CAMBRIDGE, Mass., July 21 — An algorithm developed at the Massachusetts Institute of Technology has processed stellar variability data from more than 100,000 stars in the Transiting Exoplanet Survey Satellite (TESS) catalog, a step that narrowed candidate identifications to ten and left only one target that later passed verification by the Spitzer Space Telescope. The approach, detailed by the MIT team, directly confronts a problem endemic to exoplanet detection: distinguishing genuine planetary transits from the natural flicker of stars.
TESS, the orbiting surveyor, generates roughly 27 terabytes of data each year. Much of that data stream is contaminated by stellar noise—phenomena such as starspots, flares, and surface granulation that can either mimic the periodic dimming of a transiting planet or mask it entirely.
The MIT team built a custom algorithm, dubbed The Notch Filter, designed to remove periodic stellar variability while preserving the transit signals embedded in the light curves.
The Notch Filter Approach
Applied to the full TESS catalog, the algorithm winnowed the field from over 100,000 stars to ten that showed weak, periodic dimming. Of those ten candidates, according to the record, only one passed the additional verification carried out by the Spitzer Space Telescope. The precise identity of that star and the planet candidate were not disclosed in the source material, but the filtering method itself has been documented as a tool for cleaning stellar noise at scale.
The problem of stellar noise is not unique to TESS. Earlier missions, such as the Kepler Space Telescope, faced similar challenges.
Machine learning techniques, including deep convolutional neural networks, have already been applied to identify candidate planets in both Kepler and TESS data. The Notch Filter represents a complementary statistical method: a targeted removal of periodic brightness variations that come from the star itself rather than an orbiting body.
Machine Learning and Future Directions
Data mining from large astronomical surveys is expected to yield additional hidden planets, particularly around M-dwarf stars. These small, dim stars are the most common in the galaxy, but their signals are the hardest to detect. Stellar noise on M-dwarfs can be especially persistent, and the transit signal of a potentially habitable world around such a star may be only a few hundred parts per million deep.
Surveyors like TESS produce enough data over time that algorithms capable of filtering that noise could unlock planets that would otherwise remain invisible. The timeline of the MIT work—developing the algorithm, applying it to the catalog, and verifying results with Spitzer—follows a pattern common in exoplanet science: first a large data haul, then a computational sieve, then a smaller set of targets for a space telescope to confirm.
The record shows that from more than 100,000 stars, that sieve produced exactly one confirmed detection. The remaining nine candidates from the initial group were not further discussed in the source, but the yield underscores the difficulty of the search. Sources familiar with the method described the approach as a “statistical notch filter” that removes only the specific periodic frequency of stellar rotation or pulsation, leaving the broadband transit signal intact.
The technique is not new in signal processing but is tailored here to the specific frequencies encountered in TESS data. As the survey continues to accumulate years of observations, the same algorithm could be run again on later data releases, potentially increasing the number of detections around M-dwarfs.
No timeline for wider application of The Notch Filter has been announced. The research team at MIT continues to refine the algorithm, and the scientific community will watch whether this single verification becomes a template for finding more transits hidden in the noise. For now, the record stands: one star, one planet, uncovered from a field of 100,000 by a narrow computational filter.
{“@context”: “https://schema.org”, “@type”: “NewsArticle”, “headline”: “Astronomers sift billions of data points to find planet”, “description”: “CAMBRIDGE, Mass., July 21 — An algorithm developed at the Massachusetts Institute of Technology has processed stellar variability data from more than 100,000 stars in…”, “datePublished”: “2026-07-21T09:30:00+00:00”, “author”: {“@type”: “Person”, “name”: “Ariel Antonio”}, “publisher”: {“@type”: “Organization”, “name”: “InfoPulse Today”, “url”: “https://newsfeed.infopulsetoday.com/”}}
