PTAB

IPR2026-00185

Guardant Health Inc v. Tempus Ai Inc

Log In
Key Events
Petition

1. Case Identification

2. Patent Overview

  • Title: Techniques for Analysis of Medical Images
  • Brief Description: The ’097 patent describes methods for analyzing digital images of tissue slides, such as for identifying cancerous tissue. The methods involve separating a digital image into a plurality of "tiles" and then identifying the tissue class within a given tile based on features from both the tile itself and the broader context of a larger, surrounding portion of the image.

3. Grounds for Unpatentability

Ground 1: Anticipation of Claims 1, 4-6, 9-12, 14, 15, 17, and 18 by Chukka

  • Prior Art Relied Upon: Chukka (Application # 2016/0042511).
  • Core Argument for this Ground:
    • Prior Art Mapping: Petitioner argued that Chukka, which discloses a machine learning system for automated analysis of digital pathology slides, teaches every limitation of the challenged claims. Chukka describes receiving a digital slide, tiling the image, and performing analysis by using both "single tile information from within the tiled region" and "information from the whole slide context." Petitioner asserted this contextual analysis, including considering "surrounding tiles," directly teaches the ’097 patent’s key limitation of identifying tissue features in a tile based on a larger portion of the image. Chukka also discloses creating an "overlay image" to label features, classifying various tissue types (e.g., lymphocytes, stroma), and counting cells, mapping to the limitations of independent claims 1 and 6 and numerous dependent claims.

Ground 2: Obviousness of Claims 2, 3, 7, and 8 over Chukka in view of Jones

  • Prior Art Relied Upon: Chukka (Application # 2016/0042511) and Jones (a 2005 publication titled "Voronoi-based segmentation of cells on image manifolds").
  • Core Argument for this Ground:
    • Prior Art Mapping: Petitioner contended that to the extent Chukka’s teaching of using Voronoi tessellation to identify cell boundaries is not sufficiently detailed, Jones renders these features obvious. Chukka teaches generating a foreground mask and identifying seed points (nuclei) for cell analysis. Jones teaches a specific algorithm for defining cell-to-cell boundaries by computing Voronoi regions from such pre-identified seed areas on a pixel-by-pixel basis. Together, Chukka and Jones teach generating a digital overlay drawing of the outer edge of each cell (claim 2) at an individual pixel resolution (claim 3).
    • Motivation to Combine: A POSITA implementing Chukka’s system, which suggests using Voronoi tessellation to define cell "blobs," would combine its teachings with a well-known reference like Jones that provides a specific, effective method for performing such Voronoi-based cell segmentation.
    • Expectation of Success: A POSITA would have a reasonable expectation of success because Chukka provides the exact inputs required by the Jones method—namely, seed points and an object foreground mask—to effectively outline cell boundaries.

Ground 3: Obviousness of Claims 1, 4-6, 9-12, 14, 15, 17, and 18 over Gallas in view of Chukka

  • Prior Art Relied Upon: Gallas (a 2014 journal article titled "Evaluation environment for digital and analog pathology") and Chukka (Application # 2016/0042511).

  • Core Argument for this Ground:

    • Prior Art Mapping: Petitioner argued that Gallas discloses the conventional manual method for analyzing tissue slides, which the ’097 patent merely automates. Gallas shows a digital reticle (grid) overlaid on a slide image, where a pathologist evaluates features (e.g., a mitotic figure) that span multiple grid squares, thus inherently requiring analysis of an evaluation tile in the context of its surrounding tiles. Petitioner asserted that Chukka provides the well-understood automation for this conventional, manual process described by Gallas. The combination, therefore, renders the claims obvious.
    • Motivation to Combine: A POSITA would be motivated to apply the automated methods taught by Chukka to the conventional grid-based analysis framework disclosed in Gallas to improve the efficiency and objectivity of a known manual task. Both references discuss the benefits of computer-implemented tissue analysis, providing an express motivation to combine.
    • Expectation of Success: A POSITA would expect success in automating the Gallas method with Chukka’s system, as it is a mere application of a known computer-implemented technique to a prior art process ready for improvement.

  • Additional Grounds: Petitioner asserted additional obviousness challenges based on combining Chukka or Gallas/Chukka with Sebastiao (for specific cell markers in claims 12, 13, 16) and Gallas/Chukka with Jones (for cell boundary drawing in claims 2, 3, 7, 8), relying on similar motivations to add known techniques for specific functionalities.

4. Arguments Regarding Discretionary Denial

  • Petitioner argued that discretionary denial under §325(d) would be inappropriate because none of the asserted prior art references (Chukka, Gallas, Jones, and Sebastiao) were considered by the USPTO during the original prosecution of the ’097 patent. The petition asserted that this presentation of new, non-cumulative art demonstrates a material error by the Office in allowing the challenged claims, weighing against discretionary denial.

5. Relief Requested

  • Petitioner requested the institution of an inter partes review and the cancellation of claims 1-18 of the ’097 patent as unpatentable under 35 U.S.C. §102 and §103.