DVC analysis metrics in the scientific literature: what each one measures

1. The sensing basis: a capacitance grid, not a camera

Every DVC metric ultimately comes from one raw signal, so it is worth being precise about that signal first. A sensing board carrying a grid of electrodes (a 12×4 layout in the standard configuration) is mounted under the floor of each individually ventilated cage; as the animal moves, its body changes the local dielectric properties and therefore the electrical capacitance measured at each electrode, sampled continuously and without cameras, tethers, or implants [1]. What the system records is not a position in pixels but a vector of per-electrode capacitance values, many times per second. Every metric discussed below is a transformation of that electrode-level time series — which is why understanding the grid is the key to reading the literature correctly [1].

2. Activation Density and Distance Walked: the core locomotion readouts

Activation Density and Distance Walked are the two foundational locomotion metrics: Activation Density is built from how many electrodes are activated over a time window, and Distance Walked is reconstructed from how the centroid of activation moves across the grid over time, with both validated against a conventional video tracking system [1]. In plain terms, more electrodes lighting up means more in-place or whole-cage motion, while a travelling centroid means displacement. These are the general-purpose activity indices most downstream disease- and drug-response studies are built on [1]. Because the metrics were benchmarked against video and shown to reproduce across multiple sites, the trade-off versus camera-based monitoring is one of spatial resolution, not validity [1][10].

3. The activity budget: distance, rest, and bout decomposition

Raw locomotion becomes interpretable once it is decomposed into a daily budget. In singly housed C57BL/6J mice, the raw DVC output decomposes the day into long rest (bouts of at least 40 seconds), short rest, local movement, and locomotion; mice spend roughly two-thirds of their time in long rest and travel about 330 m/day, predominantly in the dark phase, across some 7,100 discrete bouts [2]. This reference budget is what disease-model deviations are measured against — a drop in distance walked or a shift in the rest/activity split only means something relative to this baseline [2]. The dark phase is where most of the signal lives, which is precisely the window conventional daytime testing misses [2].

4. Rest Disturbance Index (RDI): a transdiagnostic rest-fragmentation biomarker

The Rest Disturbance Index quantifies the fragmentation or interruption of rest — capturing irregular activity that breaks up otherwise consolidated rest periods — and was first shown to be a sensitive, non-invasive digital biomarker in the SOD1G93A mouse model of ALS, without any EEG or EMG instrumentation [3]. What it indexes biologically is rest quality rather than rest quantity. A closely related active-phase rest metric reliably detects the excessive rest during the dark (active) phase seen in a myotonic dystrophy type 1 model, a non-invasive correlate of the hypersomnia reported in patients [4]. Because the same family of rest-disturbance signals recurs across ALS, myotonic dystrophy, narcolepsy type I, and aging colonies, it is best read as a transdiagnostic readout that generalises across models rather than a marker of any one disease [4][11][12].

5. Circadian and regularity metrics: day/night ratio, period, phase, amplitude

Treating the activity time series as a rhythm yields the classical chronobiology parameters — day/night activity ratio, circadian period, phase, and amplitude — which a DVC evaluation recovered reliably, including in clock-deficient cryptochrome-deficient mice, by using black cages with in-built LED lighting so each cage becomes an independent light-controlled chamber benchmarked against wheel running [5]. The practical consequence is that individualised light protocols and circadian phenotyping can run at rack scale without a dedicated light-tight room [5]. Biologically, these metrics index the integrity and entrainment of the circadian system rather than the amount of movement per se.

6. Spatial preference: frontality and wall activity

Because the signal is spatial, it also encodes where in the cage an animal spends its time. Metrics of spatial preference — including wall activity and "frontality," the tendency to occupy the front of the cage — are derived from which regions of the electrode grid are most active, and were shown to be strain-specific across C57BL/6N, BALB/c, and CD1 mice using only previously recorded data, with no new animals [6]. In Neuronautix's reading, these metrics index a behavioural-style and welfare dimension — disaggregation, exploration, thigmotaxis — that supports the argument for strain-specific cage-change frequency and density rather than one-size-fits-all husbandry [6].

7. Environmental channels are biomarkers too: BSI and the Urination Index

The most distinctive recent metrics do not measure the animal's movement at all — they measure its environment, from the same sensors. The Bedding Status Index is a machine-learning estimate of soiled-bedding "wetness," trained on human annotations of soiled bedding and validated across strains, ages, sexes, and densities; it reached over 90% accuracy at higher densities and supported data-driven cage-change intervals of three to six weeks — cutting cage changes by 65–70% without harming intracage ammonia, CO₂, growth, or circadian metrics [7]. The Urination Index applies the same bedding-moisture sensing to detect polyuria, and its bedding-moisture signal correlated strongly with blood glucose during hyperglycemia, enabling non-invasive, continuous in-cage detection of metabolic dysregulation and drug response [8]. The cross-cutting lesson is that "environmental" channels are biomarkers in their own right: the same raw capacitance signal serves husbandry optimisation and disease detection at once [7][8].

8. GYM500 running-wheel integration: when bulk activity is not enough

The optional GYM500 in-cage running wheel adds a voluntary-exercise channel whose running distance is logged alongside DVC activity; in a model of cancer-induced bone pain, wheel-running distance fell with the disease and tracked limb-use scores, while general home-cage activity did not — making wheel running the more sensitive spontaneous pain-like readout in that setting [9]. The design lesson is specific: bulk locomotion can miss focal or motivational phenotypes, so the wheel is worth adding when the phenotype of interest is voluntary effort rather than gross movement [9].

9. Reading these metrics correctly

Across the corpus, the metrics are strain-, sex-, and age-specific, so generic thresholds risk false signals and baselines must be matched to the cohort being studied [6][10]. None of these readouts is a black box: each is a defined transformation of the electrode-level capacitance series, with a published derivation and, in most cases, a validation against an established gold standard. In our view, the practical implication for anyone working with DVC data is to record, alongside each metric value, exactly which derivation produced it and which baseline it is being compared against — the same FAIR provenance discipline that makes any digital biomarker reusable.