colour. A scale bar relates positions to specimen distance, while an intensity scale relates displayed brightness to recorded values. Changing either display scale can change appearance without changing the underlying specimen.
Light microscopy, electron microscopy and different optical imaging arrangements answer different questions. Some observations concern live dynamics; others concern a prepared snapshot or a selected plane. A projection can place signals from different depths at the same image position. Useful interpretation therefore identifies what was imaged, which dimensions were represented and what the observation process could have changed.
Seeing a signal is different from separating two objects
Detection asks whether a signal can be distinguished from the relevant background and variation. Localisation asks where its source is estimated to lie. Resolution concerns distinguishing nearby features as separate. Magnification enlarges an image. These ideas are connected, but detecting a bright source does not establish its true dimensions, and locating a fitted centre precisely does not mean every nearby object is separately resolved.
The constructed profile uses a simple spreading rule rather than a real microscope formula. Each ideal input contributes one quarter of its value to the adjacent bin on either side and one half to its own bin. Adjacent input objects give a broad flat-topped recorded profile, whereas separated inputs give two distinct peaks under the same rule. The input and record retain different spatial descriptions.
A fourfold digital enlargement repeats the original values. It creates more display positions, not more independent detector samples. Acquiring different data can sometimes improve an inference, and a known response model can help interpret a record; neither possibility turns this toy blur into a universal optical resolution limit. Real resolution also depends on signal quality, sampling and the physical imaging system.
An enlarged image keeps the same original measurements
Read the diagram in words
Both inputs total 200 units. Adjacent bins 2/3 produce the profile 0, 25, 75, 75, 25, 0, 0, 0. Separated bins 2/6 give 0, 25, 50, 25, 0, 25, 50, 25. Repeating every recorded value four times produces 32 display values from the same eight measurements; it does not add information about the specimen.
| Offset in sample bins | Share contributed at this offset |
|---|---|
| -1 | 0.25 |
| 0 | 0.5 |
| 1 | 0.25 |
| Supplied input object | Original sample bin | Input position (µm) | Assigned intensity units |
|---|---|---|---|
| Assigned object 1 | 2 | 0.4 | 100 |
| Assigned object 2 | 3 | 0.6 | 100 |
| Original sample bin | Sample-bin centre coordinate (µm) | Recorded value after the assigned blur | Display repetitions per original value |
|---|---|---|---|
| 0 | 0 | 0 | 4 |
| 1 | 0.2 | 25 | 4 |
| 2 | 0.4 | 75 | 4 |
| 3 | 0.6 | 75 | 4 |
| 4 | 0.8 | 25 | 4 |
| 5 | 1 | 0 | 4 |
| 6 | 1.2 | 0 | 4 |
| 7 | 1.4 | 0 | 4 |
Distance between the two input positions (µm): 0.2; 200 = 200
| Supplied input object | Original sample bin | Input position (µm) | Assigned intensity units |
|---|---|---|---|
| Assigned object 1 | 2 | 0.4 | 100 |
| Assigned object 2 | 6 | 1.2 | 100 |
| Original sample bin | Sample-bin centre coordinate (µm) | Recorded value after the assigned blur | Display repetitions per original value |
|---|---|---|---|
| 0 | 0 | 0 | 4 |
| 1 | 0.2 | 25 | 4 |
| 2 | 0.4 | 50 | 4 |
| 3 | 0.6 | 25 | 4 |
| 4 | 0.8 | 0 | 4 |
| 5 | 1 | 25 | 4 |
| 6 | 1.2 | 50 | 4 |
| 7 | 1.4 | 25 | 4 |
Distance between the two input positions (µm): 0.8; 200 = 200
I(j) = Σ a(n)k(j − n)
Original one-dimensional teaching operator, eight sample-bin centres separated by 0.2 µm. Each of two ideal input objects contributes 100 assigned intensity units to its own bin before a fixed three-tap blur. This is not an optical point-spread function, Rayleigh resolution criterion, real diffraction limit or proof that a noiseless known operator cannot be inverted. A fourfold display repetition changes display size only; noise, saturation and boundary loss are absent.
A plausible appearance can arise from the measurement
An artefact is a feature or distortion introduced by the preparation, observation or processing that misrepresents the biological question being asked. Uneven illumination, unwanted fluorescence, channel leakage, specimen movement and a chosen segmentation threshold can alter apparent pattern. Light exposure can also change the specimen, including its fluorescent label. A declining image signal might reflect label bleaching rather than the loss of the labelled biological structure.
Recorded intensity includes the wanted signal together with background and measurement variation. Subtracting a justified background can address an additive offset, but it does not remove every source of noise or prove target identity. Detector saturation is a different problem: once a ceiling clips larger inputs to the same value, the original difference is no longer retained in that record.
In the cap example, subtracting ten background units recovers the two smaller supplied signals but not the largest. Repeating the capped measurement could give the same wrong underestimate every time. That illustrates why precision, meaning repeatability, differs from accuracy, meaning closeness to the relevant true quantity. Interpretation needs the measurement conditions and a question-specific comparison rather than confidence based solely on a sharp or attractive image.
Subtracting a background cannot recover a clipped signal
Read the diagram in words
Signals 20/60/120 plus background 10 would give 30/70/130. The ceiling changes the last record to 100. Subtracting 10 gives 20/60/90, so the last corrected value is below its supplied true signal 120. The capped record cannot reveal how far its input exceeded the ceiling.
| Supplied true signal | Signal plus background before clipping | Recorded detector value | Recorded value minus background | Recording ceiling reached before the full value was retained |
|---|---|---|---|---|
| 20 | 30 | 30 | 20 | × |
| 60 | 70 | 70 | 60 | × |
| 120 | 130 | 100 | 90 | ✓ |
Added background: 10; Maximum recorded value: 100
R = min(S + B,100); C = R − B
Original deterministic intensity account, arbitrary compatible detector units, nonnegative true signals, constant background 10 and recording ceiling 100. No noise or specimen change; values above the ceiling are clipped, not rescaled. The known true signals are teaching inputs, not quantities recoverable from the capped observations alone.
Worked example
Trace what enlargement and correction actually change
The eight sample bins are numbered zero through seven. The adjacent inputs at bins 2 and 3 total 200 units. Their recorded profile is 0, 25, 75, 75, 25, 0, 0, 0. At bins 2 and 6 the same total instead gives 0, 25, 50, 25, 0, 25, 50, 25. Both use the same spreading weights and retain the same total, while their shapes differ.
Repeating each of the eight original values four times gives 32 displayed values. A larger display may help viewing, but its repeated entries are not 32 new measurements. Inferring source number or separation requires the response assumptions and evidence, not the enlarged appearance alone.
For the separate detector account, true signals 20, 60 and 120 plus background 10 give 30, 70 and 130 before clipping. The ceiling retains 30, 70 and 100. Subtracting 10 gives 20, 60 and 90. The last observation does not disclose that its true input was 120; it only records a capped value.
Check your understanding
Does detecting a bright spot establish that the spot’s visible width is the object’s actual width?
Show reasoning
No. The visible signal can be spread by the imaging system and sampled into pixels, with additional effects from noise, depth and display. Detection establishes evidence for a signal under its conditions; dimensions require a justified spatial measurement and response interpretation.
Why do 32 repeated display values not provide more independent evidence than the original eight bins?
Show reasoning
Each group of four repeats comes from one same recorded value. Display enlargement changes representation rather than acquisition. It can make a feature easier to view, but it does not supply additional independent observations of the specimen or recover unrecorded detail.
Why does background subtraction fail to recover 120 in the capped-signal example?
Show reasoning
Signal plus background would be 130, but the detector retained only its ceiling of 100. Subtracting the known 10 therefore gives 90. The amount above the ceiling was not recorded, so arithmetic correction cannot retrieve that missing information from this value alone.
What to carry forward
Ask what the measurement system retained before interpreting an image as a biological object, amount or change.
References
Resources supporting this lesson’s explanations and examples.