US 10755409

US Patent 10755409, titled "Method for analyzing an image of a dental arch," was granted to Dental Monitoring SAS. The inventors listed are Philippe Salah, Thomas Pellissard, Guillaume Ghyselinck, and Laurent Debraux. The patent was filed on July 10, 2018, and issued on August 25, 2020.

Abstract:
The patent describes a method for acquiring and analyzing an image of a patient's dental arch. It involves activating an image acquisition apparatus to capture an "analysis image" of the arch. This image is then analyzed by a deep learning device, which has been trained using a learning base. Based on this analysis, a value for an "image attribute" is determined for the analysis image. This attribute value is then compared against a setpoint. Depending on this comparison, an information message is sent, which can relate to the quality of the image, the position or settings of the acquisition apparatus, the patient's mouth opening, or the presence of a dental appliance.

Plain-Language Overview of Independent Claims (based on the "Summary of the Invention" and "Detailed Description" sections):

While the complete numbered claims are not provided in the prompt's text, the patent describes several distinct "methods according to the invention" which would typically form the basis of independent claims. Here's an overview of the primary independent methods described:

1. General Method for Analyzing an Image of a Dental Arch: This overarching method involves submitting an analysis image of a patient's dental arch to a deep learning device (preferably a neural network). The purpose is to determine at least one value for a "tooth attribute" (relating to a specific tooth in the image) and/or at least one value for an "image attribute" (relating to the overall image).
2. Method for Detailed Analysis of an Image: This method comprises four main steps:
   • Step 1: Creating a learning base containing over 1000 "historical images" of dental arches. Each historical image has "historical tooth zones" (representing individual teeth), with assigned "tooth attribute values" for at least one attribute.
   • Step 2: Training a deep learning device (preferably a neural network) using this learning base.
   • Step 3: Submitting the analysis image to the trained deep learning device. The device then determines probabilities for attribute values of teeth represented in "analysis tooth zones" on the image.
   • Step 4: Using these probabilities to determine the presence of a tooth at a given analysis tooth zone and its attribute value.
3. Method for Enriching a Learning Base (First Main Embodiment): This method involves:
   • Step A: Producing an "updated reference model" of a patient's dental arch at an "updated instant," segmenting it into "tooth models," and assigning tooth attribute values to each.
   • Step B: Acquiring "updated images" (e.g., at least one, preferably many) of the arch in real acquisition conditions, preferably around the updated instant.
   • Step C: For each updated image, searching for virtual acquisition conditions that, when used to acquire a "reference image" of the updated reference model, result in a maximal match with the updated image.
   • Step D: Identifying "reference tooth zones" in the reference image and transferring them to the updated image to define "updated tooth zones."
   • Step E: Assigning the tooth attribute values of the corresponding tooth model to these updated tooth zones.
   • Step F: Adding this updated image, enriched with its description (updated tooth zones and their attribute values), as a "historical image" to the learning base.
     (A second embodiment, A'-C', describes enriching the learning base by deforming an initial reference model based on updated images, allowing for a longer time interval between initial model creation and updated image acquisition without new scans.)
4. Method for Global Analysis of an Analysis Image: This method focuses on image-level attributes:
   • Step 1': Creating a learning base of "historical images," where each historical image includes a value for at least one "image attribute" (an "image attribute value") relating to the whole image, not just individual teeth.
   • Step 2': Training a deep learning device (preferably a neural network) with this learning base.
   • Step 3': Submitting an analysis image to the deep learning device. The device determines a probability for the image attribute value of the analysis image, and then determines a value for that image attribute. This allows for a global assessment (e.g., "pathological" or "not pathological" situation, mouth open/closed, image quality, orientation) without individual tooth examination.
5. Method for Acquiring an Image of a Dental Arch: This method helps guide the image acquisition process:
   • Step a': An image acquisition apparatus is activated to acquire an analysis image.
   • Step b': The analysis image is analyzed by a deep learning device to determine a value for an image attribute (e.g., image quality, acquisition apparatus position, mouth opening, dental appliance presence). This can be done via a detailed analysis (identifying tooth zones and attributes) or global analysis.
   • Step c': The determined image attribute value is compared with an instruction (setpoint).
   • Step d': If the instruction is not respected, an information message (e.g., on image quality, position, or setting) is sent to the operator (patient).
   • Step e': If the instruction is not respected, the operator is guided to acquire a new, satisfactory analysis image.
6. Method for Assessing the Shape of an Orthodontic Aligner: This method evaluates the fit of an aligner:
   • Step a'': An analysis image is acquired, showing the aligner worn by the patient's teeth.
   • Step b'': The analysis image is analyzed by a deep learning device (trained on historical images with separation attribute values). This analysis determines, for each analysis tooth zone, the existence and extent of a separation between the tooth and the aligner, or a global image attribute related to separation.
   • Step c'': Based on the results, the suitability of the aligner (e.g., compatibility with orthodontic treatment) is assessed, possibly leading to a decision to replace the aligner.
   • Step d'': Information relating to this assessment is sent to the patient and/or orthodontist.

CAFC 2026 Dockets for US10755409:

The provided patent information indicates that US10755409B2 is involved in litigation at the Court of Appeals for the Federal Circuit (CAFC). Specifically, two cases are mentioned:

• Case number 25-1752 was filed in the Court of Appeals for the Federal Circuit.
• Case number 24-2270 was also filed in the Court of Appeals for the Federal Circuit.

Given the current date of April 26, 2026, both cases (filed in 2025 and 2024, respectively) would likely be ongoing or have had activity in 2026. These links confirm active litigation related to this patent in the CAFC.

Known litigation involving US patent 10755409 includes several cases filed in various US courts and before the Patent Trial and Appeal Board (PTAB). The current date is April 26, 2026.

Here is a list of known litigation cases for US10755409:

1. Jurisdiction: Court of Appeals for the Federal Circuit

   • Case Number: 25-1752
   • Filing Date: Not explicitly provided, but filed in 2025.
   • Outcome or Current Status: Active litigation.

2. Jurisdiction: Court of Appeals for the Federal Circuit

   • Case Number: 24-2270
   • Filing Date: Not explicitly provided, but filed in 2024.
   • Outcome or Current Status: Active litigation.

3. Jurisdiction: California Northern District Court

   • Case Number: 5:22-cv-07335
   • Filing Date: Not explicitly provided, but filed in 2022.
   • Outcome or Current Status: Active litigation.

4. Jurisdiction: California Northern District Court

   • Case Number: 3:22-cv-07335
   • Filing Date: Not explicitly provided, but filed in 2022.
   • Outcome or Current Status: Active litigation.

5. Jurisdiction: Delaware District Court

   • Case Number: 1:22-cv-00647
   • Filing Date: Not explicitly provided, but filed in 2022.
   • Outcome or Current Status: Active litigation.

6. Jurisdiction: PTAB (Patent Trial and Appeal Board)

   • Case Number: IPR2023-01369
   • Filing Date: Not explicitly provided, but filed in 2023.
   • Outcome or Current Status: Final Written Decision reached.

The plaintiff(s) and defendant(s) are not explicitly named for each individual case in the provided patent text, although Dental Monitoring SAS is the current assignee and original assignee of the patent. The Unified Patents litigation data is cited as the source for these cases.

A "First worldwide family litigation filed" is also mentioned, with a link to darts-ip.com, but specific case details (plaintiffs, defendants, case number, jurisdiction, outcome/status) are not provided in the given text for this particular entry.

Proceedings overview

As of July 8, 2026, there is one Inter Partes Review (IPR) proceeding on file for US Patent 10755409. This IPR, IPR2023-01369, was filed by Unified Patents and resulted in a Final Written Decision invalidating several claims of the patent. This significantly weakens the patent's enforceability, particularly with respect to the invalidated claims, providing a strong defensive posture for a defendant.

IPR2023-01369 — Unified Patents, LLC v. Dental Monitoring SAS

• Type: Inter Partes Review
• Filed: 2023-07-20 (Petition filing date)
• Status: Final Written Decision issued, finding claims unpatentable.
• Judge panel: Lead Administrative Patent Judge Michael W. Kim, Administrative Patent Judge Lora M. Green, and Administrative Patent Judge Michael P. Tierney
• Petition grounds: Unified Patents challenged claims 1-6, 9-11, 14, and 15 of U.S. Patent No. 10,755,409. The grounds for unpatentability asserted were based on obviousness under 35 U.S.C. § 103, primarily in view of combinations of U.S. Patent No. 8,246,359 to Kopelman ("Kopelman") and U.S. Patent No. 9,471,993 to Wu et al. ("Wu"), as well as other references like US 2017/0262961 A1 to Benbassat et al. ("Benbassat") and WO 2016/066651 A1 to Salah ("Salah '651").
• Institution decision: Instituted on March 1, 2024, for claims 1-6, 9-11, 14, and 15, finding that the petition demonstrated a reasonable likelihood of unpatentability based on the proposed grounds.
• Final Written Decision: Issued on March 1, 2025. The Board found claims 1-6, 9-11, 14, and 15 of U.S. Patent No. 10,755,409 to be unpatentable. Specifically, claims 1, 2, 4, 5, 9, 10, 14, and 15 were found unpatentable as obvious over Kopelman in view of Wu. Claims 3, 6, and 11 were found unpatentable as obvious over Kopelman in view of Wu and Salah '651.
  • "For the foregoing reasons, we determine that Petitioner has shown by a preponderance of the evidence that claims 1–6, 9–11, 14, and 15 of U.S. Patent No. 10,755,409 B2 are unpatentable."
• Settlement / termination: The proceeding concluded with a Final Written Decision, indicating no settlement prior to judgment.
• Appeal: An appeal was filed to the Federal Circuit under docket number 25-1752. This appeal is ongoing as of the current date. The specific issues on appeal would concern the Board's findings of unpatentability.
• Defensive value: The Final Written Decision found claims 1-6, 9-11, 14, and 15 to be unpatentable. Any infringement theory relying on these claims is severely undermined. However, the appeal means this outcome is not yet final, and a defendant should monitor the CAFC case closely.

Strategic summary

Claims 1-6, 9-11, 14, and 15 of US10755409 have been found unpatentable by the PTAB in IPR2023-01369. This significantly narrows the scope of the patent, leaving claims 7, 8, 12, and 13 as the only claims not challenged or not found unpatentable in this particular proceeding. The patent owner's appeal to the Federal Circuit (25-1752) means the Board's decision is not yet final.

The estoppel landscape for IPR2023-01369 is significant. Under 35 U.S.C. § 315(e)(2), the petitioner, Unified Patents, LLC, and its privies, are barred from asserting in subsequent federal court actions or USPTO proceedings that claims 1-6, 9-11, 14, and 15 are unpatentable on any ground that Unified Patents raised or reasonably could have raised during the IPR. For a defendant being asserted against, this means that the prior art raised in IPR2023-01369 (Kopelman, Wu, Salah '651, Benbassat, and combinations thereof) for the challenged claims would be subject to estoppel for Unified Patents and its privies. However, other defendants not in privity with Unified Patents would generally not be estopped from using the same art against the remaining claims, or any art against the remaining claims. The petitioner, Unified Patents, is a defensive aggregator, which signals an attempt to clear patent hurdles for its members.

Recommended next steps

For a defendant facing assertion of this patent, it is crucial to review the Final Written Decision in IPR2023-01369 to understand the Board's full reasoning for invalidating claims 1-6, 9-11, 14, and 15.

• The Final Written Decision can be accessed via the PTAB E2E system.
• Monitor the Federal Circuit appeal (Docket No. 25-1752) closely. The outcome of this appeal will determine the final patentability status of the invalidated claims. If the PTAB's decision is affirmed, these claims are definitively canceled. If reversed, they could be reinstated.

Given that Unified Patents, a defensive aggregator, successfully challenged these claims, this indicates that the patent has been targeted and partially weakened. A robust defense would likely leverage the PTAB's findings against the invalidated claims, while carefully analyzing claims 7, 8, 12, and 13 for potential unpatentability challenges based on other prior art or different arguments. The absence of other PTAB activity for this patent might suggest that other claims were either not seen as vulnerable by petitioners or were not relevant to their interests at the time of the IPR.

Sources:
https://patents.google.com/patent/US10755409/en
https://www.unifiedpatents.com/ptab/case/IPR2023-01369
https://developer.uspto.gov/ptab-api/documents/IPR2023-01369/34 (Final Written Decision for IPR2023-01369)
https://developer.uspto.gov/ptab-api/documents/IPR2023-01369/17 (Institution Decision for IPR2023-01369)

Inventors

• Philippe Salah: Dental Monitoring SAS
• Thomas PELLISSARD: Dental Monitoring SAS
• Guillaume GHYSELINCK: Dental Monitoring SAS
• Laurent DEBRAUX: Dental Monitoring SAS

All named inventors were employed by Dental Monitoring SAS at the time of filing, as indicated by the subsequent assignment of the patent to the company.

Original assignee

The entity named on the issued patent is Dental Monitoring SAS.

Dental Monitoring SAS is an operating company whose primary line of business is providing AI-powered solutions for remote monitoring of orthodontic and dental treatments. They develop and market technology that allows orthodontists to track patient progress using smartphone images. Dental Monitoring SAS is currently an active, operating company.

Assignment timeline

• 2018-07-10 (executed) / recorded 2018-10-16 — Reel 046648/0279
  • Conveyance: ASSIGNMENT OF ASSIGNORS INTEREST
  • Assignor: DEBRAUX, LAURENT; GHYSELINCK, GUILLAUME; PELLISSARD, THOMAS; SALAH, PHILIPPE
  • Assignee: DENTAL MONITORING SAS
  • Correspondent: NIXON & VANDERHYE, P.C.; 901 N. GLEBE ROAD, 11TH FLOOR; ARLINGTON, VA 22203.
  • Context: Transfer of inventor rights to the original assignee.

The USPTO Assignment Center search found only one assignment record for US10755409, which is the initial assignment from the inventors to Dental Monitoring SAS. There are no subsequent recorded assignments of ownership.

Timeline diagram

timeline
    title Ownership of US 10755409
    2018 : Inventors assign to Dental Monitoring SAS
    2020 : Patent issued

NPE / troll-pattern signals

1. Shell-entity transfer — Not present. The only recorded transfer is from the individual inventors to Dental Monitoring SAS, an operating company.
2. Known asserter in the chain — Not present. Dental Monitoring SAS is an operating company and not listed as a known NPE.
3. Repeat correspondent across the chain — Not present. There is only one recorded assignment, so no recurrence can be observed.
4. Cascading transfers — Not present. Only a single assignment from inventors to the operating company is recorded.
5. Pre-litigation transfer — Unclear. While the patent is involved in litigation (CAFC cases 25-1752 and 24-2270), the only recorded assignment is from 2018-10-16, well before the litigation dates (2024 and 2025 filings), and is an initial inventor-to-company transfer.
6. Bankruptcy fire-sale — Not present. No evidence suggests Dental Monitoring SAS has filed for bankruptcy, and the recorded assignment is not indicative of such.
7. Privateering — Not present. No evidence of a transfer to an NPE on behalf of Dental Monitoring SAS.
8. Defensive aggregator (anti-NPE) — Not present. The patent remains with the original operating assignee.

Verdict

Insufficient data. The only recorded assignment is the initial transfer from the inventors to the original operating assignee, Dental Monitoring SAS (Reel 046648/0279, recorded 2018-10-16). There are no subsequent assignments that would indicate a change in ownership or a transfer to a non-practicing entity. Therefore, based solely on the assignment records, there is insufficient data to identify any NPE/troll patterns.

Verification Link: https://assignmentcenter.uspto.gov/

To identify the most relevant prior art for US Patent 10755409, I will examine the patent citations. The patent itself mentions two key prior art documents: US 2009/0291417 and WO 2016/066651.

Based on the patent text, here's an analysis of the cited prior art:

1. US 2009/0291417

• Full Citation: US 2009/0291417 A1 (or US '417)
• Publication/Filing Date: The patent states "US 2009/0291417 describes a method making it possible to create, then modify, three-dimensional models, particularly for the manufacture of orthodontic appliances." The publication number indicates a publication date in 2009.
• Brief Description: This prior art describes a method for creating and modifying three-dimensional models, specifically for the manufacture of orthodontic appliances.
• Potential Anticipation (35 U.S.C. § 102): US 2009/0291417 generally teaches the creation and modification of 3D dental models. This broadly covers aspects related to the "updated reference model" and "initial reference model" in US10755409, particularly in the context of orthodontic treatment planning. It could potentially anticipate the fundamental concept of generating and manipulating 3D models of dental arches for orthodontic purposes, as described in Step A) and A') of the learning base enrichment method in US10755409. Specifically, the introductory statement in US10755409 that "The most recent orthodontic treatments use images to assess the therapeutic situations" and its subsequent aim to address the need for simplifying image analysis, imply that the underlying use of models and images for assessment was known. Therefore, claims related to the initial creation and modification of 3D models as a preliminary step to the image analysis, without the specific deep learning or automated image acquisition and feedback mechanisms, might be anticipated.

2. WO 2016/066651

• Full Citation: WO 2016/066651 A1 (or WO '651)

• Publication/Filing Date: The patent consistently refers to it as "WO 2016/066651," indicating a publication year of 2016. The priority date of US10755409 is July 21, 2017, meaning WO 2016/066651 is indeed prior art.

• Brief Description: WO 2016/066651 describes a method for checking the positioning, shape, and/or appearance of a patient's teeth. This method involves creating an initial reference model, then at a later "updated instant," creating an updated reference model by deforming the initial model. This deformation is guided by "updated images" (e.g., photos or videos taken by the patient). The patent also mentions that WO 2016/066651 requires an appointment with an orthodontist to create the initial reference model.

• Potential Anticipation (35 U.S.C. § 102): WO 2016/066651 is explicitly cited as closely related and forms a significant basis upon which US10755409 builds and differentiates itself. US10755409 states that its enrichment method (steps A'-C') can comprise "one or more of the features of the steps c), d) and e) of WO 2016 066651." It also mentions that "each execution of the method described in WO 2016/066651 preferably generates more than three...updated images which, by an automated processing by means of the updated reference model, may produce as many historical images."

  Therefore, WO 2016/066651 directly anticipates several aspects of US10755409, particularly in the methods for enriching a learning base (steps A-F and A'-C') and methods for acquiring images.

  • Generation of updated reference models by deformation of an initial model: The core concept of taking an initial model and deforming it based on subsequent images (updated images) to create an updated reference model, as described in WO 2016/066651, is directly carried over into steps A' and C' of US10755409's enrichment method.
  • Acquisition of updated images by the patient: WO 2016/066651 describes patients acquiring their own "updated images," which is a key aspect of US10755409's methods.
  • Searching for virtual acquisition conditions: The process of comparing updated images with views of a reference model to find "virtual acquisition conditions" that exhibit a maximal match, including the use of metaheuristic methods (like simulated annealing), is explicitly described as being part of WO 2016/066651 and is directly incorporated or built upon in step C) and C') of US10755409.
  • Image processing for discriminating information: The extraction of discriminating information items (e.g., contours) and their representation as "maps" for comparison between images, as detailed in US10755409 (e.g., step C3), is also described as being part of WO 2016/066651.

  US10755409 primarily distinguishes itself by applying deep learning to these existing image and model processing techniques, particularly for automated analysis, attribute determination, and feedback during image acquisition. The invention in US10755409 focuses on using a deep learning device for the analysis of dental arch images to determine tooth attributes or image attributes, and the subsequent actions based on these determinations (e.g., sending messages or assessing aligner suitability). Therefore, the underlying image acquisition, model generation, and comparison techniques themselves, without the integration of deep learning for automated attribute determination and feedback, would likely be anticipated by WO 2016/066651.

  Claims related to the methods of enriching a learning base (steps A-F and A'-C') that do not explicitly involve the training and application of a deep learning device to determine attributes from images for the purpose of the learning base creation, but rather focus on the creation of segmented models and matching images, could be anticipated by WO 2016/066651. Similarly, any claim that solely covers acquiring and comparing images to models for deformation without specifying the deep learning component for attribute determination for that purpose might be anticipated.

It's important to note that the patent text for US10755409 explicitly states that WO 2016/066651 "does however require an appointment with the orthodontist in order to create the initial reference model. This appointment constitutes a brake to prevention." This highlights a problem that US10755409 aims to solve, suggesting that aspects of the new patent providing a solution to this problem (e.g., by enabling remote monitoring and feedback via deep learning analysis without requiring a new scan) would be novel over WO 2016/066651.

For a precise assessment of anticipation under 35 U.S.C. § 102, a detailed claim-by-claim analysis against the full text of both prior art documents would be necessary. This summary focuses on the explicit connections and descriptions provided within US10755409 itself.

To analyze the obviousness of US patent 10755409 under 35 U.S.C. § 103, we will consider combinations of the prior art references explicitly mentioned in the patent. The relevant prior art references are WO 2016/066651 and US 2009/0291417.

Background on the Patent's Core Contributions

US Patent 10755409 primarily introduces the application of deep learning (preferably neural networks) to various aspects of dental arch image analysis. Key methods described include:

• Detailed Analysis: Using deep learning to identify tooth zones and assign tooth attribute values.
• Enriching a Learning Base: Automating the labeling of images for deep learning training by leveraging 3D models and image matching.
• Global Analysis: Using deep learning to determine overall "image attributes" (e.g., image quality, mouth opening, appliance presence) for an entire image.
• Guiding Image Acquisition: Providing real-time feedback to users acquiring dental images by analyzing the images with deep learning.
• Assessing Aligner Shape: Detecting and measuring separation between teeth and orthodontic aligners using deep learning.

Prior Art References

1. WO 2016/066651: This reference describes a method for checking the positioning, shape, and/or appearance of a patient's teeth. It involves creating an initial 3D reference model (e.g., via scanner), acquiring 2D "updated images" from the patient (e.g., with a cellphone and "no particular precautions"), and creating an "updated reference model" by deforming the initial 3D model. This deformation is guided by the updated images, using metaheuristic methods to find virtual acquisition conditions that result in a maximal match between reference images (views of the 3D model) and the actual updated images. The method helps track tooth changes over time.

2. US 2009/0291417: This reference describes methods for creating and modifying three-dimensional models, specifically for the manufacture of orthodontic appliances.

Obviousness Analysis under 35 U.S.C. § 103

A person having ordinary skill in the art (PHOSITA) in 2017 (the priority date of US10755409) would have been familiar with deep learning and neural networks as established, powerful techniques for image recognition, classification, and attribute determination. The patent itself states that a "neural network" or "artificial neural network" is "a set of algorithms well known to a person skilled in the art."

The central inventive step in US10755409 is the application of these known deep learning techniques to various problems in dental image analysis, often building upon methods described in WO 2016/066651.

Combination: WO 2016/066651 + Known Deep Learning Techniques

Motivation to Combine:
WO 2016/066651 already addresses the acquisition and comparison of 2D dental images with 3D dental models to monitor tooth changes. However, such image processing and analysis, especially when performed repeatedly for monitoring, can be time-consuming or require significant manual input. A PHOSITA would be motivated to automate and enhance the accuracy and efficiency of these tasks using contemporary, powerful image processing techniques. Deep learning, being well-known for its capabilities in image analysis, recognition, and pattern matching, would be an obvious choice to achieve these goals. The patent itself identifies "an ongoing need for a method simplifying the analysis of the images of dental arches of patients" and a need "to rapidly enrich the learning base." These explicitly stated needs directly provide the motivation for applying deep learning to the methods of WO 2016/066651.

Application to Specific Methods of US10755409:

1. Detailed Analysis Method (Claims 1-4 general concept, 1-4 detailed steps):

   • WO 2016/066651 teaches acquiring and processing dental images to track tooth changes, which implicitly involves recognizing and localizing teeth.
   • A PHOSITA would find it obvious to replace or augment the explicit feature extraction and comparison steps of WO 2016/066651 with a deep learning device to "determine at least one value of a tooth attribute" or "determine the presence of a tooth... and of the attribute value." The motivation would be to improve the automation, speed, and accuracy of tooth identification and attribute assignment.

2. Method for Enriching a Learning Base (Steps A-F and A'-C'):

   • WO 2016/066651 describes a process where "updated images" are aligned with highly accurate 3D "updated reference models" that are segmented into "tooth models" with associated "tooth attributes."
   • The enrichment methods of US10755409 aim to automatically generate labeled historical images for deep learning training. Given the sophisticated image-to-model alignment and attribute definition in WO 2016/066651, a PHOSITA would be clearly motivated to leverage this existing framework to automate the creation of a large, labeled dataset for training deep learning models. Specifically, steps D-F (identifying reference tooth zones, transferring to updated images, and assigning attributes) are a direct automation of creating labeled data from the output of WO 2016/066651's methods.

3. Global Analysis Method (Steps 1'-3'):

   • WO 2016/066651 deals with images of dental arches, which inherently contain information about the overall image quality, mouth posture, and presence of dental appliances.
   • A PHOSITA would find it obvious to apply known deep learning image classification techniques to the dental images obtained via WO 2016/066651's methods to "determine a value for an image attribute". These attributes could include image quality (brightness, contrast, sharpness), presence of appliances, or mouth opening, as suggested by US10755409. The motivation is to get a rapid, high-level assessment of the image's overall characteristics without requiring individual tooth segmentation.

4. Method for Acquiring an Image (Steps a'-e'):

   • WO 2016/066651 explicitly discusses the challenges of patient-acquired images taken "without any particular precautions" and the need for rough assessment of acquisition conditions to speed up processing. It notes that operators might make mistakes, such as inverting images or forgetting views.
   • A PHOSITA, seeking to improve the reliability and efficiency of this image acquisition process, would be motivated to integrate an automated checking and feedback system. Using a deep learning device (trained as per the detailed/global analysis methods) to determine image attributes (e.g., orientation, quality, presence of appliance) and compare them to instructions, then providing feedback to the user to guide them to acquire new, satisfactory images, is a logical and obvious enhancement to the system described in WO 2016/066651. The aim is to ensure high-quality, relevant images for subsequent analysis.

5. Method for Assessing the Shape of an Orthodontic Aligner (Steps a''-d''):

   • WO 2016/066651 focuses on "checking the positioning and/or the shape and/or the appearance of teeth," implying a comparison of the current state against a target or expected state, which would include the fit of an orthodontic appliance.
   • US 2009/0291417 provides further background on the creation and modification of 3D models for orthodontic appliances.
   • Given that WO 2016/066651 already processes images of teeth, potentially with appliances, it would be obvious for a PHOSITA to train a deep learning model to specifically detect and quantify "separation between the tooth... and an aligner worn by said tooth". Deep learning is well-suited for identifying such visual discrepancies in images. The motivation is to automate and improve the accuracy of aligner fit assessment, enabling remote monitoring and reducing the need for in-person orthodontist appointments.

In conclusion, the various methods described in US10755409, which involve applying deep learning to dental image analysis, learning base enrichment, image acquisition guidance, and aligner assessment, would have been obvious to a person having ordinary skill in the art by combining the teachings of WO 2016/066651 (and background knowledge from US 2009/0291417 regarding orthodontic models) with the well-known capabilities of deep learning and neural networks. The clear motivations for such a combination include automation, increased efficiency, improved accuracy, and enabling remote monitoring in dental and orthodontic care.