SHANGHAI UNITED IMAGING INTELLIGENCE CO., LTD. (Shanghai, CN)

The present disclosure describes a system and method for classification determination of a structure. The method may include obtaining images that represent a the structure of an object. This technique can also be utilized to identify various possible classifications for the structure, and their probabilities, by putting the data in an appropriate classification model. A backbone network is utilized to determine a structure’s backbone feature as well as a segmentation system is utilized to determine a structure’s segmentation feature. The model for classification also includes a density-based classification network that is used to determine a structure’s density feature. The goal classification for the structure could be determined using at least one of the probabilities derived from the plurality of possible classifications.

Medical imaging for diagnosis and treatment may be implemented by systems including, e.g., a X-ray imaging system, a positron emission tomography (PET) system, a magnetic resonance (MR) system, a computed tomography (CT) system, a single-photonemission computed tomography (SPECT) system, a radioisotope imaging system, etc. CT is an example of a frequently utilized diagnostic method. However, image analysis by a healthcare provider, e.g., a doctor, may be time consuming, and/or introduceinconsistencies or errors caused by differences between healthcare providers. Computer-aided diagnosis (CAD) is a method that has been used to improve diagnostic accuracy and reduce the workload of doctors, was developed in recent years. However, structures, suchas nodules pleural or pulmonary of small sizes are not yet able to distinguish their classfications (e.g. malignancies). Therefore, it is desirable to develop the system and procedure that automates the classification of structures.

A system is described in the first aspect of the disclosure. The system can comprise at least one storage media including a set of instructions, as well as at least one processor that is connected to the storage medium. Whenexecuting the set of instructions, the at least one processor may be configured to direct the system to perform operations including obtaining image data representing a structure of a subject, determining a plurality of candidate classifications of thestructure and their respective probabilities by inputting the image data into a classification model, wherein the classification model includes a backbone network for determining a backbone feature of the structure, a segmentation network for determininga segmentation feature of the structure, and a density classification network for determining a density feature of the structure, and determining a target classification of the structure based on the probabilities of the plurality of candidateclassifications.

In some embodiments, the backbone network can include several down-sampling layers. Each down-sampling layer may include a convolution layer and a batch normalization.

In certain embodiments, determining a plurality of candidate classifications of the structure inputting the image data into a classification model might include obtaining the backbone feature the segmentation feature and the density feature, by inputting the image data into the backbone network, the segmentation network, or the density classification network in turn, and then the determination of a probability for each of the many candidate classifications of the structure based on the backbone feature, the segmentation feature, and the density feature.

In certain methods, the process of determining the probability that every structure is classifiable according to its backbone feature and segmentation feature, as well as the density feature may also include the identification featureof images. This will combine the backbone feature, segmentation and density features and determine the probability of every structure being classified based on the identification feature that is identified in image data.

In some embodiments, the determination of an identification feature of the image data combing the backbone feature, the segmentation feature and the density feature could involve converting the backbone features as well as the segmentation feature and the density feature into a one-dimensional backbone feature vector as well as a one-dimensional segmentation vector and a one-dimensional feature vector, respectively, making a one-dimensional identification vector by splicing the one-dimensional backbone feature vector as well as the one-dimensional segmentation vector as well as the one-dimensional feature vector, then identifying the one-dimensional identification feature vector as the identifying feature of the image data.

Certain models allow the model of classification to be trained by using the focal loss function. At at least one weight in every focal loss function is related to one candidate classification. The focal loss function is related to the other number of.

In some embodiments, the process of obtaining images that represent the shape of a subject might consist of obtaining original image data including an image of the structure of the subject, and determining the image data by preprocessing the original image data.

In some embodiments, the determining the image data by processing the original image data might include creating a resampled picture by resampling the original image data in accordance with a resampling resolution, segmenting the image cropped in accordance with a central point of the structure, and determining the image data using normalization of the crop sizes of the image in accordance with a normalizing function.

A second aspect of the disclosure provides an apparatus. The system may comprise at least 1 storage medium, a set of instructions, and at most one processor communicating with the storage medium. The at least one processor is able to direct the system to perform the instructions. This includes obtaining the preliminary class model as well as developing a classification model in order to determine a pluralityofcandidate classifications of a subject illustrated in images. Each of these tasks can be accomplished by training the initial model with a focal loss function. A minimum of one weight of every focal loss function is related to one of several possible classifications.

A third aspect of this disclosure is a method. This method can be applied to a computing device that has at least one processor, as well as at least one computer readable storage device. The method may include obtaining, by thecomputing device, image data representing a structure of a subject, determining, by the computer, a plurality of candidate classifications of the structure as well as their probabilities of each by entering the data into a classification model, wherein the classification model includes the backbone network that determines the backbone of the structure and a segmentation network for determining a segmentation feature of the structure as well as a density classification network that determines a density feature of the structure and determining, using the computer, a target classification of the structure based on the probabilities of the various candidates for classification.

Additional characteristics will be explained in part in the following description. Parts of these features will be obvious to experienced art professionals when they look at the following and the drawings that accompany it. Other features could be made evident by the production of or operation of the examples. The advantages of this disclosure can be accomplished and achieved through practice or use of various aspects of the methods instruments, combinations and methods described in the specific examples below.

Click here to view the patent on USPTO website.

 

Get Patents with PatentPC

What is a patent?

A patent is issued by the government in order to protect an invention. It gives the inventor the right to create, use and market the idea. Society benefits when new technologies are introduced for sale. These benefits could be directly realized when individuals are able to accomplish previously unattainable feats or indirectly through the opportunities for economic growth which innovation can bring (business growth, jobs).

Patent protection is demanded by a variety of pharmaceutical companies and university researchers for research and development. Patents are granted for a product, process, or method of creating new materials. In order to be granted protection under a patent the invention must be useful unique, innovative, and not apparent to others within the same area.

Patents are awarded to inventors who have commercially successful inventions. They act as a motivator for inventors to create. Small companies and inventors can rest sure that they will earn an income from the investment they make in technology development through patents. They can earn a living by their work.

Patents play a vital role in businesses with the ability to:

Protect new products and services that are innovative;

Improve the value, the appearance, and visibility of your products on the market

Differentiate yourself and your products from the rest.

Find out about business and technical information.

Avoid the risk of accidentally using proprietary third-party content, or losing your valuable information, innovative outputs, or other creative output.

Patents transform the knowledge of inventors into a marketable asset, which opens up new opportunities for employment creation through licensing and joint ventures.

Investors who are involved in the development and commercialization of technology will find small companies with patent protection more appealing.

Patenting may lead to the creation of new ideas and inventions. This information can encourage the development of new ideas and could qualify for protection under patents.

Patents can serve as a deterrent to untrustworthy third parties who profit from an invention’s efforts.

Patent-protected technology that is commercially profitable could be used to fund technological research and development (R&D) which can increase the chance of better technology in future.

Intellectual property ownership is a way to convince lenders and investors that there are genuine opportunities to commercialize your product. One patent that is powerful could open the door for numerous financing options. Patents can be used along with other IP assets as collateral or security to secure financing. Investors may also look at your patent assets in order to increase the value of your company. Forbes and other publications have pointed out that each patent could increase company valuation by anywhere from $500,000 to $1 million.

Startups require a well-designed business plan that leverages the IP to demonstrate that your product or service is distinctive, superior, or innovative. Investors will also be impressed if you have IP rights are secured or in the process to being secure, and if they agree with your business plan.

It is essential to keep your invention secret until you apply to protect it with patents. Public disclosure of an invention before it is filed can typically degrade the novelty of an invention and make it ineligible for patent protection. Pre-filing disclosures, such as for investors, test marketing, or any other business partners, must be done after signing a confidentiality contract.

There are a variety of patents. Understanding the different types of patents is vital to safeguard your invention. Patents for utility are used to protect the development of new methods and machines. Design patents cover ornamental designs. Utility patents are best to protect the proprietor from copies and competitors. In most cases they are granted to modify or improve existing inventions. They can also be used to improve or modify existing inventions. A process patent could be a way to describe the actions or methods to perform a particular action. A chemical composition will include the combination of ingredients.

What is the length of time a patent will last? Patents that are utility-related last for 20 years from the initial filing dates, but their expiration dates can be extended because of delays at the patent office for instance.

Do you want to protect your idea? As patents are only granted for applicants who are first to file, you need to file quickly – call PatentPC to speak with a patent attorney PatentPC to protect your idea now!

When you are writing your patent application when you are writing a patent application, it is important to conduct an internet search for patents, since the search will give you some insight into other people’s concepts. You’ll be able to reduce the nature of your idea. In addition, you can learn about state of the art in your area of invention. This will allow you to comprehend the scope of your invention and prepare you to file your patent application.

How to Search for Patents

The first step in obtaining your patent is to perform the patent search. You can do a google patent search or do a USPTO search. Patent-pending is the name used to describe the product that has been protected by the patent application. You can search for the public pair to locate the patent application. Once the patent office has approved your application, you will be able to conduct an online search for a patent number and locate the issued patent. The product you are selling will be patented. Alongside the USPTO search engine, you can also utilize other search engines like espacenet as described below. A patent lawyer or attorney can advise you on the procedure. In the US patents are issued by the US trademark and patent office, or the United States patent and trademark office, which also examines trademark applications.

Interested in finding more similar patents? Here are the steps:

1. Create a list of terms for your invention according to its function, composition, or use.

Begin by writing down a concise detailed description of your idea. Do not use generic terms such as “device,” “process,” and “system.” Instead, think about synonyms for the terms you selected initially. Also, make note of key technical terms as well as keywords.

To help you find keywords and concepts, use the questions below.

  • What’s the purpose of the invention? Is it a utilitarian device or an ornamental design?
  • Is the invention a method of creating something or performing some function? Or is it a product or procedure?
  • What is the basis of the invention? What is the physical composition of the invention?
  • What is the goal of this invention?
  • What are the technical terms and terms used to describe an invention’s nature? To assist you in finding the correct terms, consult the technical dictionary.

2. Use these terms to find relevant Cooperative Patent Classifications on the Classification Text Search Tool. If you’re not able to locate the appropriate classification to describe your invention, look through the classification’s Schemas of classes (class schedules) and try again. Consider substituting the words that you’ve used to describe your invention if you do not get any results from the Classification Text Search with synonyms such as the terms you used in step 1.

3. Check the CPC Classification Definition to confirm the validity of the CPC classification you found. The link to the CPC classification definition is available when the classification you have selected is a blue square with a “D” to its left. CPC classification definitions will help determine the relevant classification’s scope and therefore you’re certain to choose the one that is relevant. The definitions could also contain search tips or other suggestions that could be helpful for further investigation.

4. The Patents Full-Text Database and the Image Database allow you to retrieve patent documents that include the CPC classification. By focusing your search on abstracts and representative drawings it is possible to narrow your search to find the most relevant patent publications.

5. This collection of patent publications is the most appropriate to check for similarity to your invention. Pay close attention to the specification and claims. Consult the applicant and patent examiner for any additional patents.

6. Find patent applications published in the public domain using the CPC classification you picked in Step 3 from the Applications Full-Text and Image Database. You can use the same method of search in Step 4 to narrow your search results down to the most relevant patent application by examining the abstract and representative drawings on each page. Then, you must carefully review the published patent applications, paying particular attention to the claims as well as additional drawings.

7. You can find other US patent publications using keywords in the AppFT and PatFT databases, as well as the classification search of patents that are that are not issued by the United States according to below. Also, you can make use of search engines on the internet to find non-patent literature disclosures about inventions. For instance:

  • Add keywords to your search. Keyword searches may turn up documents that are not well-categorized or have missed classifications during Step 2. For example, US patent examiners often supplement their classification searches with keyword searches. Think about the use of technical engineering terminology rather than everyday words.
  • Search for foreign patents using the CPC classification. Then, re-run the search using international patent office search engines such as Espacenet, the European Patent Office’s worldwide patent publication database of over 130 million patent publications. Other national databases include:
  • Search non-patent literature. Inventions can be made public in many non-patent publications. It is recommended that you search journals, books, websites, technical catalogs, conference proceedings, and other print and electronic publications.

To review your search, you can hire a registered patent attorney to assist. A preliminary search will help one better prepare to talk about their invention and other related inventions with a professional patent attorney. In addition, the attorney will not spend too much time or money on patenting basics.