MIM Software Inc v. EXini Diagnostics Ab Inc

1. Case Identification

• Case #: IPR2025-00827
• Patent #: 11,941,817
• Filed: April 4, 2025
• Petitioner(s): MIM Software Inc.
• Patent Owner(s): EXINI Diagnostics AB.
• Challenged Claims: 1-5, 7-14, 16-19, 22-26, and 28-32

2. Patent Overview

• Title: Systems and Methods for Platform Agnostic Whole Body Image Segmentation
• Brief Description: The ’817 patent discloses systems and methods for automatically processing 3D medical images to identify cancerous lesions. The claimed method involves using machine learning to segment a 3D anatomical image (e.g., CT) into multiple target volumes of interest (VOIs), creating a 3D segmentation map from these VOIs, and then mapping it to a 3D functional image (e.g., PET) to detect hotspots representing lesions.

3. Grounds for Unpatentability

Ground 1: Anticipation by Renisch or Obviousness over Renisch in view of Zhao - Claims 1-5, 7, 10-14, 16, 19, and 26 are anticipated by Renisch, or in the alternative, are obvious over Renisch in view of Zhao.

• Prior Art Relied Upon: Renisch (Application # 2012/0123253) and Zhao (Patent 10,140,544).
• Core Argument for this Ground:
  • Prior Art Mapping: Petitioner argued that Renisch discloses every element of the independent claims. Renisch teaches a system that automatically processes 3D anatomical (CT) and 3D functional (PET) images to identify lesions. It explicitly discloses using machine learning (e.g., neural networks) for a "segmentation unit" to identify anatomical structures like the brain, heart, and liver (target VOIs). These segmented structures are then used to analyze the functional image to detect hotspots while suppressing regions of normal physiological uptake. For the alternative obviousness ground, Petitioner contended that Zhao explicitly teaches implementing image segmentation using a "multi-value mask," which corresponds to the claimed "3D segmentation map," and individual "segmentation masks," which correspond to the segmented VOIs.
  • Motivation to Combine (for §103 ground): A POSITA would combine Renisch with Zhao to implement the known segmentation process of Renisch using the explicit mask-based techniques taught by Zhao. Zhao provides a well-understood method for representing and using segmented regions (ROIs) to filter image data and mark organ locations, which would predictably improve the process of differentiating regions within the functional image as described by Renisch.
  • Expectation of Success (for §103 ground): A POSITA would have a high expectation of success, as Zhao’s teachings on segmentation masks are directly applicable to the 3D medical image segmentation performed in Renisch.

Ground 2: Obviousness over Baker in view of Zhao - Claims 1-2, 7-11, 16-18, 22-25, and 29-32 are obvious over Baker in view of Zhao.

• Prior Art Relied Upon: Baker (Application # 2018/0144828) and Zhao (Patent 10,140,544).
• Core Argument for this Ground:
  • Prior Art Mapping: Petitioner presented Baker as an alternative primary reference that is nearly identical to the ’817 patent. Baker disclosed automatically analyzing 3D anatomical (CT) and 3D functional (PET) images to detect cancer. It taught using machine learning techniques like convolutional neural networks for "automated segmentation of CT scans" to identify 3D boundaries of specific organs. Baker further taught transferring these identified 3D boundaries to the PET image via mapping to determine levels of cancerous tissue based on detected hotspots. The combination with Zhao was argued for the same reason as in Ground 1: to explicitly supply the "segmentation map" and "segmentation masks" terminology for Baker's process of creating and using multiple distinct organ volumes.
  • Motivation to Combine (for §103 ground): A POSITA would combine Baker with Zhao to implement Baker's organ segmentation using Zhao's express teachings on segmentation masks. This would provide a known and predictable way to represent the segmented 3D organ volumes for transfer to the functional image, enhancing the ability to differentiate between anatomical regions for hotspot analysis.
  • Expectation of Success (for §103 ground): Success would be expected, as both references are in the same field of 3D medical image segmentation, and applying Zhao’s mask generation techniques to Baker's segmented volumes is a straightforward implementation choice.

Ground 3: Obviousness over Baker-Zhao in view of Eiber - Claims 3-5 and 12-14 are obvious over Baker in view of Zhao and further in view of Eiber.

• Prior Art Relied Upon: Baker (Application # 2018/0144828), Zhao (Patent 10,140,544), and Eiber (a 2018 journal article introducing the PROMISE criteria).

• Core Argument for this Ground:

  • Prior Art Mapping: This ground built upon the Baker-Zhao combination to address dependent claims requiring the use of reference tissues to determine "hotspot index values." Petitioner argued that Eiber explicitly recommended a standardized method (PROMISE criteria) for prostate cancer evaluation that involves comparing the uptake intensity of a lesion (hotspot) to the uptake in reference organs like the liver and parotid gland. This comparison yields a standardized "miPSMA score," which Petitioner equated to the claimed "hotspot index value."
  • Motivation to Combine (for §103 ground): A POSITA would combine the Baker-Zhao system with Eiber's teachings to standardize the cancer evaluation process. Eiber provided an express motivation to adopt its PROMISE criteria to "aid reproducibility" and "enhance communication" in reporting PSMA PET/CT findings. Applying this standardized scoring method to the lesions detected by the Baker-Zhao system would be a logical step to improve its clinical utility.
  • Expectation of Success (for §103 ground): A POSITA would have a reasonable expectation of success because Baker already taught segmenting the liver, and Eiber taught using the measured uptake in the liver as a reference value. It would be straightforward to use the data already generated by Baker's system to perform the comparison taught by Eiber.

• Additional Grounds: Petitioner asserted additional obviousness challenges, including grounds based on Renisch/Renisch-Zhao in view of Baker or Eiber for claims related to PSMA-specific radiopharmaceuticals, and a ground based on Baker-Zhao in view of Suehling for claims related to excluding background voxels from hotspot detection.

4. Key Claim Construction Positions

• "3D segmentation map" (claims 1, 3, 10, 12, 19, 26): Petitioner proposed this term be construed as "a plurality of 3D segmentation masks distinguishing a plurality of regions within a 3D image." This construction was based on the patent's intrinsic evidence and was argued to be critical for showing that prior art references like Renisch and Baker, which create and use multiple distinct organ segmentations, meet this limitation even without using the exact phrase "segmentation map."

5. Arguments Regarding Discretionary Denial

• Petitioner argued that institution is appropriate under 35 U.S.C. §325(d) because the Examiner did not properly consider the most relevant prior art during prosecution. The petition contended that the Examiner erroneously identified a patent to Hamadeh as the "closest prior art," despite Hamadeh being limited to 2D images and not using machine learning for segmentation. Petitioner asserted that its primary references (Renisch, Baker, and Zhao) were either not considered or not substantively addressed, and these references teach the core claimed features, thus presenting arguments and combinations that are substantially different from those considered by the Office.

6. Relief Requested

• Petitioner requests institution of an inter partes review and cancellation of claims 1-5, 7-14, 16-19, 22-26, and 28-32 of the ’817 patent as unpatentable.