# US 2019/0318822 A1 · Deep image classification of medical images > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2019/0318822 A1 - Document ID: US20190318822A1 - Status: Published application - Kind: A1 - Application number: 15/952,361 - Filed: 2018-04-13 - Published: 2019-10-17 - Earliest claimed: 2018-04-13 - Assignees: International Business Machines Corporation - Applicants: International Business Machines Corporation - Inventors: Xin ZHANG; Guo Tong XIE; Xiaolu ZHANG; Xiu Li LI; Peng GAO - Web page: https://patentgrep.com/patent/US20190318822A1 ## CPC - G16H 30/40: for processing medical images, e.g. editing - G06F 18/2148: {characterised by the process organisation or structure, e.g. boosting cascade} - G06F 18/217: {Validation; Performance evaluation; Active pattern learning techniques} - G06F 18/24: Classification techniques - G06N 3/045: Combinations of networks - G06N 3/0464: Convolutional networks [CNN, ConvNet] - G06N 3/082: modifying the architecture, e.g. adding, deleting or silencing nodes or connections - G06N 3/084: Backpropagation, e.g. using gradient descent - G06N 3/09: Supervised learning - G06V 10/454: {Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]} - G06V 10/82: using neural networks - G06V 20/698: {Matching; Classification} - G16H 50/20: for computer-aided diagnosis, e.g. based on medical expert systems - G16H 50/70: for mining of medical data, e.g. analysing previous cases of other patients - G06V 2201/03: Recognition of patterns in medical or anatomical images - G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING ## Abstract Methods, systems, and storage components for utilizing a deep neural network(s) for classifying at least one medical image. A deep neural network (DNN) can be configured by an image processing component to contain at least one revived neuron, where the revived neuron has an adjusted value based on a focus-learning function that is transferred to the DNN by the image processing component, where the focus-learning function provides the adjustment by updating the DNN with data that contains a corrected classification with respect to at least one normal image sample derived from a medical image, and where the correction is based on the focus-learning function comparing an annotation associated with an abnormal image sample derived from the medical image to another annotation associated with the at least one normal image. ## Claims ### Claim 1 A computer implemented method for improving classification of a medical image comprising: extracting, by a processor, a plurality of samples from the medical image, wherein the plurality of samples includes: i) at least one image sample with a normal portion of organic matter and ii) at least one image sample with an abnormal portion of organic matter; and training a focus-learning function for an image analysis component with the plurality of samples by: comparing a first classification made by the image analysis component as to at least one of: i) the at least one normal image sample, and ii) the at least one abnormal image sample, to at least one annotation associated with at least one of: i) the at least one normal image sample, and ii) the at least one abnormal image sample; determining that the classification was erroneous based on the annotation; and updating the focus-learning function to reflect a correct classification of at least one of the i) the normal image sample and ii) the abnormal image sample based on the determination that the classification was erroneous. ### Claim 2 (depends on 2) The computer implemented method of claim 2, wherein the abnormal portion of organic matter comprises a lesion on organic tissue, and wherein the normal portion of organic matter comprises organic tissue without a lesion. ### Claim 3 (depends on 1) The computer implemented method of claim 1, wherein the annotation comprises a first set of coordinates that identify a location on the at least one medical image associated with the abnormal portion of organic matter. ### Claim 4 (depends on 3) The computer implemented method of claim 3, wherein the first classification erroneously determines that the at least one normal image sample is an abnormal image sample, and wherein determining that the first classification was erroneous based on the annotation further comprises: comparing the first set of coordinates to a second set of coordinates, wherein the second set of coordinates are associated with the normal image sample; and determining, based on the coordinate comparison, that the first classification was erroneous. ### Claim 5 (depends on 4) The computer implemented method of claim 4, wherein updating the focus-learning function to reflect a correct classification of at least one of the i) the normal image sample and ii) the abnormal image sample based on the determination that the classification was erroneous comprises: updating the classification function to reflect the correct classification based on the coordinate comparison. ### Claim 6 (depends on 1) The computer implemented method of claim 1 further comprising: receiving a deep-neural network (DNN) based on the extracted plurality of samples; and updating the DNN based on the focus learning function. ### Claim 7 (depends on 6) The computer implemented method of claim 6, wherein the DNN includes one or more dead neurons related to the extracted plurality of samples, wherein at least one of the one or more dead neurons is associated with the incorrectly classified sample. ### Claim 8 (depends on 7) The computer implemented method according to claim 7 further comprising: reviving the at least one neuron associated with the incorrectly classified sample based on the updated focus-learning function. ### Claim 9 A non-transitory computer-readable storage medium containing computer program code which, when executed by operation of one or more computer processors, performs an operation comprising: extracting a plurality of samples from the at least one medical image, wherein the plurality of samples includes i) at least one image sample with a normal portion of organic matter and ii) at least one image sample with an abnormal portion of organic matter; training a focus-learning function for an image analysis component with the plurality of samples by comparing a first classification made by the image analysis component as to at least one of i) the at least one normal image sample and ii) the at least one abnormal image sample to at least one annotation associated with at least one of i) the at least one normal image sample and ii) the at least one abnormal image sample; determining that the classification was erroneous based on the annotation; and updating the focus-learning function to reflect a correct classification of at least one of the i) the normal image sample and ii) the abnormal image sample based on the determination that the classification was erroneous. ### Claim 10 (depends on 9) The computer-readable storage medium of claim 9, wherein the abnormal portion of organic matter refers to a lesion on organic tissue, and wherein the normal portion of organic matter refers to organic tissue without a lesion. ### Claim 11 (depends on 9) The computer-readable storage medium of claim 9, wherein the annotation refers to a first set of coordinates that identify a location on the at least one medical image associated with the abnormal portion of organic matter. ### Claim 12 (depends on 11) The computer-readable storage medium of claim 11, wherein the first classification erroneously determines that the at least one normal image sample is an abnormal image sample, and determining that the first classification was erroneous based on the annotation further comprises: comparing the first set of coordinates to a second set of coordinates, wherein the second set of coordinates are associated with the normal image sample; and based on the coordinate comparison, determining that the first classification was erroneous. ### Claim 13 (depends on 12) The computer-readable storage medium of claim 12, wherein the updating the focus-learning function to reflect a correct classification of at least one of the i) the normal image sample and ii) the abnormal image sample based on the determination that the classification was erroneous comprises: updating the classification function to reflect the correct classification based on the coordinate comparison. ### Claim 14 (depends on 9) The computer-readable storage medium of claim 9 configured to further perform: receiving an initial deep-neural network (DNN) based on the extracted plurality of samples; and updating the DNN based on the focus learning function. ### Claim 15 (depends on 14) The computer-readable storage medium of claim 14, wherein the DNN includes one or more dead neurons related to the extracted plurality of samples, wherein at least one of the one or more dead neurons is associated with the incorrectly classified sample. ### Claim 16 (depends on 15) The computer-readable storage medium of claim 15 configured to further perform: reviving the at least one neuron associated with the incorrectly classified sample based on the updated focus-learning function. ### Claim 17 A system comprising: one or more computer processors; and a memory containing computer program code that, when executed by operation of the one or more computer processors, performs an operation comprising: extracting a plurality of samples from the at least one medical image, wherein the plurality of samples includes i) at least one image sample with a normal portion of organic matter and ii) at least one image sample with an abnormal portion of organic matter; training a focus-learning function for an image analysis component with the plurality of samples by: comparing a first classification made by the image analysis component as to at least one of i) the at least one normal image sample and ii) the at least one abnormal image sample to at least one annotation associated with at least one of i) the at least one normal image sample and ii) the at least one abnormal image sample; determining that the classification was erroneous based on the annotation; and updating the focus-learning function to reflect a correct classification of at least one of the i) the normal image sample and ii) the abnormal image sample based on the determination that the classification was erroneous. ### Claim 18 (depends on 17) The system according to claim 17, wherein the abnormal portion of organic matter refers to a lesion on organic tissue, and wherein the normal portion of organic matter refers to organic tissue without a lesion. ### Claim 19 (depends on 17) The system according to claim 17, wherein the annotation refers to a first set of coordinates that identify a location on the at least one medical image associated with the abnormal portion of organic matter. ### Claim 20 (depends on 19) The system according to claim 19, wherein the first classification erroneously determines that the at least one normal image sample is an abnormal image sample, and determining that the first classification was erroneous based on the annotation further comprises: comparing the first set of coordinates to a second set of coordinates, wherein the second set of coordinates are associated with the normal image sample; and based on the coordinate comparison, determining that the first classification was erroneous. ### Claim 21 (depends on 20) The system according to claim 20, wherein the updating the focus-learning function to reflect a correct classification of at least one of the i) the normal image sample and ii) the abnormal image sample based on the determination that the classification was erroneous comprises: updating the classification function to reflect the correct classification based on the coordinate comparison. ### Claim 22 (depends on 17) The system of claim 17 configured to further perform: receiving an initial deep-neural network (DNN) based on the extracted plurality of samples; and updating the DNN based on the focus learning function. ### Claim 23 (depends on 22) The system medium of claim 22, wherein the DNN includes one or more dead neurons related to the extracted plurality of samples, wherein at least one of the one or more dead neurons is associated with the incorrectly classified sample. ### Claim 24 The system 23 configured to further perform: reviving the at least one neuron associated with the incorrectly classified sample based on the updated focus-learning function. ### Claim 25 A memory component for storing computer executable instructions in a computer device comprising: a deep neural network (DNN) configured by an image processing component to contain at least one revived neuron, wherein the revived neuron has an adjusted value based on a focus-learning function that is transferred to the DNN by the image processing component, wherein the focus-learning function provides the adjustment by updating the DNN with data that contains a corrected classification with respect to at least one normal image sample derived from a medical image, and wherein the correction is based on the focus-learning function comparing an annotation associated with an abnormal image sample derived from the medical image to another annotation associated with the at least one normal image.