Patent Drafting Analysis of FUJIFILM Corporation’s Cell Differentiation Prediction Apparatus | US 2024/0112479 A1
Patent Drafting Analysis of FUJIFILM's Cell Differentiation Prediction Apparatus | US 2024/0112479 A1
A structural and strategic analysis of FUJIFILM's image-based stem cell differentiation prediction patent, examining claim architecture, drafting quality, §101 eligibility posture, critical gaps, and prosecution positioning across apparatus, method, and CRM claim types.
Spec Words
8,450
Across 6 sections
Draft now ↗
Total Claims
12
3 independent · 9 dependent
Draft now ↗
Figure Sheets
18
System architecture, cell imaging, processing flows, feature extraction
Draft now ↗
Published by PatSnap Insights Team · · 11 min read Verified by PatSnap Eureka Data
Overview
Structural Overview
The detailed description dominates at approximately 61% of total estimated words (~5,200 of ~8,450), providing extensive technical support across 20 figures and 117 numbered paragraphs. The claim set is compact at 12 total claims (3 independent, 9 dependent), establishing apparatus, method, and non-transitory computer-readable storage medium (CRM) coverage. The 18 figure sheets provide comprehensive coverage of system architecture, cell imaging workflows, feature extraction algorithms, and statistical validation data including ROC curves and AUC performance tables.
Section Word Distribution
↗ Click bars to exploreFigure Inventory — 18 Sheets
| Figure | Description | Role |
|---|---|---|
| FIG. 1 | Schematic diagram of the microscope observation system 2, showing information processing apparatus 10 connected to microscope apparatus 20 with mount 21, light source 22, imaging apparatus 23, and driving unit 24.Search in Eureka ↗ | System architecture |
| FIG. 2 | Diagram of imaging target ranges IR (1 through N) covering the circular imaging region R of a cell-culture container, showing the serpentine scan path of the imaging apparatus.Search in Eureka ↗ | Key embodiment |
| FIG. 3 | Block diagram of internal hardware configuration of information processing apparatus 10, showing CPU 42, memory 41, storage device 40, communication unit 43, display 11, and input device 14.Search in Eureka ↗ | System architecture |
| FIG. 4 | Functional block diagram of information processing apparatus 10, illustrating imaging control unit 50, RW control unit 51, processing unit 52, feature quantity extraction unit 53, prediction unit 54, and display control unit 55 operating with storage device 40.Search in Eureka ↗ | Key embodiment |
| FIG. 5 | Diagram showing processing unit 52 feeding input image PI through trained model LM to generate output image PO with semantic segmentation into three classes: cell nucleus (class 1), cytoplasm (class 2), and culture medium (class 3).Search in Eureka ↗ | Claim support |
| FIG. 6 | Diagram of feature quantity extraction unit 53 computing average number of cells (50) as feature quantity F by counting cells in each of output images PO1 through PON and dividing their sum by N.Search in Eureka ↗ | Claim support |
| FIG. 7 | Diagram illustrating the determination criterion K as a boundary value (cutoff) between distributions DP (successful differentiation) and DN (unsuccessful differentiation) plotted against feature quantity F, showing TP, TN, FP, and FN regions.Search in Eureka ↗ | Claim support |
| FIG. 8 | Diagram of a receiver operating characteristic (ROC) curve showing the relationship between true positive rate (TPR) and false positive rate (FPR), with the area under the curve (AUC) shaded.Search in Eureka ↗ | Other |
| FIG. 9 | Flowchart of the differentiation prediction process (steps S10–S18), from causing the microscope apparatus to perform imaging through acquiring input images, generating output images, counting cells, deriving average cell number as feature quantity, and comparing to determination criterion K to output a prediction result.Search in Eureka ↗ | Flow diagram |
| FIG. 10 | Block diagram of training unit 60 showing how stained cell images TD1 and not-stained cell images TD2 constitute training data TD input to training model M, with adjustment unit 62 comparing output images PO to stained images TD1 to produce trained model LM stored in storage device 40.Search in Eureka ↗ | Key embodiment |
| FIG. 11 | Flow diagram showing the three-phase cell culture process: P1 (start of cell culture), repeated P2 (predict differentiation), and P3 (induce differentiation), illustrating the timing of prediction relative to differentiation induction.Search in Eureka ↗ | Flow diagram |
| FIG. 12 | First modification diagram showing processing unit 52 generating output image PO with four-class semantic segmentation: cell nucleus (class 1), cytoplasm (class 2), culture medium (class 3), and central region of cell nucleus (class 4, shown as 31C).Search in Eureka ↗ | Key embodiment |
| FIG. 13 | Second modification flow diagram showing alternating P2A (perform imaging) and P2B (predict differentiation) sub-processes during the culture period, culminating in P3 (induce differentiation).Search in Eureka ↗ | Flow diagram |
| FIG. 14 | Second modification feature quantity extraction diagram showing computation of increase rate NR = log2(NT2/NT1) from two sets of output images PO1–PON at different imaging times, yielding total cell counts NT1 (1000) and NT2 (1500) and feature quantity NR (0.585).Search in Eureka ↗ | Claim support |
| FIG. 15 | Third modification feature quantity extraction diagram showing per-region feature quantities F extracted individually from output images PO1 through PON, where the number of cells in each image is used as a separate feature quantity.Search in Eureka ↗ | Claim support |
| FIG. 16 | Display example for the third modification showing prediction results superimposed on the cell-culture container map, with hatched regions indicating successful differentiation prediction and dashed regions indicating unsuccessful prediction, yielding ~26% successful area.Search in Eureka ↗ | UI/interface |
| FIG. 17 | Table comparing AUC, sensitivity, and specificity for 19 different feature quantities F (including increase rate in number of cells, number of cells, nucleus area, N/C ratio, circularity, and confluency, each with overall, average, and standard deviation variants).Search in Eureka ↗ | Other |
| FIG. 18 | Graphs comparing confluency frequency distributions before and after differentiation induction for a case where confluency variation is large, showing the emergence of a low-density region after induction.Search in Eureka ↗ | Other |
| FIG. 19 | Graphs comparing confluency frequency distributions before and after differentiation induction for a case where confluency variation is small, illustrating that low-density region is not generated and differentiation is highly likely to be successful.Search in Eureka ↗ | Other |
| FIG. 20 | Scatter plot illustrating the correlation between Nodal gene expression level and nucleus area (median) for 29 batches of iPS cells induced to differentiate into cardiac muscle, with determination criterion K at ~142 μm², showing successful vs. unsuccessful differentiation outcomes.Search in Eureka ↗ | Claim support |
Analysis powered by PatSnap Eureka. Patent text and figures publicly available from USPTO. Draft a Similar Patent
Claims
Claim Architecture Analysis
The claim set contains 3 independent claims: Claim 1 (apparatus), Claim 11 (method), and Claim 12 (non-transitory CRM), providing tripartite enforcement coverage. The dependent:independent ratio of 3.0:1 is below the norm for this bio-imaging IPC class, suggesting limited fallback depth. The structural strategy of maintaining functionally identical core limitations across all three independent claim types while differentiating dependent claims only under the apparatus claim (Claims 2–10) creates a methodological gap where the rich dependent claim fallbacks are not mirrored in the method or CRM branches.
Core inventive concept: Claims 1, 11, and 12 each recite an apparatus/method/program that predicts success or failure of cell differentiation by "extracting a feature quantity based on an appearance of the cells by performing image processing on the captured image" and "predicting the success or failure of differentiation of the cells on the basis of the extracted feature quantity" — solving the problem that prior art (WO2019/240008A) required complex cell information and process history data by instead relying solely on the visual appearance of cells in a microscope image to make the differentiation prediction.
Independent Claim Dissection
| Claim | Preamble | Transition | Key Body Elements |
|---|---|---|---|
| Claim 1 | An information processing apparatus that predicts a success or failure of differentiation of cells on the basis of a captured image obtained by imaging a cell population cultured in a cell-culture container | comprising | at least one processor configured to: extract a feature quantity based on an appearance of the cells by performing image processing on the captured image; predict the success or failure of differentiation of the cells on the basis of the extracted feature quantitySearch prior art ↗ |
| Claim 11 | An information processing method for predicting a success or failure of differentiation of cells on the basis of a captured image obtained by imaging a cell population cultured in a cell-culture container | comprising | extracting a feature quantity based on an appearance of the cells by performing image processing on the captured image; predicting the success or failure of differentiation of the cells on the basis of the extracted feature quantitySearch prior art ↗ |
| Claim 12 | A non-transitory computer-readable storage medium storing a program for causing a computer to execute a process of predicting a success or failure of differentiation of cells on the basis of a captured image obtained by imaging a cell population cultured in a cell-culture container | comprising | extracting a feature quantity based on an appearance of the cells by performing image processing on the captured image; predicting the success or failure of differentiation of the cells on the basis of the extracted feature quantitySearch prior art ↗ |
Claim Dependency Tree
1 Apparatus claim — processor configured to extract feature quantity from cell image and predict differentiation success/failureSearch Claim 1 prior art ↗
2 Adds: prediction based on only the feature quantity based on appearance of cellsSearch in Eureka ↗
3 Adds: processor detects cell nuclei from at least one captured image and extracts information related to cell nuclei as feature quantitySearch in Eureka ↗
4 Further: feature quantity includes number of cells, area of cell nucleus, N/C area ratio, or proportion of cell-occupied area (confluency) (depends on Claim 3)Search in Eureka ↗
5 Further: feature quantity includes increase rate in number of cells calculated from plurality of captured images at different image-capturing times (depends on Claim 3)Search in Eureka ↗
6 Further: cell nuclei detected using trained model from machine learning using fluorescence staining images as training data (depends on Claim 3)Search in Eureka ↗
7 Adds: predicts success/failure for each of plurality of regions into which cell-culture container is divided, using per-region captured imagesSearch in Eureka ↗
8 Further: processor calculates ratio of area for which differentiation is successful to cell-culture area (depends on Claim 7)Search in Eureka ↗
9 Adds: cells are stem cellsSearch in Eureka ↗
10 Further: stem cells are in an undifferentiated state (depends on Claim 9)Search in Eureka ↗
11 Method claim — extracting feature quantity and predicting differentiation from captured cell imageSearch Claim 11 prior art ↗
12 CRM claim — non-transitory computer-readable storage medium for cell differentiation prediction processSearch Claim 12 prior art ↗
| Metric | This Application | Med. Device / Bio-Imaging Industry Norm |
|---|---|---|
| Total claims | 12 | 15 – 25 |
| Independent claim count | 3 | 2 – 4 |
| Dependent : Independent ratio | 3.0 : 1 | 5 – 8 : 1 |
| Method claims present? | Yes — Claim 11 | Usually yes |
| System / apparatus claims? | Yes — Claim 1 | Always |
Analysis powered by PatSnap Eureka. Patent text and figures publicly available from USPTO. Draft a Similar Patent
Drafting Quality
Drafting Quality Signals
The claim set benefits from a clean tripartite structure (apparatus/method/CRM) with well-grounded technical content, and the detailed description provides strong support through 20 figures including empirical AUC validation data in FIG. 17. However, the independent claims' extreme functional minimalism — Claim 1 recites only two functional processor steps with no structural constraints — creates significant §101 exposure and leaves the claims vulnerable to prior art attacks that the sparse dependent fallback chain (concentrated entirely under Claim 1) cannot adequately address for Claims 11 and 12.
✅
Antecedent Basis
The claim set is clean with respect to antecedent basis. Claim 1 introduces "at least one processor" and subsequent dependent claims 2 through 10 uniformly reference "the at least one processor," providing proper antecedent basis throughout. Claims 11 and 12 use method/CRM gerund constructions that do not rely on article-referencing, avoiding antecedent basis issues entirely. No "the [element]" references appear without prior introduction.
✅
Spec–Claim Consistency
The specification maps comprehensively onto all claim limitations. FIG. 4 (¶[0057]) directly supports the processor architecture in Claim 1. FIG. 5 and ¶[0060]–[0064] support the "image processing" and "feature quantity extraction" limitations. FIG. 6 and ¶[0065]–[0066] support Claim 4's cell-count feature quantity. FIG. 10 and ¶[0077]–[0081] support the trained model limitation in Claim 6. FIG. 17 provides empirical validation of the AUC performance for multiple feature quantity variants recited in Claims 4 and 5.
✅
Transition Word Usage
All independent claims and dependent claims use "comprising" as the transition, which is the strategically correct open-ended transition for this technology space where additional functional components (e.g., additional image processing modules, display components) will inevitably be present in any infringing embodiment. The use of "comprising" at both the preamble level and the processor step level in Claim 1 is appropriate. No missed opportunity exists for "consisting essentially of" narrowing, as the claims are correctly kept open.
⚠️
§112(f) Means-Plus-Function Risk
No explicit "means for" language appears in any claim, reducing formal §112(f) exposure. However, Claim 1's recitation of "at least one processor configured to: extract a feature quantity... and predict the success or failure" uses purely functional claiming language that — while not triggering §112(f) presumption — may receive examiner scrutiny as to whether sufficient structural definition is provided. The dependent claims cure this somewhat: Claim 6 grounds the trained model in a specific neural network structure (CNN, U-Net, SegNet, ResNet as described in ¶[0059]), but this detail is only reached via a dependent claim rather than being anchored in the independent claims.
⚠️
§101 Eligibility Risk
Claims 1, 11, and 12 face meaningful Alice/Mayo exposure because the core innovation — "extracting a feature quantity" and "predicting success or failure" — can be characterized as an abstract idea of data analysis applied to biological observations. The hardware tie-in in Claim 1 ("at least one processor") provides a Step 2A, Prong 1 defense, but the claimed steps do not recite any unconventional hardware arrangement. Claim 6's trained machine learning model (CNN/U-Net/SegNet) provides the strongest §101 defense element but is only in a dependent claim. A stronger filing would have included at least one structural tie-in to the imaging hardware or the trained model architecture in each independent claim.
⚠️
Dependent Claim Fallback Quality
The 9 dependent claims (Claims 2–10) all depend from apparatus Claim 1, creating a structural asymmetry: method Claim 11 and CRM Claim 12 have zero dependent claims, leaving them completely without fallback positions. Within the apparatus chain, Claims 4 and 5 add meaningful technical specificity (specific feature quantity types: cell count, N/C ratio, confluency, increase rate), and Claims 7–8 add the per-region prediction and area-ratio calculation that support the commercial embodiment shown in FIG. 16. However, Claims 9–10 (stem cells / undifferentiated state) are weak fallbacks that merely narrow to the intended use case without adding patentable technical distinction.
⚠️
Abstract Quality
An examiner reading only the abstract may correctly identify the core apparatus claim — the abstract accurately states that the processor "extracts a feature quantity based on an appearance of the cells by performing image processing" and "predicts the success or failure of differentiation of the cells on the basis of the extracted feature quantity." However, the abstract omits the key differentiating elements that distinguish this from the prior art WO2019/240008A — specifically, that the prediction is based solely on appearance (no gene information or process history required) and that cell nuclei detection via a trained ML model drives the feature extraction. These omissions may slow examiner orientation to the novelty.
✅
Figure Support Quality
The 18 figure sheets provide excellent coverage of all claim limitations. FIG. 4 supports the processor/functional unit architecture of Claim 1; FIG. 5 supports the image-processing/segmentation step; FIG. 6 supports the feature quantity extraction (average cell count); FIG. 7 supports the determination criterion K used in prediction; FIGS. 14–15 support the increase-rate feature quantity of Claim 5 and the per-region prediction of Claim 7; FIG. 17 provides empirical AUC data for the feature quantity variants in Claims 4–5. No claim limitation appears to lack figure support. The empirical data in FIG. 17 and FIG. 20 is particularly valuable as written description support for enablement.
Analysis powered by PatSnap Eureka. Patent text and figures publicly available from USPTO. Draft a Similar Patent
Scorecard
Strategic Intent Scorecard
Multi-dimensional assessment of this application's patent strategy quality, based on claim structure, specification depth, and prosecution positioning.
Claim Breadth
3.5
Prosecution Defensibility
3
Spec–Claim Consistency
4.5
Dependent Claim Coverage
2.5
Claim Type Diversity
4
Figure Support Quality
4.5
Key observation: Spec–Claim Consistency and Figure Support Quality are joint highest-scoring dimensions (both 4.5/5.0) because the 18-figure set and 117-paragraph detailed description provide exhaustive empirical and structural support for every claim limitation, including AUC validation data in FIG. 17 that directly evidences the claimed feature quantity effectiveness. The lowest-scoring dimension is Dependent Claim Coverage (2.5/5.0), caused by the structural weakness that all 9 dependent claims depend only from apparatus Claim 1, leaving method Claim 11 and CRM Claim 12 completely unsupported by fallback positions — a practitioner prosecuting this application should immediately file continuation claims adding dependent method and CRM claims mirroring the apparatus fallback chain.
See how your own draft compares — Open Eureka IP Drafting →
Critical Gaps
3 Critical Gaps in This Claim Set
A senior-attorney lens on the three highest-priority structural weaknesses — what each exposes in prosecution and litigation, and what a stronger filing would have done differently.
🔒
3 Critical Gaps in This Claim Set
See the full attorney-level analysis of what this application leaves unprotected — and how to draft it more defensively for your own filings.
No dependent fallbacks for method/CRM claims Trained ML model absent from independent claims No integrated microscope-processor system claim
Unlock Full Analysis — Free Frequently asked questions
US 2024/0112479 A1 — key questions answered
Still have questions? PatSnap Eureka can answer them from patent data instantly. Search in Eureka
Disclaimer: This analysis is generated by PatSnap Eureka AI based on publicly available patent data from the USPTO. It does not constitute legal advice and should not be relied upon as such. Patent data may be subject to change as prosecution progresses. Scores and assessments reflect automated analysis and may not capture all relevant legal or technical nuances. Always consult a qualified patent attorney for formal legal opinions on patentability, freedom to operate, or infringement.
Ask anything about this patent.
PatSnap Eureka searches patents and data to answer instantly.
PatSnap Eureka searches patents and data to answer instantly.
Powered by PatSnap Eureka
Link copied to clipboard