Real recorded data Synthetic parts labelled SYNTHETIC inside

Cancer as Coherence Collapse: A Reversible State Transition

Framework, Detection System, Verified Evidence and Research Program — Current Claims

Author: Paul Samuel Guarino
Framework: SEIS-UGFM Coherence Framework
Contact: [email protected]
Date: 22 October 2026
Status: Research white paper. Computational findings re-executed from raw public data on 3 October 2026. Pre-clinical: no patient has been diagnosed or treated under this framework.

Provenance key

Every claim in this paper carries one of four marks, so that a reader can see at a glance what kind of statement it is.

1. The claims

1. Malignancy is a loss of organization, and the loss can be measured in an image. — VERIFIED. Across five public datasets spanning prostate, leukemia, lymph-node histology and dermatoscopy — 10,015 dermatoscopic images in the latest run — malignant tissue is less spectrally ordered than healthy tissue. The direction is the same in every dataset. The magnitude is modest on histology (Cohen's d ≈ 0.2–0.8) and strong on morphometrics (AUC 0.895). Stated precisely: the metrics read spatial organization, and malignancy is the largest driver of its loss — basal cell carcinoma is the strongest spectral signal in the corpus, and a disorganized benign lesion reads as disorganized.

2. Cancer is a state transition between two stable regimes, and the transition is reversible inside a window. — HYPOTHESIS of this framework, now independently supported in gene-expression space. Shin et al. (KAIST, Advanced Science, January 2025) modeled colorectal tumorigenesis as an attractor landscape from patient-derived matched organoids and found the tumor transition state to be unstable and reversible, with a reversion switch that returns cells to the normal attractor.

3. The restorative lever is coherence, not killing. — HYPOTHESIS. The therapeutic principle of this framework is to re-establish the phase relationship that the healthy state keeps — to phase-lock cells back to their memory of the healthy state — rather than to destroy them. The KAIST reversion switch is the molecular form of the same principle.

4. Coherence is a relationship, not a note. — HYPOTHESIS. The 700/N harmonic cascade is a mutually coupled set of frequencies; 41.176 Hz = 700/17 is one member and the framework's candidate anchor. The pre-registered frequency sweep tests whether any specific frequency, including 41.176 Hz, governs the signature.

5. A detection system exists, and its image channel measures real images today. — DESIGNED, with an operating image channel. The Cancer Coherence Detection System (Clinical Edition v3.0, interface release v3.1 of 23 October 2026) and the Klein CDIS Workstation provide an image-analysis channel that runs the verified metric pipeline of Section 4 on an uploaded image, a bioelectric impedance channel, a biophoton channel, a combined Resonance Score, a frequency-sweep module, patient records, batch analysis, machine-learning hooks, and export and regulatory-package generators.

6. The nearest publishable experiment is the time course. — PROGRAM. Image the transition rather than the endpoint. If the spectral signature moves in the window where reversion is still possible, this system becomes the spatial readout — cheap, non-invasive, run on pixels — of the exact window that KAIST targets molecularly.

2. The framework

2.1 Coherence collapse

Healthy tissue maintains recurrent spatial organization — glandular architecture, nuclear spacing, stromal alignment — arising from coordinated cellular behaviour. Malignancy disrupts that coordination. If the disruption is real, it is detectable as a shift in spatial-frequency structure: organized tissue concentrates spectral energy; disorganized tissue disperses it. This is the directional prediction the framework makes, and it is testable on any labeled image dataset (Section 4).

2.2 One physics, two signs

Coherence, in this framework, is a relationship between oscillators — phase and tempo — never a single frequency. The same lever that builds order can break pathological order: where a disease is runaway synchrony, structured timing opposes it; where a disease is lost organization, as in cancer, structured timing restores it. The framework therefore treats cancer and its reversal as two signs of one physics. — HYPOTHESIS.

2.3 The 700/N cascade

The framework's physical anchor is the number 700, an observed input that lies within 0.036% of (mₑc² / E_Thomson) × (1/α). The cascade members 700/N (N = 16, 17, 18, …) are claimed to form a mutually coupled system: the framework is about the relationships between members — cross-frequency coupling, harmonic locking at integer ratios, mutual phase structure — and about the temporal stability of prime-indexed members. 41.176 Hz = 700/17 is one member. Its significance is membership in the cascade, not amplitude relative to its neighbours. — HYPOTHESIS.

2.4 Spatial–temporal translation

Images are not measured at 41.176 Hz; a fixed tissue section has no temporal frequency. In the image domain the 700/N ratios serve as an informing geometry — a comb of relative proportions on a normalized radial spatial-frequency axis — against which power concentration, dispersion and drift are measured. Temporal 41.176 Hz belongs to bioelectric and neural measurement, which is the role of the instrument channels in Section 5. The two domains are coupled by the hypothesis that the same organizational mathematics governs both; they are measured separately.

2.5 The therapeutic principle: restoration

An intervention that re-establishes the coherent relationship, applied in the window before the malignant state locks in, is predicted to move tissue back toward the healthy regime. In molecular terms this is what the KAIST reversion switch does in gene-expression space. In the physical terms of this framework it is entrainment — a structured field at cascade frequencies — whose effect on coherence metrics is the primary outcome of the frequency-sweep protocol (Section 7). — HYPOTHESIS.

3. Convergence with the attractor-landscape model (KAIST, 2025)

Shin, D., Gong, J.-R., Jeong, S. D., Cho, Y., Kim, H. P., Kim, T.-Y., Cho, K.-H. Attractor Landscape Analysis Reveals a Reversion Switch in the Transition of Colorectal Tumorigenesis. Advanced Science, 2025. DOI 10.1002/advs.202412503. The REVERT framework reconstructs a core gene-regulatory network from single-cell transcriptomes of patient-derived matched colon organoids (normal and tumor), maps the attractor landscape of the normal-to-cancer transition, and identifies transcription-factor interventions that push cells back to the normal attractor. The same laboratory published the attractor-landscape picture of colorectal tumorigenesis in 2016 (Cho & Cho, BMC Systems Biology).

Two independent programs, from opposite directions, describe the same structure: two stable regimes, a critical transition, and a window in which the transition can still be reversed. The convergence is in the shape of the model and in the therapeutic philosophy — restore the state rather than kill the cell. The readout spaces differ, and that difference is this program's contribution: KAIST can find the transition in expression space; this system is built to watch it in an image.

The mathematical bridge is the theory of critical transitions. As a dynamical system approaches a transition between attractors it shows generic early-warning signals — rising variance, critical slowing, spectral reddening, rising entropy (Scheffer et al., Nature 2009). The coherence metric set of Section 4 — noise coefficient of variation, Brightness Stability Ratio, spectral entropy, spectral flatness — is that family of statistics computed on tissue images. The verified endpoint data already show the early-warning pattern: at the malignant endpoint, entropy is up, stability is down, variability is up. What has not yet been measured by any program is the signature across the transition itself. That is the experiment in Section 7.

4. Verified evidence

All datasets are public; every figure below was re-executed from the raw files on 3 October 2026. Metrics computed per image: 2D-FFT power spectrum; spectral purity; spectral flatness; spectral entropy; bicoherence; brightness statistics including the Brightness Stability Ratio (BSR, in its mean²/variance and mean/CV formulations); gradient density. Classification by logistic regression with stratified 5-fold cross-validation. — VERIFIED.

4.1 Prostate — outcome coupling (Stanford prostate.data, n = 97). A three-feature model (log cancer volume, log prostate weight, seminal-vesicle invasion) predicts log-PSA with R² = 0.636. Patients with capsular breach show mean log-PSA 3.72 against 2.14 without. Boundary integrity, volume and the biochemical marker move together.

4.2 Prostate — malignant vs benign morphometrics (Kaggle, n = 100; 62 M / 38 B).

4.3 Leukemia — C-NMC (n = 1,867 cells; 1,219 ALL / 648 normal). Spectral and textural features separate leukemic from normal cells with small, consistent effects: best single metric spectral purity, d = 0.37, AUC 0.62; combined 11-feature model AUC 0.67.

4.4 Histopathology — PatchCamelyon (n = 4,000 patches; 2,000 / 2,000). Six coherence metrics yield d = 0.17–0.39, every one significant at n = 4,000 (p ≤ 1.6 × 10⁻⁷); the strongest are the resonance ratio (d = 0.39) and spectral entropy (d = 0.38). BSR is lower in cancer.

4.5 Skin lesions — HAM10000 melanoma vs nevus (n = 2,613; 1,113 / 1,500).

All four directional predictions — dimmer, noisier, lower BSR, higher entropy — hold simultaneously. On the histology-confirmed subset (n = 1,676), spectral flatness alone reaches AUC 0.742. The BSR reduction seen on PatchCamelyon replicates here: melanoma is about 44% lower than nevus on the mean²/variance formulation.

Extension to the full dataset — all 10,015 images, seven classes (computed 22 October 2026). The same frozen metric set was run on every image. Per class against nevus, on BSR (mean/CV):

Pooled malignant (mel + bcc + akiec, n = 1,954) against benign (nv + bkl + df + vasc, n = 8,061): combined AUC 0.647 (nine features, logistic regression, 5-fold). Basal cell carcinoma is the strongest spectral signal in the corpus. The result also sharpens the claim: dermatofibroma — benign but structurally chaotic — separates from nevus as strongly as carcinoma does, while benign keratosis, an organized lesion, reads near-null. The metrics measure spatial disorganization, of which malignancy is the largest driver. That is the precise statement of what the instrument reads, and it is the reason the pooled malignant-versus-benign figure is lower than the per-class figures.

4.6 Ovarian subtype structure — UBC-OCEAN (513–538 whole-slide images). Coherence and texture features read subtype structure among cancers at balanced accuracy 0.611 ± 0.079 on a three-class task (chance 0.33); this is a subtype task among malignancies, not cancer-versus-healthy detection.

4.7 Specificity. The same features applied to a generic image-forgery set return F1 = 0.000 — they do not respond to artificial manipulation of images. The metrics read tissue organization, not a universal image statistic.

What the evidence supports. Tissue organization is measurable from an image, and malignant tissue exhibits a measurable, cross-dataset shift toward spectral disorder — malignancy being the largest driver of that shift, with disorganized benign lesions reading as disorganized and organized benign lesions reading as organized. The shift is distributed and overlapping — a probabilistic displacement of the whole population, not a binary boundary — which is exactly the form a state transition takes when it is read at the population level. The evidence supports a tissue-organization biomarker and a contributing diagnostic feature today, and a transition readout once the time course is run; it does not yet speak to any single frequency parameter, which is the sweep's question.

5. The detection system

The Cancer Coherence Detection System — Clinical Edition v3.0 (February 2026; interface release v3.1, 23 October 2026) — and the Klein CDIS Workstation (July 2026) are the working embodiment of the framework. Release v3.1 adds the image-analysis channel and a clinical-workstation interface built to the program's calm-delivery standard: results are stated exactly in words and numbers and presented in composed colours, with no alarm theatre, and every simulated element is labeled as simulated on screen. The system is a network of interoperable subsystems. The couplings described below are embodiment-dependent and non-exclusive; any subsystem may operate independently, or be omitted, while the others operate. — DESIGNED.

5.1 Image-analysis channel. The real measurement channel of the system. An uploaded image is converted to grayscale, centre-cropped and resized to 512 × 512, and the nine-metric set of Section 4 is computed on it — 2D-FFT power spectrum, spectral purity, spectral flatness, spectral entropy, bicoherence, brightness, noise coefficient of variation, BSR in both formulations, and gradient density — by a JavaScript engine (coherence_metrics.js) that mirrors the verified Python analysis code. Processing is entirely local: no image leaves the machine. The channel scores the image against measured reference distributions for the tissue type selected — dermatoscopy (HAM10000, all seven lesion classes, n = 1,007 calibration images) or blood-cell microscopy (C-NMC leukemia and normal cells, n = 400) — and reports four readouts: each metric's value and its z-position against the benign reference; a Disorganization Index (mean |z| against the benign reference); the nearest reference class (nearest-centroid match across the lesion classes); and a Malignant-Pattern Score from a logistic model fitted on the calibration set. The interface states on screen that this is a pattern score, not a diagnosis. The metric set is VERIFIED on public data (Section 4); the channel itself is DESIGNED and operating, with its calibration references filed beside the verified analysis code.

5.2 Bioelectric impedance channel. Receives a time-series from an impedance spectrometer over USB (Web Serial), computes the spectrum, and reports the detected peak within a ±10 Hz window of 41.176 Hz, the frequency drift of that peak from 41.176 Hz, and the spectral purity (peak power over total power). It provides input to the Resonance Score. Instrument interface built; instrument to be connected.

5.3 Biophoton channel. Receives a photon-count time-series from a photomultiplier or EMCCD detector and reports mean emission, coefficient of variation and BSR. It provides input to the Resonance Score. The channel's decision rule will be set from measured emission statistics: the ultraweak-photon-emission literature reports elevated and more variable emission in malignant cells (Takeda et al. 2004; a 2020 thirteen-cell-line replication, ~90% discriminant accuracy; delayed-luminescence discrimination of melanoma from fibroblast, Musumeci et al. 2005), so the discriminating quantity in this channel is variability and temporal structure rather than brightness alone.

5.4 Resonance Score. Integrates the channels into a 0–100 index: 0.20 × spectral purity + 0.15 × (inverse) frequency drift + 0.30 × BSR + 0.35 × mean emission, each term normalized to a reference range, with a logistic cancer probability centered at a score of 40 and three bands — HEALTHY (≥ 60), SUSPICIOUS (40–60), CANCER (< 40). The weights and bands are design parameters. The Resonance Score takes its inputs from the bioelectric and biophoton channels, and its reference ranges were set from the SYNTHETIC parameter set used to exercise the pipeline; they will be re-fitted on instrument data before any clinical use. The image channel (5.1) is scored separately, against measured distributions, and does not pass through the Resonance Score.

5.5 Frequency-sweep module. Implements the pre-registered sweep protocol: 30–60 Hz at 0.1 Hz resolution, a set exposure duration per frequency, replications, PEMF amplitude in millitesla, four cell lines (MCF-7 breast adenocarcinoma, PC-3 prostate adenocarcinoma, THP-1 monocytic leukemia, B16-F10 murine melanoma), randomized double-blind frequency order, with coherence recovery versus frequency and a frequency × cell-line heatmap as outputs. The module commands the 41.176 Hz PEMF generator and the Arduino Due frequency controller and receives measurements from the channels above. Its on-screen output is a placeholder until the instruments are connected.

5.6 Clinical and data subsystems. Patient records (local store or PostgreSQL / MongoDB / MySQL) with role-based access and an audit log; batch analysis with sensitivity, specificity and ROC reporting; machine-learning training hooks (dense, convolutional, random forest, SVM) with TensorFlow.js export; data export in CSV, JSON and DICOM-compatible XML; a patient diagnostic report with a physician signature block; and generators for FDA 510(k), PMA and IDE submission packages and for IRB submissions. The Workstation adds a validation page, batch reporting and model export in a single application.

5.7 Hardware set (specified). Bioelectric impedance spectrometer; 41.176 Hz PEMF generator; photomultiplier / EMCCD photon counter; Arduino Due controller for frequency precision. A demo mode exercises the instrument channels, the sweep and the batch view on SYNTHETIC signals; the interface carries a persistent "Demo mode · simulated signals" indicator, and every simulated chart and figure carries a "Simulated" tag.

Regulatory status: Research Use Only; investigational device, pre-submission. The verified component is the image-domain metric set, which the image channel now runs on real images. The instrument channels, the sweep and the learning hooks are instrument-ready designs whose decision thresholds will be fitted on measured data.

6. Therapeutic and extension lines

6.1 Entrainment. The frequency sweep (Section 5.5, Section 7) is the first physical test of restoration: whether a structured field at cascade frequencies raises coherence metrics in malignant cell lines, with sham controls and operator-blinded frequency assignment, coherence restoration as the primary outcome. — HYPOTHESIS under test.

6.2 Regeneration as the inverse problem. Where cancer is framed as coherence disruption, regeneration is framed as coherence restoration. The framework maps the phases of wound healing — hemostasis, inflammation, proliferation, remodeling — onto bands of the cascade (N = 16 through 21, 43.75 to 33.33 Hz) and proposes that chronic wounds are stuck at a band and that fibrotic healing skips one. — HYPOTHESIS; a theoretical extension on file.

6.3 Collapse-threshold agents and photon emission. The framework names four substrates of biological coherence — cytoskeletal order, mitochondrial membrane potential, metabolic state, and ROS dynamics. Repurposed antiparasitic agents act on exactly these substrates (benzimidazoles on tubulin, ivermectin on membrane potential, fenbendazole on glycolysis). The framework reads them as collapse-threshold agents — they do not restore coherence; they push already-marginal cells over the edge — and proposes a first-in-literature experiment: photomultiplier measurement of ultraweak photon emission during graded compound exposure, with caspase-inhibitor controls, as a real-time coherence readout. — HYPOTHESIS; full protocol on file.

7. Research program

1. Tumorigenesis time course (the bridge experiment). Patient-derived matched organoids, normal → transition → carcinoma, four arms of at least twelve wells, daily imaging for 21 days; the frozen metric set of Section 4 on every frame; metric trajectories aligned to an independently measured transition state; the reversion intervention applied at the detected inflection in one arm and after lock-in in another, and the metrics re-read. Costs pixels, not transcriptomes. Prediction: the spectral signature begins its excursion at or before the transition state and recovers after reversion. Protocol v1.0 frozen for pre-registration on 22 October 2026 (on file).

2. Pre-registered frequency sweep. Section 5.5 as specified; thresholds, metric set and analysis frozen before data collection.

3. Living-cell photon emission under compound exposure. Section 6.3 as specified.

4. Correlative study. Histology-texture metrics from a resected specimen against fresh-tissue bioelectric and photon measurements from the same patient — the direct test of the spatial–temporal coupling in Section 2.4.

5. Integration. The image channel is in the interface as of v3.1 (Section 5.1). Next: extend its calibration from the subsample references to the full verified datasets and add histopathology references; connect the impedance spectrometer, photon counter and PEMF generator to the channels of Section 5; fit the Resonance Score reference ranges on measured data.

6. Whole-slide completion of the UBC-OCEAN extraction and matched-source expansion of the image pipeline to further public datasets.

7. Blind third-party validation. The standing offer to any qualified laboratory: data this program has never seen, labels withheld until results are returned.

8. Intellectual property

Provisional patent applications are on file: 63/937,225 (filed 7 November 2025, universal coherence detection at 41.176 Hz) and 64/028,112 (filed 5 January 2026), with the cancer detector consolidated into a non-provisional filing in November 2026.

9. References

• Shin D., Gong J.-R., Jeong S. D., Cho Y., Kim H. P., Kim T.-Y., Cho K.-H. Attractor Landscape Analysis Reveals a Reversion Switch in the Transition of Colorectal Tumorigenesis. Advanced Science, 2025. DOI 10.1002/advs.202412503.
• Cho S.-H., Cho K.-H. Attractor landscape analysis of colorectal tumorigenesis and its reversion. BMC Systems Biology, 2016.
• Scheffer M. et al. Early-warning signals for critical transitions. Nature 461, 53–59 (2009).
• Takeda M. et al. Biophoton detection as a novel technique for cancer imaging. Cancer Science 95, 656–661 (2004).
• Thirteen-cell-line ultraweak-photon-emission replication, 2020 (PMC7226102).
• Musumeci F., Scordino A., Tudisco S., Niggli H. J. Discrimination between normal and cancer cells by delayed luminescence. Applied Physics Letters, 2005.
• Datasets: Stanford prostate (prostate.data); Kaggle prostate morphometrics; C-NMC leukemia (ISBI 2019); PatchCamelyon; HAM10000; UBC-OCEAN.

Appendix — Provenance of this paper

• Verified findings and code: RESEARCH_LIBRARY\Cancer_Research\Verified_Coherence_Findings\2026-10-03_verified_coherence_findings_whitepaper.md; CNMC_Validation\ (metric code, per-image outputs, artifact provenance audit of 3 October 2026; the all-class HAM10000 run of 22 October 2026: ham10000_full_multiclass.py, ham10000_all_classes_metrics.csv).
• Attractor-landscape translation and the time-course design: Transition_Bridge\2026-10-22_attractor_landscape_spatial_coherence_translation.md; pre-registered protocol: Transition_Bridge\2026-10-22_preregistered_transition_timecourse_protocol.md.
• Antiparasitic / photon-emission analysis and protocol: Antiparasitic_UPE_Analysis\2026-10-03_antiparasitic_upe_coherence_analysis.md.
• Regeneration extension: Cellular_Regeneration_Framework_Application.md.
• Detection system: Inventions\07_Cancer_Detector\cancer_detection_clinical_gui_v3.1.html (current release, with the image channel) and its engine coherence_metrics.js; screenshots in 2026-10-23_v3.1_screenshots\; the February build cancer_detection_clinical_gui.html; Klein CDIS Workstation (standalone).html; Klein CDIS Research Deck (standalone).html. Image-channel calibration references: CNMC_Validation\2026-10-03_cnmc_matched_source_validation\2026-10-23_js_pipeline_calibration_refs.json.
• Earlier syntheses (January, May and July 2026) are retained in the library under SUPERSEDED_ labels; their figures are not used here.

Paul Samuel Guarino SEIS-UGFM Coherence Framework [email protected]