Rondevoo Tech LLC v. Genedata USA Inc

I. Executive Summary and Procedural Information

• Parties & Counsel:
  • Plaintiff: Rondevoo Technologies, LLC (California)
  • Defendant: Genedata (USA) Inc. (Delaware)
  • Plaintiff’s Counsel: Stamoulis & Weinblatt LLC; Sand, Sebolt & Wernow Co., LPA
• Case Identification: Rondevoo Technologies, LLC v. Genedata (USA) Inc., 1:19-cv-01553, D. Del., 08/20/2019
• Venue Allegations: Plaintiff alleges venue is proper because Defendant is a Delaware corporation and therefore resides in the District of Delaware.
• Core Dispute: Plaintiff alleges that Defendant’s image analysis software infringes three patents related to methods and systems for generating special-purpose image analysis algorithms through user-guided training.
• Technical Context: The technology at issue involves automated, computer-based image analysis, a field critical for accelerating research and diagnostics in areas such as histology, material science, and pharmaceutical development by quantifying and classifying features within digital images.
• Key Procedural History: The complaint does not reference any prior litigation, Inter Partes Review (IPR) proceedings, or licensing history related to the patents-in-suit.

Case Timeline

II. Technology and Patent(s)-in-Suit Analysis

U.S. Patent No. 7,088,854 - Method and apparatus for generating special-purpose image analysis algorithms (Issued Aug. 8, 2006)

The Invention Explained

• Problem Addressed: The patent describes the difficulty for computer systems to accurately and consistently classify and count objects ("entities") within complex images, a task that often requires expert human judgment to interpret subtle or varied features (’854 Patent, col. 1:41-47). Existing systems lacked a mechanism for incorporating this expert knowledge to automate the process effectively (’854 Patent, col. 2:50-54).
• The Patented Solution: The invention proposes a system for generating a custom "product algorithm" for image analysis. The process begins with an "evolving algorithm" that receives an image and identifies entities based on initial parameters. A user then provides "judgment" or feedback on the accuracy of the classification. This feedback refines the evolving algorithm, which is then stored as a final "product algorithm" capable of automatically classifying entities in new images without further user input (’854 Patent, Abstract; Fig. 2).
• Technical Importance: This technology aimed to standardize and automate the labor-intensive process of expert image quantification, particularly in scientific fields like histology, allowing for more reproducible and scalable data analysis (’854 Patent, col. 3:1-14).

Key Claims at a Glance

• The complaint asserts infringement of at least Claim 1 (Compl. ¶14).
• The essential elements of independent Claim 1 are:
  • A computer usable medium with program code configured to:
  • obtain an image with multiple "chromatic data points";
  • generate an "evolving algorithm" that partitions these data points into an entity based on a "user's judgment"; and
  • store a version of this evolving algorithm as a "product algorithm" that can automatically classify entities in a second image.

U.S. Patent No. 7,254,266 - Method and apparatus for generating special-purpose image analysis algorithms (Issued Aug. 7, 2007)

The Invention Explained

• Problem Addressed: As a continuation of the application leading to the ’854 Patent, the ’266 Patent addresses the same fundamental problem: the need for a system that can learn from an expert to automate the classification of image data (’266 Patent, col. 1:12-20).
• The Patented Solution: This patent claims a specific method for creating a product algorithm through an iterative training mode. The method involves a first user training an "evolving algorithm" by providing sequential feedback. The system presents a first set of identified entities, obtains feedback, executes the algorithm with that feedback, presents a second refined set of entities, obtains user approval, and then stores the trained algorithm. This "product algorithm" can then be provided to a second user for automated analysis on new images (’266 Patent, col. 5:21-42).
• Technical Importance: The invention formalizes a workflow for capturing expert knowledge in a discrete, iterative training session and packaging that knowledge into a deployable tool for broader, potentially non-expert, use (’266 Patent, col. 3:1-14).

Key Claims at a Glance

• The complaint asserts infringement of at least Claim 1 (Compl. ¶19).
• The essential elements of independent Claim 1 are:
  • A method for automating expert image quantification, comprising:
  • obtaining a product algorithm for analyzing image data via a "training mode" that uses iterative input from a first user to an "evolving algorithm";
  • The training mode comprises the steps of: presenting a first set of entities for feedback; obtaining feedback; executing the evolving algorithm with the feedback; presenting a second set of entities for feedback; obtaining approval; and storing the evolving algorithm as a product algorithm;
  • providing the stored product algorithm to a second user for application against a second set of image data.

Multi-Patent Capsule: U.S. Patent No. 8,687,879 - Method and apparatus for generating special-purpose image analysis algorithms (Issued Apr. 1, 2014)

• Technology Synopsis: This patent, from the same family as the ’854 and ’266 patents, claims a non-transitory computer program product for automating image quantification. The invention centers on generating a "locked evolving algorithm" through an iterative training mode where a user provides feedback to refine entity recognition. This trained algorithm is then stored for subsequent automated use on other image sets (’879 Patent, Abstract).
• Asserted Claims: The complaint asserts infringement of at least Claim 1 (Compl. ¶24).
• Accused Features: Plaintiff alleges that the Accused System's use of a training mode for a convolutional neural network (CNN), which relies on "user manual annotation of objects of interest" and subsequent "user feedback," infringes the claims of the ’879 Patent (Compl. ¶44).

III. The Accused Instrumentality

Product Identification

• The accused instrumentality is the "GeneData Imagence® for HCS Image Analysis" system ("Accused System") (Compl. ¶25).

Functionality and Market Context

• The complaint alleges the Accused System is a software solution that "enables image analysis based on product algorithms" for "detecting complex phenotypes" (Compl. ¶25, ¶27). Its relevant technical functionality is described as generating a classification algorithm based on a user's manual annotation of objects, which trains a convolutional neural network (CNN). The system then allegedly executes this algorithm based on user feedback to further train the CNN (Compl. ¶34). The complaint does not provide detail on the product's specific market positioning. No probative visual evidence provided in complaint.

IV. Analysis of Infringement Allegations

The complaint alleges that the Accused System’s method of training a convolutional neural network (CNN) through user annotation and feedback meets the limitations of the asserted claims. The narrative allegations suggest a direct mapping of the patented training process onto the Accused System's functionality.

’854 Patent Infringement Allegations

• Identified Points of Contention:
  • Scope Questions: A central question may be whether the term "evolving algorithm," as understood in light of the patent's specification (which provides examples using Bayesian methods), can be construed to read on the "convolutional neural network (CNN)" allegedly used by the Defendant (Compl. ¶34). The patent’s disclosure of a "neural network" as one type of classification algorithm may support Plaintiff's position (’854 Patent, col. 6:8-9).
  • Technical Questions: The claim requires identification "in accordance with a user's judgment." The complaint alleges this is met through "user manual annotation" and "user feedback" (Compl. ¶34). The specific nature and mechanism of this user interaction in the Accused System will be a key factual question in determining if it meets this limitation.

’266 Patent Infringement Allegations

• Identified Points of Contention:
  • Scope Questions: Claim 1 recites a specific, multi-step iterative sequence: "presenting a first set...obtaining feedback...presenting a second set...obtaining approval." The complaint's allegations that this is accomplished by "generating and executing the algorithm based on user feedback" (Compl. ¶35) raises the question of whether the Accused System's training workflow actually follows these discrete claimed steps or operates as a more continuous feedback loop.
  • Technical Questions: A factual dispute may arise over what constitutes "approval" from the user in the Accused System's workflow. The defense could argue that the system's process for finalizing a trained model does not meet the claim's specific "obtaining approval" step.

V. Key Claim Terms for Construction

• The Term: "evolving algorithm" (asserted in ’854 Claim 1 and ’266 Claim 1)
  • Context and Importance: This term is the technological core of the asserted claims. Its construction will determine whether the patents' scope is broad enough to cover modern machine learning approaches like the accused Convolutional Neural Network (CNN) or is limited to the specific types of algorithms detailed in the specification.
  • Intrinsic Evidence for Interpretation:
    • Evidence for a Broader Interpretation: The specification explicitly lists "a neural network" as an example of an "image classification algorithm" that the system may utilize, suggesting the inventors contemplated technologies beyond the primary embodiments (’854 Patent, col. 6:8-9).
    • Evidence for a Narrower Interpretation: The patent’s detailed description and figures focus heavily on an implementation using Bayes' Theorem to determine pixel-class probabilities, which could be argued to limit the scope of "evolving algorithm" to that specific approach or closely related statistical methods (’854 Patent, Fig. 8; col. 15:60-16:5).
• The Term: "user's judgment" (asserted in ’854 Claim 1)
  • Context and Importance: This term defines the nature of the human input required to train the "evolving algorithm." The breadth of this term is critical, as it dictates what types of user interactions constitute infringement.
  • Intrinsic Evidence for Interpretation:
    • Evidence for a Broader Interpretation: The specification describes user input as potentially including selecting an area of an image that contains an entity or identifying portions of the image that are background, suggesting a range of possible interactions (’854 Patent, col. 8:30-54).
    • Evidence for a Narrower Interpretation: The primary examples describe the user selecting a "sample set of chromatic data points" to define "pixel classes" (’854 Patent, col. 9:10-16). A defendant may argue that "judgment" is limited to this specific act of defining classes by sampling pixels, rather than more abstract annotations like drawing a bounding box.

VI. Other Allegations

• Indirect Infringement: The complaint does not plead facts sufficient to support a claim for either induced or contributory infringement.
• Willful Infringement: The complaint alleges that Defendant had knowledge of the patents-in-suit "at least as of the service of the present Complaint" (Compl. ¶53). This allegation, if proven, may support a finding of willful infringement for conduct occurring only after the complaint was filed.

VII. Analyst’s Conclusion: Key Questions for the Case

This dispute appears to center on the application of patent claims, drafted in the early 2000s, to a modern machine-learning system. The outcome may depend on the court's interpretation of key claim terms and a factual analysis of the accused software's precise operation.

• A core issue will be one of definitional scope: can the term "evolving algorithm," which the patents describe in detail using Bayesian classification methods, be construed broadly enough to read on the accused product’s alleged use of a Convolutional Neural Network (CNN)?
• A key evidentiary question will be one of operational correspondence: does the accused software's workflow for user-guided training perform the specific, multi-step iterative sequence of presenting sets of entities, obtaining feedback, and securing approval as required by Claim 1 of the ’266 patent, or is there a fundamental mismatch in its technical operation?