DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Notice to Applicant
This office action is in response to application filed on 01/18/2025.
Limitations appearing inside of {} are intended to indicate the limitations not taught by said prior art(s)/combinations.
Claims 1-5 are pending in the application.
Information Disclosure Statement
No Information Disclosure Statement (IDS) was filed; therefore, no applicant-submitted references were considered.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-5 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim 1 recites the limitation "the last layer" in line 22. It is unclear which layer is the last layer. There is insufficient antecedent basis for this limitation in the claim.
Claim 1 recites “the dense layer” in lines 24-25. It is unclear which dense layer is being referred to. Is it “a fully connected dense layer” (line 24), the first, second or third dense layers (lines 11, 13, and 15, respectively).
Claim 1 recites “a previous layer” in line 25. It is unclear which layer is being referred to: weighted global average pooling layer (line 7), first, second and third dense layer (lines 11, 13, and 15, respectively), the last layer (line 22), a fully connected dense layer (line 24), the dense layer (line 25).
Claim 4 recites “the previous layer” in line 2. It is unclear which layer is being referred to: weighted global average pooling layer (line 7), first, second and third dense layer (lines 11, 13, and 15, respectively), the last layer (line 22), a fully connected dense layer (line 24), the dense layer (line 25), the previous layer (line 25) , a last fully connected layer (line 27).
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claims 1-3, and 5 are rejected under 35 U.S.C. 103 as being unpatentable over “Ghosh” (Ghosh, Susmita, and Abhiroop Chatterjee. "Introducing feature attention module on convolutional neural network for diabetic retinopathy detection." arXiv preprint arXiv:2308.02985 (2023).) in view of “Akbar” (Akbar M, Ullah M, Shah B, Khan RU, Hussain T, Ali F, Alenezi F, Syed I, Kwak KS. An effective deep learning approach for the classification of Bacteriosis in peach leave. Front Plant Sci. 2022 Nov 24;13:1064854. doi: 10.3389/fpls.2022.1064854. PMID: 36507379; PMCID: PMC9730539.).
1. Ghosh teaches a method for detecting diabetic retinopathy by integrating a deep channel attention mechanism (DCAM) into a pre-trained VGG19 Convolutional Neural Network (CNN), comprising:
receiving features from the pre-trained VGG19 Convolutional Neural Network (CNN) (features obtained through the pretrained VGG19 as input; Ghosh, [§III, p2, col 1, ¶1]) ;
computing weighted averages of one or more input feature maps along the spatial dimensions thereof by a weighted global average pooling layer (the first layer applied is the Weighted Global Average Pooling which takes the features obtained through the pretrained VGG19 as input. It computes the weighted average of the feature maps along the spatial dimensions, resulting in a tensor with reduced spatial dimension (batch_size, 1, 1, channels); Ghosh, [§III, p2, col 1, ¶2])
generating an input feature vector with reduced spatial dimensions (It computes the weighted average of the feature maps along the spatial dimensions, resulting in a tensor with reduced spatial dimension (batch_size, 1, 1, channels); Ghosh, [§III, p2, col 1, ¶2]),
wherein the weighted averages of the input feature maps are computed using learnable weights of each channel (channel attention; Ghosh, [§III, p2, col 2 , ¶1]);
providing at least three dense layers in the DCAM, (See Fig 2, shown below, exhibits two dense layers and an attention feature tensor which are all interpreted as dense layers; Ghosh, [§III, p2, col 2]
PNG
media_image1.png
366
397
media_image1.png
Greyscale
),
the layers comprising:
a first dense layer with a rectified linear unit (ReLU) activation function configured to reduce the number of channels to C/ratio(The first one uses a rectified linear unit (ReLU) activation function reducing the number of channels to chan_dim/ratio.; Ghosh, [§III, col 1, ¶3]);
a second dense layer with a rectified liner unit (ReLU) activation function (The second dense layer utilizes a sigmoid activation function to produce an output tensor; Ghosh, [§III, p2, col 1, ¶3]) configured to reduce the number of channels to C/(ratio*2) (fc2 represents the output of the second dense layer (Eq. 3). It further reduces the dimensionality of the input by multiplying the features from the first dense layer fc1(i,1,1,d) with the corresponding weights W2; Ghosh, [§III Feature Attention Mechanism:, p 3, col 2. ¶2]; and See Fig 2, shown above, exhibits the second dense layer with ); and
a third dense layer {with a sigmoid activation function} to generate an output vector, wherein dimensions of the output vector of the third dense layer is (batch size, 1, 1, C), (See ATTENTION FEATURE block of Fig 2 is the third dense layer, and “resulting in a tensor with reduced spatial dimension (batch_size, 1, 1, channels”); Ghosh, [§III, p2, col 1, ¶2]),
wherein C is the channel and ratio is a tuneable parameter (C represents the number of input channels; Ghosh, [§III Feature Attention Mechanism:, p3, col 1 ¶3]),
generating an attention feature vector by multiplying the weighted averages and the input feature vector, wherein the multiplication is an element-wise multiplication (attention weights are generated to provide a measure of relevance/importance of each channel … these weights are then applied to the feature map using element-wise multiplication; Ghosh, [§III Feature Attention Mechanism:, p2, col 2, ¶3]);
adding the attention feature vector obtained from the element-wise multiplication and the extracted input feature vector from the last layer of the pre -trained VGGl19 architecture to generate a residual connection (See Equation 5, shown below, exhibits adding the attention feature vector a_f with the input feature vector i_f; Ghosh,[§III. Attention Feature Mechanism:, p3, col 2, ¶5]:
PNG
media_image2.png
76
392
media_image2.png
Greyscale
),
wherein a skip connection is utilized for adding the said features (a skip connection is incorporated to add the attention feature tensor obtained from the element-wise multiplication to the input feature tensor; Ghosh, [§III, p2, col 2, ¶1]);
providing a fully connected dense layer with the ReLU activation function, the dense layer connects every neuron from the previous layer to every neuron available therein (See Equation 2, shown below, fully connected layer fc1 contains a RELU function; Ghosh, [§III Feature Attention Mechanism:, p 3, col 1, ¶5]:
PNG
media_image3.png
69
399
media_image3.png
Greyscale
); and
providing a last filly connected dense layer with 5 neurons corresponding to the number of classes in a classification task(There are five classes in the image dataset, Ghosh, [§IV. (A), p 3, col 2, ¶1]).
Ghosh discloses a third dense layer to generate an output vector, but does not explicitly disclose a third dense layer with a sigmoid activation function.
However, Akbar, a similar field of endeavor of deep learning approach for classification using a VGG19 model, teaches a third dense layer with a sigmoid activation function (Three Fully Connected (FC) layers. The final layer was the Sigmoid layer; Akbar, [Table 2 caption], and see arrow in Fig 2, shown below:
PNG
media_image4.png
520
749
media_image4.png
Greyscale
).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to include a sigmoid activation function on the third layer as taught by Akbar to the invention of Ghosh. The motivation to do so would be because the sigmoid function prevents output value overflow and is saturated, where a precise prediction of X above 2 or less typically places the value of Y very close to 1 or 0 at the edge of the curve shown in Figure 8, allowing precise predictions..
2. The combination of Ghosh and Akbar teaches the method of claim 1.
Ghosh further teaches wherein the dimensions of the output vector of the third dense layer and the first dense layer are same (The output tensor has the same shape as that of the input feature, i.e., (batch_size, height, width, chan_dim), Ghosh, [§III, p2, col 2, ¶1]).
3. The combination of Ghosh and Akbar teaches the method of claim 1.
Ghosh further teaches further comprising the step of image preprocessing applied to plurality of retinal fundus images obtained from a dataset (The APTOS dataset consists of a large collection of high-resolution retinal fundus images; Ghosh, [§IV. A., p3, col 2, ¶1]).
5. The combination of Ghosh and Akbar teaches the method of claim 1.
Ghosh further teaches wherein the VGG19 Convolutional Neural Network is pre- trained on an APTOS DR Dataset (the proposed neural network model is conducted using APTOS dataset; Ghosh, [§IV, p3, col 2, ¶1]).
Claim 4 is rejected under pre-AIA 35 U.S.C. 103(a) as being unpatentable over Ghosh in view of Akbar, and further in view of “Choudhary” (Choudhary A, Ahlawat S, Urooj S, Pathak N, Lay-Ekuakille A, Sharma N. A Deep Learning-Based Framework for Retinal Disease Classification. Healthcare (Basel). 2023 Jan 10;11(2):212. doi: 10.3390/healthcare11020212. PMID: 36673578; PMCID: PMC9859538.).
4. The combination of Ghosh and Akbar teaches the method of claim 1.
The combination does not explicitly teach wherein the activation function is a softmax activation function, wherein the activation function is a softmax activation function, the softmax activation function converts the raw scores from the previous layer into class probabilities.
However, Choudhary, a similar field of endeavor of retinal disease classification using VGG-19 architecture teaches wherein the activation function is a softmax activation function, (output dense layer (None, 4), which is used for the prediction of output with softmax activation; Choudhary [§3.1, ¶2]), the softmax activation function converts the raw scores from the previous layer into class probabilities (Softmax activation helps in multiclass classification and works on relative probabilities; Choudhary, [§3, ¶¶3-4]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to include a softmax activation function that converts the raw scores to class probabilities as taught by Choudhary to the combined invention of Ghosh and Akbar. The motivation to do so would be because softmax activation helps in multiclass classification .
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. See PTO-892 Notice of References Cited for full citations.
Das et al., (2024) teaches diabetic retinopathy detection in retinal fundus images using VGG19 architecture, and would been relied upon for teaching limitations in the independent and dependent claims.
Jha et al., (2023) teaches diabetic retinopathy detection in retinal fundus images using VGG16 architecture, and would been relied upon for teaching limitations in the independent and dependent claims.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to CHANDHANA PEDAPATI whose telephone number is 571-272-5325. The examiner can normally be reached M-F 8:30am-6pm (ET).
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Chan Park can be reached at 571-272-7409. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/CHANDHANA PEDAPATI/Examiner, Art Unit 2669 /CHAN S PARK/Supervisory Patent Examiner, Art Unit 2669