Science
Living cells give off a faint light. HelioFlux is building a passive platform to measure that signal and test what its color and timing can tell us about cancer.
A decade of published work in cells and animals, plus exploratory measurements in healthy adults. The proposed next study is in people with cancer.
Oxidative chemical reactions in living tissue can release faint light. A photon counter converts some arriving photons into electrical signals.
This drawing shows the principle. The real measurements come from dishes of cells, whole animals and, in the 2025 study, recordings made near the human head. How photon detection works ↗
No dye, no needle, no added radiation. The light is already there. The challenge is collecting enough photons above background, then determining what their color and timing can tell us.
A small fraction of reactions linked to oxidative metabolism release photons.
Only a minute fraction leaves the sample. The signal remains far below visual detection.
In a dark chamber, its photon-counting detector is meant to record individual events as they arrive.
The aim is to evaluate spectral and temporal features, not only total brightness.
Living cells can emit ultraweak photons as a by-product of oxidative chemistry. The signal is invisible to the eye but measurable with a photon-counting detector.
In Dr. Murugan’s cell studies, malignant and non-malignant cell lines showed different spectral and temporal features. HelioFlux is developing tools to measure and analyze those patterns. See the studies
Left · Non-cancer cellsRight · Cancer cells
Bars, left to right: 420 nm 620 nm 950 nm
Counting every photon, the totals are close but not identical.
An illustration of the 2018 finding, not study data. In that study, total counts could not separate the groups; counts through 420, 620 and 950 nm filters together classified 69% of recordings correctly. Across all six filters, cancer lines leaned weakly toward shorter, violet-to-ultraviolet wavelengths and non-cancer lines toward near-infrared.
Reading the experiment
Total brightness alone could not separate the groups. Filtering the light by wavelength did better.
Cancer and non-cancer lines, each recorded in triplicate.
One wavelength band per recording.
A photon counter under the dish for close to a day.
Color describes the light’s wavelength. Timing describes how the measured signal varies over time. They are different ways of studying the same light.
The roughly 20 Hz feature in the 2020 paper is a rhythm in the photon count over time. It has nothing to do with the color of the light. Read the 2020 paper ↗
The platform in development is designed to pair photon-counting hardware with analysis software for the spectral and temporal features of the recorded signal.
Designed to record individual photon events.
QSense is a research-stage system in development, designed to measure the ultraweak light a sample gives off on its own. It is meant to work in darkness, without illuminating the sample or adding a fluorescent label.
Designed to analyze color and timing.
LuminAI is being developed to evaluate spectral and temporal features in photon-count data. Its role is to test whether those features contain reproducible biological information. Human cancer performance has not yet been established.
The tissue emits the signal without an external light source or added contrast agent. QSense is designed to collect that emission in darkness. The proposed study in people with cancer would compare those measurements with biopsy results to test what, if anything, the signal adds.
Dr. Nirosha Murugan’s published work spans cell cultures, a mouse melanoma model, and exploratory measurements in healthy adults. Each study addresses a different level of evidence.
Published in
In a small in-vitro study, light counted through three wavelength filters classified 69% of recordings correctly as cancer or non-cancer. Setting aside one of the two non-cancer lines raised that to 92%.
Cancer cells leaned toward near-ultraviolet light and non-cancer cells toward near-infrared, a weak but opposite trend across the six filters. Over the 23-hour recording, the ultraviolet-to-infrared ratio differed between the groups only between hours 13 and 15. Three filters, at 420, 620 and 950 nm, did the separating. Unfiltered counts could not. With all six lines included, the three filters together classified 69% of recordings correctly; the 92% figure came after setting aside HBL100, one of the two non-cancer lines, which the authors argue is transformed rather than truly normal. In that subset, every cancer recording was classified correctly, but only 63% of non-cancer recordings were. This is cell-dish evidence. The proposed study in people with cancer is described later on this page.
Murugan et al. · Biophotonic markers of malignancy: Discriminating cancers using wavelength-specific biophotons. Biochemistry and Biophysics Reports. Read the original paper ↗
The study reported two distinct findings. In the cell work, signal-frequency features sorted cancer and non-cancer samples correctly about 70% of the time with all 13 human and mouse lines pooled, and about 83% for pairs from the same tissue (the abstract reports 90%). In a separate series of four melanoma experiments with 47 mice in all, group-level differences appeared at 24 hours. The mouse study did not report diagnostic accuracy.
The cell work covered 13 human and mouse lines, recorded for just 100 seconds each. Analyzing how the photon count rose and fell, rather than the raw total, gave the paper’s accuracy figures: about 70% across all 13 lines and about 83% for cancer and non-cancer lines from the same tissue; the abstract gives a single figure of 90%.
The mouse experiment compared 23 animals given live melanoma cells, 8 given UV-killed cells and 16 left alone, 47 in all. At 24 hours the live-tumor group showed a difference near 22 Hz. A six-mouse subset showed another group difference at day 19. These were group-level statistical findings, not an animal diagnostic-accuracy result.
Murugan et al. · Ultraweak photon emissions as a non-invasive, early-malignancy detection tool: An in vitro and in vivo study. Cancers. Read the original paper ↗
In 20 healthy adults, photon measurements near the head differed from background. Some varied with EEG rhythms, though with eyes closed a background detector did too. It was an exploratory study of brain activity, not cancer detection; light from the human body had already been imaged in 2009.
Light recorded near the head differed from background light in its variability and rhythm, mostly below 1 Hz. With eyes closed, it varied with EEG alpha activity, as did the background detector pointed away from the head, so the authors call for experiments with electromagnetic shielding. The study did not investigate cancer or establish which tissue produced the detected photons.
Casey et al. · Exploring ultraweak photon emissions as optical markers of brain activity. iScience. Nirosha Murugan, senior and corresponding author. Read the original paper ↗
The full research timeline
From the first experiments in 1923 to modern photon imaging, with illustrations and original papers.
Record the tissue’s photon signal.
Match the measurement to the clinical result.
Test whether the signal adds clinically useful information.
Short answers, with the sources behind them.
Because the signal sits close to instrument background. Modern photon-counting detectors, controlled dark measurements and statistical methods have made it more practical to separate biological emission from noise and examine patterns in the counts. Follow the research history →
No. Thermal imaging records infrared radiation associated with temperature. Ultraweak photon-emission research measures a much fainter signal associated with chemical processes. In the well-known 2009 human-imaging paper, the thermal image is a separate comparison, not the photon image. Read the human-imaging paper ↗
Recordings are made in darkened enclosures. Detector dark counts and environmental background are measured, and samples are compared with controls recorded under the same conditions. Some early claims did not survive careful re-testing. Those failures helped shape the controls used today. See the 1934 negative result →
A passive readout related to ongoing cell chemistry, measured without an excitation source or added contrast agent. Preclinical studies found spectral or temporal differences in malignant cell samples and group-level changes in a mouse melanoma model. Whether those patterns add clinically useful information in people is what a biopsy-matched human study must test.
No. Biopsy remains the clinical reference. The question is whether a passive light measurement can add useful information before biopsy or help identify which cases warrant further evaluation.
Success would require a repeatable signal measured under a consistent protocol and quantified against biopsy-confirmed status. Those results would determine whether a larger validation study is warranted.
In the science press
Biophoton science in focus
Mentions Dr. Nirosha Murugan, HelioFlux’s co-founder and chief scientist.
Science journalist Jo Marchant reports on biophoton research and its possible applications. This is a Nature news feature, not a new experimental study.
Read the Nature feature (opens in a new tab)Science journalism, not validation or endorsement of HelioFlux. Full feature may require a subscription.
Request the research, development roadmap and technical background.
