WO2022130308A1 - Techniques for cancer determination - Google Patents

Techniques for cancer determination Download PDF

Info

Publication number
WO2022130308A1
WO2022130308A1 PCT/IB2021/061907 IB2021061907W WO2022130308A1 WO 2022130308 A1 WO2022130308 A1 WO 2022130308A1 IB 2021061907 W IB2021061907 W IB 2021061907W WO 2022130308 A1 WO2022130308 A1 WO 2022130308A1
Authority
WO
WIPO (PCT)
Prior art keywords
response data
temperature response
breasts
features
steady state
Prior art date
Application number
PCT/IB2021/061907
Other languages
French (fr)
Inventor
Siddhartha Panda
Prashant Singh
Karun MALHOTRA
Ragash THACHAT
Original Assignee
Indian Institute Of Technology Kanpur
Murata Manufacturing Co., Ltd.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Indian Institute Of Technology Kanpur, Murata Manufacturing Co., Ltd. filed Critical Indian Institute Of Technology Kanpur
Priority to CN202180085873.5A priority Critical patent/CN116648186A/en
Publication of WO2022130308A1 publication Critical patent/WO2022130308A1/en

Links

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/01—Measuring temperature of body parts ; Diagnostic temperature sensing, e.g. for malignant or inflamed tissue
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/43—Detecting, measuring or recording for evaluating the reproductive systems
    • A61B5/4306—Detecting, measuring or recording for evaluating the reproductive systems for evaluating the female reproductive systems, e.g. gynaecological evaluations
    • A61B5/4312—Breast evaluation or disorder diagnosis
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H40/00—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices
    • G16H40/60—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices for the operation of medical equipment or devices
    • G16H40/63—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices for the operation of medical equipment or devices for local operation
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]

Definitions

  • Figure 1 illustrates a system for determining presence of cancerous cells in breasts, in accordance with an example implementation of the present subject matter
  • Figure 2 illustrates the system for determining presence of the cancerous cells in the breasts, in accordance with another example implementation of the present subject matter
  • Figure 3 illustrates a method for determining the presence of the cancerous cells in the breasts, in accordance with an example implementation of the present subject matter
  • Figure 4 illustrates a method for determining the presence of the cancerous cells in the breasts, in accordance with another example implementation of the present subject matter.
  • mammography In mammography, a breast is compressed and put between two plates. A low energy X-ray is then passed through the breast and an image is recorded on an X-ray film placed under the breast. The recorded image indicates the cancerous cells in higher contrast as compared to normal cells. However, for subjects having higher breast density, mammography is considered to be less sensitive in early stages.
  • MRI Magnetic ray imaging
  • breast thermography and breast thermometry includes devices that aims to detect the presence of cancerous cells based on an elevation in temperature of the breasts. It is a well-established notion in the field of medicine that local temperature in and around cancerous cells have elevated temperature in comparison to the normal cells.
  • the breast thermography includes recording of an infrared (IR) thermogram of the breast using an IR camera followed by analysis of the infrared thermogram for determining the presence of the cancerous cells.
  • IR infrared
  • the images obtained by the IR camera are captured from a distance, they have reduced sensitivity in determination of the temperature of the cells at early stages, especially when the temperature elevation is lower. As a result, the breast thermography is rendered ineffective in determination of the presence of the cancerous cells at early stages.
  • breast thermometry which involves taking the temperature through multiple thermal sensing elements which are in direct contact with the breasts.
  • the breast thermometry further involves creation of temperature profile which could be converted to a thermal map for each of the breasts based on one or more temperature readings received from the thermal sensing elements.
  • the thermal map may then be analysed by medical professionals for ascertaining the presence of the cancerous cells in the breasts.
  • the analysis of the thermal map varies based on the expertise and experience of each of the medical professional, thereby raising concerns relating to reliability of such diagnosis.
  • transient temperature response data for each of the breasts may be received.
  • the transient temperature response may be received from multiple thermal sensing elements which may be embedded in contact thermometry based wearable breast caps, where a breast cap may be worn by a subject on each of the breasts.
  • the number of thermal sensing elements embedded on each of the breast caps may vary based on the geometry of each of the breasts.
  • the transient temperature response data may then be stored and converted to steady state temperature response data.
  • the steady state temperature response data may then be processed to extract a set of features to be fed to various machine learning models for determination of presence of the cancerous cells in the breasts.
  • machine learning models may include, but are not limited to, artificial neural network (ANN), Support vector machine (SVM), Fuzzy Inference System (FIS), k-Nearest Neighbors (k-NN), Linear Discriminant Analysis (LDA), Naive Bayes Classifier, and Decision tree.
  • ANN artificial neural network
  • SVM Support vector machine
  • FIS Fuzzy Inference System
  • k-NN k-Nearest Neighbors
  • LDA Linear Discriminant Analysis
  • Naive Bayes Classifier and Decision tree.
  • Figure 1 illustrates a system 100 for determining presence of cancerous cells in breasts, in accordance with an example of the present subject matter.
  • system 100 may include, but are not limited to, a desktop, laptop, personal digital assistant, and smartphones.
  • the system 100 may be connected to a contact thermometry based wearable breast cap (not shown), where the breast cap has multiple thermal sensing elements embedded therein.
  • the breast cap may be worn by a subject in a manner, such that, the thermal sensing elements may contact the skin and may be able to sense transient temperature response data from a breast.
  • the system 100 may be connected to the breast cap via a data acquisition system (not shown), where the data acquisition system may collect and store the transient temperature response data from the thermal sensing elements.
  • the system 100 may include an accumulation engine 102 for receiving the transient temperature response data sensed by the thermal sensing elements.
  • the accumulation engine 102 may receive the transient temperature response data from a plurality of sources.
  • the accumulation engine 102 may receive the transient temperature response data directly from the breast cap.
  • the accumulation engine 102 may receive the transient temperature response data from the data acquisition system. The accumulation engine 102 may then convert the transient temperature response data into steady state temperature response data.
  • the system 100 may further include an analysis engine 104 that may analyze the steady state temperature response data to extract a set of features to be fed to various machine learning models for determination of presence of the cancerous cells in the breasts.
  • the set of features may include, but are not limited to, temperature features and fuzzy symmetry features.
  • the analysis engine 104 may analyze the set of features via the machine learning models trained based on data obtained from subjects having confirmed presence of cancerous cells in breasts. The data used for training may be obtained from subjects having various stages of breast cancer to enable efficient identification of the cancerous cells by the machine learning models.
  • machine learning models may include, but are not limited to, artificial neural network (ANN), Support vector machine (SVM), Fuzzy Inference System (FIS), k-Nearest Neighbours (k-NN), Linear Discriminant Analysis (LDA), Naive Bayes Classifier, and Decision tree.
  • ANN artificial neural network
  • SVM Support vector machine
  • FIS Fuzzy Inference System
  • k-NN k-Nearest Neighbours
  • LDA Linear Discriminant Analysis
  • Naive Bayes Classifier and Decision tree.
  • the system may also include a determination engine 106 that may collect results from the various machine learning models and combine the results through a decision fusion process for classification of the breasts as normal or cancerous.
  • a determination engine 106 may collect results from the various machine learning models and combine the results through a decision fusion process for classification of the breasts as normal or cancerous.
  • Figure 2 illustrates the system 100 for determining the presence of the cancerous cells in the breasts, in accordance with another example of the present subject matter.
  • the system 100 may include a processor 202 and a memory 204 coupled to the processor 202.
  • the functions of the various elements shown in the Figures, including any functional blocks labelled as “processor(s)”, may be provided through the use of dedicated hardware as well as hardware capable of executing instructions. When provided by a processor, the functions may be provided by a single dedicated processor, by a single shared processor, or by a plurality of individual processors, some of which may be shared.
  • explicit use of the term “processor” would not be construed to refer exclusively to hardware capable of executing instructions, and may implicitly include, without limitation, digital signal processor (DSP) hardware, network processor, application specific integrated circuit (ASIC), field programmable gate array (FPGA). Other hardware, standard and/or custom, may also be coupled to processor 202.
  • DSP digital signal processor
  • ASIC application specific integrated circuit
  • FPGA field programmable gate array
  • the memory 204 may include any computer-readable medium including, for example, volatile memory (e.g., Random Access Memory (RAM)), and/or non-volatile memory (e.g., ROM, EPROM, flash memory, etc.).
  • volatile memory e.g., Random Access Memory (RAM)
  • non-volatile memory e.g., ROM, EPROM, flash memory, etc.
  • the system 100 may also include engine(s) 206, which may include the accumulation engine 102, the analysis engine 104, and the determination engine 106.
  • the engine(s) 206 may be implemented as a hardware, firmware and a combination thereof.
  • the firmware of the engine may be processor executable instructions stored on a non-transitory machine-readable storage medium and the hardware for the engine may include a processing resource (for example, implemented as either a single processor or a combination of multiple processors), to execute such instructions.
  • the machine-readable storage medium may store instructions that, when executed by the processing resource, implement the functionalities of the engine.
  • the system 100 may include the machine -readable storage medium for storing the instructions and the processing resource to execute the instructions.
  • machine -readable storage medium may be located within the system 100. However, in other examples, the machine-readable storage medium may be located at a different location but accessible to system 100 and the processor 202.
  • the system 100 may further include data 208, that serves, amongst other things, as a repository for storing data that may be fetched, processed, received, or generated by the engine(s) 206.
  • the data 208 may include accumulation data 210, analysis data 212, determination data 214, and other data 216.
  • the data 208 may be stored in the memory 204.
  • a subject may be wearing a breast cap 218 on each of the left breast and the right breast.
  • the system 100 may be coupled to each of the breast caps 218 worn by the subject.
  • Each of the breast caps 218 may include ‘ 16’ to ‘512’ thermal sensing elements.
  • the number of sensing elements included in each of the breast caps are not fixed and may vary in accordance with the geometry of each of the breasts.
  • each of the thermal sensing elements may start generating the transient temperature response data.
  • the breast caps 218 may be coupled to the system 100 in a number of ways. In an example, the breast caps 218 may be directly coupled to the system 100. In another example, the breast caps 218 may be coupled to the system 100 via a data acquisition system 220. In said example, the data acquisition system 220 may acquire and store the transient temperature response data.
  • the accumulation engine 102 may receive the transient temperature response data for each of the thermal sensing elements. Based on the topology, the transient temperature response data for each of the thermal sensing elements may either be received directly from the breast caps 218 or from the data acquisition system 220. The transient temperature response data for each of the thermal sensing elements may be stored in the accumulation data 210. It would be noted that the usage of the term ‘transient temperature response data’ hereinafter, encompasses the transient temperature response data for each of the thermal sensing elements.
  • the accumulation engine 102 may filter out the transient temperature response data having fluctuations.
  • the accumulation engine 102 may filter out the transient temperature response data having fluctuations more than a threshold value.
  • the transient temperature response data having fluctuations more than the threshold value may be considered as bad quality data and generation of such data may be attributed to either loose connections or any other fault in the breast caps and therefore, may be discarded.
  • the threshold value for fluctuations may be ⁇ 0.75 °C. It would be noted that the above-mentioned threshold value for the fluctuations is exemplary and may be varied as per requirements.
  • the transient response data may then be processed for determining the presence of cancerous cells.
  • the accumulation engine 102 may denoise the transient temperature response data.
  • the transient temperature response data may be denoised by median filtering.
  • the filtered transient temperature response data may be stored in the accumulation data 210.
  • the accumulation engine 102 may then convert the transient temperature response data to steady state temperature response data.
  • the accumulation engine 102 may first select a saturation plateau region with least fluctuation from the transient temperature response data collected by each of the thermal sensing elements.
  • the accumulation engine 102 may then compute the mean temperature value of the saturation plateau region to extract the steady state temperature response data corresponding to the transient temperature response data collected by each of the thermal sensing elements.
  • the steady state temperature response data may be stored in the accumulation data 210.
  • T n represents steady state temperature value from a n th thermal sensing element.
  • the steady state temperature response data for the left breast and the right breast of a subject may be represented as follows:
  • T Right [T 1 ,T 2 ,T 3 ,T 4 ,T 5 ,T 6 ,T 7 ,T 8 ,T 9 ,T 10 ,T 11 ,T 12 ,T 13 ,T 14 ,T 15 ,T 16 j
  • T Left [T 17 ,T 18 ,T 19 ,T 20 ,T 21 ,T 22 ,T 23 ,T 24 ,T 25 ,T 26 ,T 27 ,T 28 ,T 29 ,T 30 ,T 31 ,T 32 ]
  • T Right is 1x16 size steady state temperature response data for the right breast.
  • T Left is 1x16 size steady state temperature response data for the left breast.
  • the analysis engine 104 may then analyze the steady state temperature response data to extract a set of features to be fed to various machine learning models for determination of presence of the cancerous cells in the breasts.
  • the set of features may include, but are not limited to, temperature features and fuzzy symmetry features.
  • the temperature features may be computed by computing a mean temperature value (M) of the steady state temperature data for the left breast and the right breast of the subject.
  • M mean temperature value
  • the mean temperature value be represented as follows:
  • T Right , T Left mean (T 1 ,T 2 ,T 3 , . . T 30 ,T 31 ,T 32 )
  • the analysis engine 104 may then normalize the steady state temperature response by the mean temperature value, to obtain the temperature features.
  • T mn represents the mean normalized temperature response value from the n th sensor.
  • the mean normalized temperature data for the left breast and the right breast may be represented as follows:
  • T m(Right) [T m1 , T m2 , T m3 , • • • T m14 , T m15 ,T m16 ]
  • T m(Left) [T m17 , T m18 , T m19 , • • • T m30 , T m31 ,T m32 ]
  • T m(Right) is mean normalized temperature data for the right breast
  • T m(Left) is mean normalized temperature data for the left breast.
  • the combined mean normalized temperature data from left and right breast may be considered as temperature features for the subject.
  • the temperature features may be represented as follows:
  • T T [T m1 , T m2 , T m3 , . . . T m30 , T m31 ,T m32 ]
  • T T is 1x32 size temperature features for the subject.
  • the analysis engine 104 may also compute the fuzzy symmetry features from the temperature responses.
  • absolute difference between a minimum value, a maximum value, and a mean value of the steady state temperature response data of the left breast and the right breast may be computed.
  • the fuzzy symmetry features may then be computed by assigning a fuzzy value between 0 and 1 based on the absolute difference between the minimum value, the maximum value, and the mean value of the steady state temperature response data of the left breast and the right breast.
  • T Right(min) may be a minimum value of the steady state temperature response data T Right for the right breast
  • T Left(min) may be a minimum value of the steady state temperature response data T Left for the left breast.
  • a fuzzy value T F(min) may be assigned between ‘0’ to ‘ 1’ in a graded manner.
  • the fuzzy value T F(min) may be assigned on the basis of a fuzzy scale.
  • ‘0’ on the fuzzy scale may correspond to a minimum value of absolute difference of the minimum temperature ⁇ T min
  • ‘ 1’ on the fuzzy scale may correspond to a threshold of the absolute difference of minimum temperature ⁇ T min .
  • the threshold of the absolute difference of minimum temperature ⁇ T min may be user-defined. A user may further be allowed to define other grades on the fuzzy scale. In this manner, the fuzzy value T F(min) corresponding to the ⁇ T min may be determined based on the fuzzy scale.
  • T Right(max) may be a maximum value of the steady state temperature response data T Right for the right breast
  • T Left(max) may be a maximum value of the steady state temperature response data T Left for the left breast.
  • a fuzzy value T F(max) may be assigned between ‘0’ to ‘ 1’ in a graded manner.
  • the fuzzy value T F(max) may be assigned on the basis of a fuzzy scale.
  • ‘0’ on the fuzzy scale may correspond to a minimum value of absolute difference of the maximum temperature ⁇ T max and ‘ 1’ on the fuzzy scale may correspond to a threshold of the absolute difference of maximum temperature ⁇ T max .
  • the threshold of the absolute difference of maximum temperature ⁇ T max may be user-defined. A user may further be allowed to define other grades on the fuzzy scale. In this manner, the fuzzy value T F(max) corresponding to the ⁇ T max may be determined based on the fuzzy scale.
  • T Right(mean) may be a mean value of the steady state temperature response data T Right for the right breast
  • T Left(mean) may be a mean value of the steady state temperature response data Tpeft for the left breast.
  • a fuzzy value T F(mean) may be assigned between ‘0’ to ‘ 1’ in a graded manner.
  • the fuzzy value T F(mean) may be assigned on the basis of a fuzzy scale.
  • ‘0’ on the fuzzy scale may correspond to a minimum value of absolute difference of the mean temperature ⁇ T mean and '1' on the fuzzy scale may correspond to a threshold of the absolute difference of mean temperature ⁇ T mean .
  • the threshold of the absolute difference of mean temperature ⁇ T mean may be user-defined. A user may further be allowed to define other grades on the fuzzy scale. In this manner, the fuzzy value T F(mean) corresponding to the ⁇ T mean may be determined based on the fuzzy scale.
  • the extracted features represent the temperature symmetry between both breasts and hence called fuzzy symmetry features.
  • T C is 1x35 combined feature set for the subject.
  • the combined feature set may be extracted for multiple subjects in the manner described above.
  • the analysis engine 104 may then perform principal component analysis on the combined feature set extracted for the multiple subjects.
  • the data may be divided in training data set and test data set.
  • the training data set may include combined feature set for all subjects who have already been classified into normal and cancerous, while the test data set may include the combined feature set for the subject under test.
  • the training data set may then be combined with the test data set for the subject under test in a sequential manner. Thereafter, the analysis engine 104 may perform principal component analysis for information extraction and data dimension reduction.
  • Performing the principal component analysis on the combined feature set of the subject under test reduces the dimensionality of the combined feature set significantly, thereby reducing the consumption of computational resources in further analysis of the combined feature set.
  • performing the principal component analysis reduce the size of the combined feature set to 1x3.
  • the analysis engine 104 may then analyze the combined feature set via multiple machine learning models that have been trained based on data obtained from subjects having confirmed presence of cancerous cells in the breasts.
  • the data used for training may be obtained from subjects having various stages of breast cancer to enable efficient identification of the cancerous cells by the machine learning models.
  • the machine learning models may include, but are not limited to, artificial neural network (ANN), Support vector machine (SVM), Fuzzy Inference System (FIS), k-Nearest Neighbours (k-NN), Linear Discriminant Analysis (LDA), Naive Bayes Classifier, and Decision tree.
  • the determination engine 106 may then perform decision fusion of the results obtained from the various machine learning models.
  • the decision fusion may be performed based on a majority vote rule.
  • the decision fusion may classify the breast to be normal or cancerous based on the results obtained from a majority of machine learning models. For instance, if the combined feature set is subjected to five machine learning models and three of the machine learning models confirms the presence of cancerous cells, the decision fusion may classify the breasts as cancerous.
  • Figure 3 illustrates a method 300 for determining the presence of cancerous cells in the breasts, in accordance with an example of the present subject matter.
  • the method 300 may be implemented in a variety of systems, but for the ease of explanation, the description of the method 300 is provided in reference to the above-described system 100.
  • the order in which the method 300 is described is not intended to be construed as a limitation, and any number of the described method blocks may be combined in any order to implement the method 300, or an alternative method.
  • blocks of the method 300 may be performed in the system 100.
  • the blocks of the methods 300 may be executed based on instructions stored in a non-transitory computer-readable medium, as will be readily understood.
  • the non-transitory computer-readable medium may include, for example, digital memories, magnetic storage media, such as magnetic disks and magnetic tapes, hard drives, or optically readable digital data storage media.
  • transient temperature response data for the breasts may be received.
  • the transient temperature response data may be received by an accumulation engine 102 of the system 100.
  • the transient temperature response data may be received directly from the breast caps worn by a subject.
  • the transient temperature response data may be received from a data acquisition system, where the data acquisition system may have received the transient temperature response from the breast caps worn by the subject.
  • the transient temperature response data may be converted to steady state temperature response data.
  • the transient temperature response data may be converted to steady state temperature response data by the accumulation engine 102.
  • the steady state temperature response data may be analysed to extract a set of features.
  • the set of features may be extracted to be fed to various machine learning algorithms to determine the presence of cancerous cells in the breasts.
  • the set of features may include, but are not limited to, temperature features and fuzzy symmetry features.
  • the analysis of the steady state temperature response data for extraction of the set of features may be done by the analysis engine 104.
  • the set of features may be analysed by various machine learning models, such as artificial neural network (ANN), Support vector machine (SVM), Fuzzy Inference System (FIS), k-Nearest Neighbours (k-NN), Linear Discriminant Analysis (LDA), Naive Bayes Classifier, and Decision tree, for determining the presence of cancerous cells in the breasts.
  • ANN artificial neural network
  • SVM Support vector machine
  • FIS Fuzzy Inference System
  • k-NN k-Nearest Neighbours
  • LDA Linear Discriminant Analysis
  • Naive Bayes Classifier a Bayes Classifier
  • results obtained from the various machine learning models may be combined through a decision fusion process for classification of the breast as normal or cancerous.
  • the results obtained from the various machine learning models may be combined by the determination engine 106.
  • Figure 4 illustrates a method 400 for determining the presence of the cancerous cells in the breasts, in accordance with another example of the present subject matter.
  • the method 400 may be implemented in a variety of systems, but for the ease of explanation, the description of the method 400 is provided in reference to the above-described system 100.
  • the order in which the method 400 is described is not intended to be construed as a limitation, and any number of the described method blocks may be combined in any order to implement the method 400, or an alternative method.
  • blocks of the method 400 may be performed in the system 100.
  • the blocks of the methods 400 may be executed based on instructions stored in a non-transitory computer-readable medium, as will be readily understood.
  • the non-transitory computer-readable medium may include, for example, digital memories, magnetic storage media, such as magnetic disks and magnetic tapes, hard drives, or optically readable digital data storage media.
  • transient temperature response data for the breasts may be received.
  • the transient temperature response data may be received by an accumulation engine 102 of the system 100.
  • the transient temperature response data may be denoised.
  • the transient temperature response data may be denoised by median filter denoising.
  • the transient temperature response data may be denoised by the accumulation engine 102.
  • the denoised transient temperature response data may be converted to steady state temperature response data.
  • the denoised transient temperature response data may be converted to steady state temperature response data by the accumulation engine 102.
  • the conversion may include selection of a saturation plateau region with least fluctuation from the transient temperature response data for each of the thermal sensing elements.
  • the conversion may further include computation of the mean temperature value of the saturation plateau region to extract the steady state temperature response data for each of the thermal sensing elements.
  • the steady state temperature response data may be processed to extract a set of features to be fed to various machine learning models for determining the presence of cancerous cells in the breasts.
  • the set of features may include temperature features and fuzzy symmetry features.
  • the extraction of temperature features includes computation of mean temperature value of the steady state temperature data for the left breast and the right breast of a subject, followed by normalization of the steady state temperature response data by the mean temperature value.
  • the extraction of the fuzzy symmetry features include, but is not limited to, computation of absolute difference between a minimum value, a maximum value, and a mean value of the steady state temperature response data of the left breast and the right breast, followed by assigning a fuzzy value between 0 and 1 based on the absolute difference between the minimum value, the maximum value, and the mean value of the steady state temperature response data of the left breast and the right breast.
  • the temperature features and fuzzy symmetry features may then be concatenated to get a combined feature set for the subject.
  • principal component analysis may be performed on the set of features, i.e., the combined feature set.
  • principal component analysis may be performed by the analysis engine 104.
  • Performing the principal component analysis on the combined feature set reduces the dimensionality of the combined feature set significantly, thereby reducing the consumption of computational resources in further analysis of the combined feature set.
  • the combined feature set may be analyzed via the various machine learning models to determine the presence of cancerous cells in the breasts.
  • the various machine learning models may be trained based on data obtained from subjects having confirmed presence of cancerous cells in the breasts. The data used for training may be obtained from subjects having various stages of breast cancer to enable efficient identification of the cancerous cells by the machine learning models.
  • the analysis of the combined feature set by the various machine learning models may be done by the analysis engine 104.
  • decision fusion of the results obtained from the various machine learning models may be performed to classify the breasts as cancerous or normal.
  • the decision fusion may be performed by a determination engine of the system 100.

Landscapes

  • Health & Medical Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Medical Informatics (AREA)
  • Public Health (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Pathology (AREA)
  • Data Mining & Analysis (AREA)
  • Epidemiology (AREA)
  • Primary Health Care (AREA)
  • Physics & Mathematics (AREA)
  • Surgery (AREA)
  • Artificial Intelligence (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Molecular Biology (AREA)
  • Databases & Information Systems (AREA)
  • Animal Behavior & Ethology (AREA)
  • Veterinary Medicine (AREA)
  • Biophysics (AREA)
  • General Business, Economics & Management (AREA)
  • Business, Economics & Management (AREA)
  • Evolutionary Computation (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physiology (AREA)
  • Psychiatry (AREA)
  • Signal Processing (AREA)
  • Gynecology & Obstetrics (AREA)
  • Reproductive Health (AREA)
  • Measuring Or Testing Involving Enzymes Or Micro-Organisms (AREA)

Abstract

Techniques for determining presence of cancerous cells in breasts of a subject are described. In an example, transient temperature response data for the breasts is received. The transient temperature response data is then converted to steady state temperature response data. A set of features for the breasts are then extracted based on the steady state temperature response data. The set of features are then analyzed by at least one machine learning model, where the at least one machine learning model is trained based on data obtained from a plurality of subjects with confirmed presence of cancerous cells in the breasts. The presence of cancerous cells in the breasts is determined based on the analysis of the set of features by the at least one machine learning model.

Description

TECHNIQUES FOR CANCER DETERMINATION
BACKGROUND
[0001] Breast cancer is one of the major diseases responsible for female mortality in the countries worldwide. According to a World Health Organisation (WHO) report, approximately 2.1 million women get impacted by the breast cancer every year. In year 2018, it is estimated that 627,000 women died because of breast cancer which is approximately 15% of all cancer deaths among women.
BRIEF DESCRIPTION OF DRAWINGS
[0002] The following detailed description references the drawings, wherein: [0003] Figure 1 illustrates a system for determining presence of cancerous cells in breasts, in accordance with an example implementation of the present subject matter,
[0004] Figure 2 illustrates the system for determining presence of the cancerous cells in the breasts, in accordance with another example implementation of the present subject matter,
[0005] Figure 3 illustrates a method for determining the presence of the cancerous cells in the breasts, in accordance with an example implementation of the present subject matter, and
[0006] Figure 4 illustrates a method for determining the presence of the cancerous cells in the breasts, in accordance with another example implementation of the present subject matter.
DETAILED DESCTRIPTION
[0007] Early breast cancer diagnosis can effectively reduce mortality rate. One of the conventionally used techniques for determining the presence of cancerous cells in breasts include mammography. In mammography, a breast is compressed and put between two plates. A low energy X-ray is then passed through the breast and an image is recorded on an X-ray film placed under the breast. The recorded image indicates the cancerous cells in higher contrast as compared to normal cells. However, for subjects having higher breast density, mammography is considered to be less sensitive in early stages.
[0008] An alternate technique for determining the presence of cancerous cells in breasts is Magnetic ray imaging (MRI). While MRI is effective for subjects having higher breast density, MRI requires injection of a contrast agent into a subject’s body for achieving high specificity, thereby rendering MRI unsuitable for subjects having allergic reactions to the contrast agent.
[0009] Other tools and techniques for determining the presence of the cancerous cells in breasts include electrical impedance scanning (EIS), mammary ductoscopy, breast thermography, and breast thermometry. Among the above- mentioned tools and techniques, the breast thermography and breast thermometry includes devices that aims to detect the presence of cancerous cells based on an elevation in temperature of the breasts. It is a well-established notion in the field of medicine that local temperature in and around cancerous cells have elevated temperature in comparison to the normal cells. The breast thermography includes recording of an infrared (IR) thermogram of the breast using an IR camera followed by analysis of the infrared thermogram for determining the presence of the cancerous cells. In breast thermography, since the images obtained by the IR camera are captured from a distance, they have reduced sensitivity in determination of the temperature of the cells at early stages, especially when the temperature elevation is lower. As a result, the breast thermography is rendered ineffective in determination of the presence of the cancerous cells at early stages.
[00010] The issues related to sensitivity in determination of the temperature are addressed by breast thermometry which involves taking the temperature through multiple thermal sensing elements which are in direct contact with the breasts. The breast thermometry further involves creation of temperature profile which could be converted to a thermal map for each of the breasts based on one or more temperature readings received from the thermal sensing elements. The thermal map may then be analysed by medical professionals for ascertaining the presence of the cancerous cells in the breasts. However, the analysis of the thermal map varies based on the expertise and experience of each of the medical professional, thereby raising concerns relating to reliability of such diagnosis.
[00011] According to example implementations of the present subject matter, techniques for determining the presence of cancerous cells in breasts are described. [00012] In an example, transient temperature response data for each of the breasts may be received. The transient temperature response may be received from multiple thermal sensing elements which may be embedded in contact thermometry based wearable breast caps, where a breast cap may be worn by a subject on each of the breasts. The number of thermal sensing elements embedded on each of the breast caps may vary based on the geometry of each of the breasts.
[00013] In said example, the transient temperature response data may then be stored and converted to steady state temperature response data. The steady state temperature response data may then be processed to extract a set of features to be fed to various machine learning models for determination of presence of the cancerous cells in the breasts. Examples of machine learning models may include, but are not limited to, artificial neural network (ANN), Support vector machine (SVM), Fuzzy Inference System (FIS), k-Nearest Neighbors (k-NN), Linear Discriminant Analysis (LDA), Naive Bayes Classifier, and Decision tree. The results obtained from the various machine learning models may then be combined through a decision fusion process for classification of the breast as normal or cancerous.
[00014] The determination of the presence of the cancerous cells based on the analysis of the temperature response data using various machine learning models reduces the role of factors associated with experiences and expertise of the medical professionals, thereby improving the reliability of breast cancer diagnosis.
[00015] The above techniques are further described with reference to Figure 1 to Figure 4. It would be noted that the description and the Figures merely illustrate the principles of the present subject matter along with examples described herein and would not be construed as a limitation to the present subject matter. It is thus understood that various arrangements may be devised that, although not explicitly described or shown herein, embody the principles of the present subject matter. Moreover, all statements herein reciting principles, aspects, and implementations of the present subject matter, as well as specific examples thereof, are intended to encompass equivalents thereof.
[00016] Figure 1 illustrates a system 100 for determining presence of cancerous cells in breasts, in accordance with an example of the present subject matter. Examples of system 100 may include, but are not limited to, a desktop, laptop, personal digital assistant, and smartphones.
[00017] The system 100 may be connected to a contact thermometry based wearable breast cap (not shown), where the breast cap has multiple thermal sensing elements embedded therein. The breast cap may be worn by a subject in a manner, such that, the thermal sensing elements may contact the skin and may be able to sense transient temperature response data from a breast. In an example, the system 100 may be connected to the breast cap via a data acquisition system (not shown), where the data acquisition system may collect and store the transient temperature response data from the thermal sensing elements.
[00018] The system 100 may include an accumulation engine 102 for receiving the transient temperature response data sensed by the thermal sensing elements. The accumulation engine 102 may receive the transient temperature response data from a plurality of sources. In an example, the accumulation engine 102 may receive the transient temperature response data directly from the breast cap. In another example, the accumulation engine 102 may receive the transient temperature response data from the data acquisition system. The accumulation engine 102 may then convert the transient temperature response data into steady state temperature response data.
[00019] The system 100 may further include an analysis engine 104 that may analyze the steady state temperature response data to extract a set of features to be fed to various machine learning models for determination of presence of the cancerous cells in the breasts. The set of features may include, but are not limited to, temperature features and fuzzy symmetry features. The analysis engine 104 may analyze the set of features via the machine learning models trained based on data obtained from subjects having confirmed presence of cancerous cells in breasts. The data used for training may be obtained from subjects having various stages of breast cancer to enable efficient identification of the cancerous cells by the machine learning models. Examples of machine learning models may include, but are not limited to, artificial neural network (ANN), Support vector machine (SVM), Fuzzy Inference System (FIS), k-Nearest Neighbours (k-NN), Linear Discriminant Analysis (LDA), Naive Bayes Classifier, and Decision tree.
[00020] Moreover, the system may also include a determination engine 106 that may collect results from the various machine learning models and combine the results through a decision fusion process for classification of the breasts as normal or cancerous. A detailed explanation of the system 100 is further provided with reference to an example implementation in Figure 2.
[00021] Figure 2 illustrates the system 100 for determining the presence of the cancerous cells in the breasts, in accordance with another example of the present subject matter.
[00022] The system 100 may include a processor 202 and a memory 204 coupled to the processor 202. The functions of the various elements shown in the Figures, including any functional blocks labelled as “processor(s)”, may be provided through the use of dedicated hardware as well as hardware capable of executing instructions. When provided by a processor, the functions may be provided by a single dedicated processor, by a single shared processor, or by a plurality of individual processors, some of which may be shared. Moreover, explicit use of the term “processor” would not be construed to refer exclusively to hardware capable of executing instructions, and may implicitly include, without limitation, digital signal processor (DSP) hardware, network processor, application specific integrated circuit (ASIC), field programmable gate array (FPGA). Other hardware, standard and/or custom, may also be coupled to processor 202.
[00023] The memory 204 may include any computer-readable medium including, for example, volatile memory (e.g., Random Access Memory (RAM)), and/or non-volatile memory (e.g., ROM, EPROM, flash memory, etc.). [00024] Further, the system 100 may also include engine(s) 206, which may include the accumulation engine 102, the analysis engine 104, and the determination engine 106.
[00025] In an example, the engine(s) 206 may be implemented as a hardware, firmware and a combination thereof. In examples described herein, such combinations of hardware and firmware may be implemented in several different ways. For example, the firmware of the engine may be processor executable instructions stored on a non-transitory machine-readable storage medium and the hardware for the engine may include a processing resource (for example, implemented as either a single processor or a combination of multiple processors), to execute such instructions.
[00026] In accordance with implementations of the present subject matter, the machine-readable storage medium may store instructions that, when executed by the processing resource, implement the functionalities of the engine. In such implementations, the system 100 may include the machine -readable storage medium for storing the instructions and the processing resource to execute the instructions.
[00027] In an example of the present subject matter, machine -readable storage medium may be located within the system 100. However, in other examples, the machine-readable storage medium may be located at a different location but accessible to system 100 and the processor 202.
[00028] The system 100 may further include data 208, that serves, amongst other things, as a repository for storing data that may be fetched, processed, received, or generated by the engine(s) 206. The data 208 may include accumulation data 210, analysis data 212, determination data 214, and other data 216. In an example, the data 208 may be stored in the memory 204.
[00029] In an example, a subject may be wearing a breast cap 218 on each of the left breast and the right breast. In said example, the system 100 may be coupled to each of the breast caps 218 worn by the subject. Each of the breast caps 218 may include ‘ 16’ to ‘512’ thermal sensing elements. As described earlier, the number of sensing elements included in each of the breast caps are not fixed and may vary in accordance with the geometry of each of the breasts. As soon as the subject wears the breast caps 218, each of the thermal sensing elements may start generating the transient temperature response data.
[00030] The breast caps 218 may be coupled to the system 100 in a number of ways. In an example, the breast caps 218 may be directly coupled to the system 100. In another example, the breast caps 218 may be coupled to the system 100 via a data acquisition system 220. In said example, the data acquisition system 220 may acquire and store the transient temperature response data.
[00031] In an example implementation of the present subject matter, the accumulation engine 102 may receive the transient temperature response data for each of the thermal sensing elements. Based on the topology, the transient temperature response data for each of the thermal sensing elements may either be received directly from the breast caps 218 or from the data acquisition system 220. The transient temperature response data for each of the thermal sensing elements may be stored in the accumulation data 210. It would be noted that the usage of the term ‘transient temperature response data’ hereinafter, encompasses the transient temperature response data for each of the thermal sensing elements.
[00032] In an example, the accumulation engine 102 may filter out the transient temperature response data having fluctuations. In said example, the accumulation engine 102 may filter out the transient temperature response data having fluctuations more than a threshold value. The transient temperature response data having fluctuations more than the threshold value may be considered as bad quality data and generation of such data may be attributed to either loose connections or any other fault in the breast caps and therefore, may be discarded. In an example, the threshold value for fluctuations may be ±0.75 °C. It would be noted that the above-mentioned threshold value for the fluctuations is exemplary and may be varied as per requirements. The transient response data may then be processed for determining the presence of cancerous cells.
[00033] In an example, the accumulation engine 102 may denoise the transient temperature response data. In said example, the transient temperature response data may be denoised by median filtering. The filtered transient temperature response data may be stored in the accumulation data 210.
[00034] The accumulation engine 102 may then convert the transient temperature response data to steady state temperature response data. In an example, to convert the transient temperature response data, the accumulation engine 102 may first select a saturation plateau region with least fluctuation from the transient temperature response data collected by each of the thermal sensing elements. The accumulation engine 102 may then compute the mean temperature value of the saturation plateau region to extract the steady state temperature response data corresponding to the transient temperature response data collected by each of the thermal sensing elements. The steady state temperature response data may be stored in the accumulation data 210.
[00035] In an illustrative example, let ‘Tn’ represents steady state temperature value from a nth thermal sensing element. In said example, the steady state temperature response data for the left breast and the right breast of a subject may be represented as follows:
T Right = [T1,T2,T3,T4,T5,T6,T7,T8,T9,T10,T11,T12,T13,T14,T15,T16j
TLeft=[T17,T18,T19,T20,T21,T22,T23,T24,T25,T26,T27,T28,T29,T30,T31,T32] where,
TRight is 1x16 size steady state temperature response data for the right breast, and
TLeft is 1x16 size steady state temperature response data for the left breast.
[00036] The analysis engine 104 may then analyze the steady state temperature response data to extract a set of features to be fed to various machine learning models for determination of presence of the cancerous cells in the breasts. The set of features may include, but are not limited to, temperature features and fuzzy symmetry features.
[00037] In an example, the temperature features may be computed by computing a mean temperature value (M) of the steady state temperature data for the left breast and the right breast of the subject. In an illustrative example, the mean temperature value be represented as follows:
M=mean ( TRight, TLeft) = mean (T1,T2,T3, . . T30,T31,T32 )
[00038] The analysis engine 104 may then normalize the steady state temperature response by the mean temperature value, to obtain the temperature features. In an illustrative example, let Tmn represents the mean normalized temperature response value from the nth sensor. In said example, the mean normalized temperature data for the left breast and the right breast may be represented as follows:
Tm(Right) = [Tm1, Tm2, Tm3, • • • Tm14, Tm15,Tm16]
Tm(Left) = [Tm17, Tm18, Tm19, • • • Tm30, Tm31,Tm32] where,
Tm(Right) is mean normalized temperature data for the right breast, and
Tm(Left) is mean normalized temperature data for the left breast.
[00039] The combined mean normalized temperature data from left and right breast may be considered as temperature features for the subject. In an illustrative example, the temperature features may be represented as follows:
TT = [Tm1, Tm2, Tm3, . . . Tm30, Tm31,Tm32] where,
TT is 1x32 size temperature features for the subject.
[00040] The analysis engine 104 may also compute the fuzzy symmetry features from the temperature responses. In an example, absolute difference between a minimum value, a maximum value, and a mean value of the steady state temperature response data of the left breast and the right breast, may be computed. The fuzzy symmetry features may then be computed by assigning a fuzzy value between 0 and 1 based on the absolute difference between the minimum value, the maximum value, and the mean value of the steady state temperature response data of the left breast and the right breast.
[00041] In an illustrative example, let TRight(min) may be a minimum value of the steady state temperature response data TRight for the right breast, and TLeft(min) may be a minimum value of the steady state temperature response data TLeft for the left breast. The absolute difference of minimum temperature ΔTmin of both breasts may be computed as: ΔTmin = absolute (TRight(min) - TLeft(min))
[00042] Based on the absolute difference of minimum temperature, a fuzzy value TF(min) may be assigned between ‘0’ to ‘ 1’ in a graded manner. The fuzzy value TF(min) may be assigned on the basis of a fuzzy scale. In an example, ‘0’ on the fuzzy scale may correspond to a minimum value of absolute difference of the minimum temperature ΔTmin and ‘ 1’ on the fuzzy scale may correspond to a threshold of the absolute difference of minimum temperature ΔTmin. In said example, the threshold of the absolute difference of minimum temperature ΔTmin may be user-defined. A user may further be allowed to define other grades on the fuzzy scale. In this manner, the fuzzy value TF(min) corresponding to the ΔTmin may be determined based on the fuzzy scale.
[00043] Similarly, let TRight(max) may be a maximum value of the steady state temperature response data TRight for the right breast, and TLeft(max) may be a maximum value of the steady state temperature response data TLeft for the left breast. The absolute difference of maximum temperature ΔTmax of both breasts may be computed as: ΔTmax = absolute ( TRight(max) - TLeft(max))
[00044] Based on the absolute difference of maximum temperature, a fuzzy value TF(max) may be assigned between ‘0’ to ‘ 1’ in a graded manner. The fuzzy value TF(max) may be assigned on the basis of a fuzzy scale. In an example, ‘0’ on the fuzzy scale may correspond to a minimum value of absolute difference of the maximum temperature ΔTmax and ‘ 1’ on the fuzzy scale may correspond to a threshold of the absolute difference of maximum temperature ΔTmax. In said example, the threshold of the absolute difference of maximum temperature ΔTmax may be user-defined. A user may further be allowed to define other grades on the fuzzy scale. In this manner, the fuzzy value TF(max) corresponding to the ΔTmax may be determined based on the fuzzy scale.
[00045] Further, let TRight(mean) may be a mean value of the steady state temperature response data TRight for the right breast, and TLeft(mean) may be a mean value of the steady state temperature response data Tpeft for the left breast. The absolute difference of mean temperature ΔTmean of both breasts may be computed as: ΔTmean = absolute (TRight(mean) - TLeft(mean))
[00046] Based on the absolute difference of mean temperature, a fuzzy value TF(mean) may be assigned between ‘0’ to ‘ 1’ in a graded manner. The fuzzy value TF(mean) may be assigned on the basis of a fuzzy scale. In an example, ‘0’ on the fuzzy scale may correspond to a minimum value of absolute difference of the mean temperature ΔTmean and '1' on the fuzzy scale may correspond to a threshold of the absolute difference of mean temperature ΔTmean. In said example, the threshold of the absolute difference of mean temperature ΔTmean may be user-defined. A user may further be allowed to define other grades on the fuzzy scale. In this manner, the fuzzy value TF(mean) corresponding to the ΔTmean may be determined based on the fuzzy scale.
[00047] As the aforementioned features are computed using the absolute difference between the maximum values, minimum values, and mean values of the steady state temperature response data from each of the left breast and the right breast, the extracted features represent the temperature symmetry between both breasts and hence called fuzzy symmetry features. Collectively, the fuzzy symmetry features may be represented as follows: TF = [ TF(min), TF(max),TF(mean) ]
[00048] The temperature features TT and fuzzy symmetry features TF may then be concatenated to get a combined feature set, which may be represented as follows: TC= [TT, TF] where,
TC is 1x35 combined feature set for the subject.
[00049] In an example, the combined feature set may be extracted for multiple subjects in the manner described above. The analysis engine 104 may then perform principal component analysis on the combined feature set extracted for the multiple subjects. In operation, the data may be divided in training data set and test data set. The training data set may include combined feature set for all subjects who have already been classified into normal and cancerous, while the test data set may include the combined feature set for the subject under test. The training data set may then be combined with the test data set for the subject under test in a sequential manner. Thereafter, the analysis engine 104 may perform principal component analysis for information extraction and data dimension reduction. Performing the principal component analysis on the combined feature set of the subject under test reduces the dimensionality of the combined feature set significantly, thereby reducing the consumption of computational resources in further analysis of the combined feature set. In an example, for a subject having combined feature set of size 1x35, performing the principal component analysis reduce the size of the combined feature set to 1x3.
[00050] The analysis engine 104 may then analyze the combined feature set via multiple machine learning models that have been trained based on data obtained from subjects having confirmed presence of cancerous cells in the breasts. The data used for training may be obtained from subjects having various stages of breast cancer to enable efficient identification of the cancerous cells by the machine learning models. Examples of the machine learning models may include, but are not limited to, artificial neural network (ANN), Support vector machine (SVM), Fuzzy Inference System (FIS), k-Nearest Neighbours (k-NN), Linear Discriminant Analysis (LDA), Naive Bayes Classifier, and Decision tree.
[00051] The determination engine 106 may then perform decision fusion of the results obtained from the various machine learning models. In an example, the decision fusion may be performed based on a majority vote rule. According to the majority vote rule, the decision fusion may classify the breast to be normal or cancerous based on the results obtained from a majority of machine learning models. For instance, if the combined feature set is subjected to five machine learning models and three of the machine learning models confirms the presence of cancerous cells, the decision fusion may classify the breasts as cancerous.
[00052] Figure 3 illustrates a method 300 for determining the presence of cancerous cells in the breasts, in accordance with an example of the present subject matter. Although the method 300 may be implemented in a variety of systems, but for the ease of explanation, the description of the method 300 is provided in reference to the above-described system 100. The order in which the method 300 is described is not intended to be construed as a limitation, and any number of the described method blocks may be combined in any order to implement the method 300, or an alternative method.
[00053] It may be understood that blocks of the method 300 may be performed in the system 100. The blocks of the methods 300 may be executed based on instructions stored in a non-transitory computer-readable medium, as will be readily understood. The non-transitory computer-readable medium may include, for example, digital memories, magnetic storage media, such as magnetic disks and magnetic tapes, hard drives, or optically readable digital data storage media.
[00054] At block 302, transient temperature response data for the breasts may be received. The transient temperature response data may be received by an accumulation engine 102 of the system 100. In an example, the transient temperature response data may be received directly from the breast caps worn by a subject. In another example, the transient temperature response data may be received from a data acquisition system, where the data acquisition system may have received the transient temperature response from the breast caps worn by the subject.
[00055] At block 304, the transient temperature response data may be converted to steady state temperature response data. The transient temperature response data may be converted to steady state temperature response data by the accumulation engine 102.
[00056] At block 306, the steady state temperature response data may be analysed to extract a set of features. The set of features may be extracted to be fed to various machine learning algorithms to determine the presence of cancerous cells in the breasts. The set of features may include, but are not limited to, temperature features and fuzzy symmetry features. In an example, the analysis of the steady state temperature response data for extraction of the set of features may be done by the analysis engine 104. [00057] At block 308, the set of features may be analysed by various machine learning models, such as artificial neural network (ANN), Support vector machine (SVM), Fuzzy Inference System (FIS), k-Nearest Neighbours (k-NN), Linear Discriminant Analysis (LDA), Naive Bayes Classifier, and Decision tree, for determining the presence of cancerous cells in the breasts. The set of features may be analysed by the various machine learning models by the analysis engine 104.
[00058] At block 310, results obtained from the various machine learning models may be combined through a decision fusion process for classification of the breast as normal or cancerous. In an example, the results obtained from the various machine learning models may be combined by the determination engine 106.
[00059] Figure 4 illustrates a method 400 for determining the presence of the cancerous cells in the breasts, in accordance with another example of the present subject matter. Although the method 400 may be implemented in a variety of systems, but for the ease of explanation, the description of the method 400 is provided in reference to the above-described system 100. The order in which the method 400 is described is not intended to be construed as a limitation, and any number of the described method blocks may be combined in any order to implement the method 400, or an alternative method.
[00060] It may be understood that blocks of the method 400 may be performed in the system 100. The blocks of the methods 400 may be executed based on instructions stored in a non-transitory computer-readable medium, as will be readily understood. The non-transitory computer-readable medium may include, for example, digital memories, magnetic storage media, such as magnetic disks and magnetic tapes, hard drives, or optically readable digital data storage media.
[00061] At block 402, transient temperature response data for the breasts may be received. In an example, the transient temperature response data may be received by an accumulation engine 102 of the system 100.
[00062] At block 404, the transient temperature response data may be denoised. In an example, the transient temperature response data may be denoised by median filter denoising. In an example, the transient temperature response data may be denoised by the accumulation engine 102. [00063] At block 406, the denoised transient temperature response data may be converted to steady state temperature response data. In an example, the denoised transient temperature response data may be converted to steady state temperature response data by the accumulation engine 102. The conversion may include selection of a saturation plateau region with least fluctuation from the transient temperature response data for each of the thermal sensing elements. The conversion may further include computation of the mean temperature value of the saturation plateau region to extract the steady state temperature response data for each of the thermal sensing elements.
[00064] At block 408, the steady state temperature response data may be processed to extract a set of features to be fed to various machine learning models for determining the presence of cancerous cells in the breasts. The set of features may include temperature features and fuzzy symmetry features. In an example, the extraction of temperature features includes computation of mean temperature value of the steady state temperature data for the left breast and the right breast of a subject, followed by normalization of the steady state temperature response data by the mean temperature value. In said example, the extraction of the fuzzy symmetry features include, but is not limited to, computation of absolute difference between a minimum value, a maximum value, and a mean value of the steady state temperature response data of the left breast and the right breast, followed by assigning a fuzzy value between 0 and 1 based on the absolute difference between the minimum value, the maximum value, and the mean value of the steady state temperature response data of the left breast and the right breast. The temperature features and fuzzy symmetry features may then be concatenated to get a combined feature set for the subject.
[00065] At block 410, principal component analysis may be performed on the set of features, i.e., the combined feature set. In an example, principal component analysis may be performed by the analysis engine 104. Performing the principal component analysis on the combined feature set reduces the dimensionality of the combined feature set significantly, thereby reducing the consumption of computational resources in further analysis of the combined feature set. [00066] At block 412, the combined feature set may be analyzed via the various machine learning models to determine the presence of cancerous cells in the breasts. In an example, the various machine learning models may be trained based on data obtained from subjects having confirmed presence of cancerous cells in the breasts. The data used for training may be obtained from subjects having various stages of breast cancer to enable efficient identification of the cancerous cells by the machine learning models. In an example, the analysis of the combined feature set by the various machine learning models may be done by the analysis engine 104.
[00067] At block 414, decision fusion of the results obtained from the various machine learning models may be performed to classify the breasts as cancerous or normal. In an example, the decision fusion may be performed by a determination engine of the system 100.
[00068] Although examples of the present subject matter have been described in language specific to methods and/or structural features, it is to be understood that the present subject matter is not limited to the specific methods or features described. Rather, the methods and specific features are disclosed and explained as examples of the present subject matter.

Claims

We Claim:
1. A method comprising: receiving transient temperature response data for breasts of a subject, wherein the transient temperature response data represents temperature data of the breasts collected over a period of time; converting the transient temperature response data to steady state temperature response data; extracting a set of features for the breasts based on the steady state temperature response data; and analyzing the set of features by at least one machine learning model, wherein the at least one machine learning model is trained based on data obtained from a plurality of subjects with confirmed presence of cancerous cells in the breasts; and determining presence of cancerous cells in the breasts of the subject based on the analysis of the set of features by the at least one machine learning model.
2. The method as claimed in claim 1, further comprising filtering out the transient temperature response data with fluctuations greater than a threshold value.
3. The method as claimed in claim 2, further comprising denoising the transient temperature response data, wherein the transient temperature response data is denoised by median filtering.
4. The method as claimed in claim 1, wherein converting the transient temperature response data to the steady sate temperature response data comprises: selecting a saturation plateau region with least fluctuations from the transient temperature response data; and computing a mean temperature value for the saturation plateau region.
5. The method as claimed in claim 1, wherein the set of features comprises temperature features, and wherein extracting the temperature features comprises: computing a mean temperature value of the steady state temperature response data for the breasts; and normalizing the steady state temperature response data for the breasts by the mean temperature value.
6. The method as claimed in claim 1, wherein the set of features comprises fuzzy symmetry features, and wherein extracting the fuzzy symmetry features comprises: computing a minimum value, a maximum value, and a mean value of the steady state temperature response data for each of the breasts; computing absolute differences between the minimum value of the steady state temperature response data for each of the breasts, the maximum values of the steady state temperature response data for each of the breasts, and the mean values of the steady state temperature response data for each of the breasts; and assigning a fuzzy value based on each of the absolute differences between the minimum value of the steady state temperature response data for each of the breasts, the maximum values of the steady state temperature response data for each of the breasts, and the mean values of the steady state temperature response data for each of the breasts.
7. The method as claimed in claim 1, further comprising performing principal component analysis on the set of features for reducing a dimension of the set of features.
8. The method as claimed in claim 1, further comprising: analysing the set of features by at least two machine learning models; and determining the presence of cancerous cells based on a combination of results of analysis of the set of features by the at least two machine learning models.
9. A system (100) comprising: an accumulation engine (102) to: receive transient temperature response data for breasts of a subject, wherein the transient temperature response data represents temperature data of the breasts collected over a period of time; and convert the transient temperature response data to steady state temperature response data; an analysis engine (104) coupled to the accumulation engine (102) to: extract a set of features for the breasts based on the steady state temperature response data; and analyze the set of features by at least one machine learning model, wherein the at least one machine learning model is trained based on data obtained from a plurality of subjects with confirmed presence of cancerous cells in the breasts; and a determination engine (106) coupled to the analysis engine (104) to determine presence of cancerous cells in the breasts of the subject based on the analysis of the set of features by the at least one machine learning model.
10. The system (100) as claimed in claim 9, wherein the accumulation engine (102) is to filter out the transient temperature response data with fluctuations greater than a threshold value.
11. The system (100) as claimed in claim 9, wherein the accumulation engine (102) is to denoise the transient temperature response data, wherein the transient temperature response data is denoised by median filtering.
12. The system (100) as claimed in claim 9, wherein to convert the transient temperature response data to the steady sate temperature response data, the accumulation engine is to: select a saturation plateau region with least fluctuations from the transient temperature response data; and compute a mean temperature value for the saturation plateau region.
13. The system as claimed in claim 9, wherein the set of features comprises temperature features, and wherein to extract the temperature features, the analysis engine (104) is to: compute a mean temperature value of the steady state temperature data for the breasts; and normalize the steady state temperature response data for the breasts by the mean temperature value.
14. The system (100) as claimed in claim 9, wherein the set of features comprises fuzzy symmetry features, and wherein to extract the fuzzy symmetry features, the analysis engine (104) is to: compute a minimum value, a maximum value, and a mean value of the steady state temperature response data for each of the breasts; compute absolute differences between the minimum value of the steady state temperature response data for each of the breasts, the maximum values of the steady state temperature response data for each of the breasts, and the mean values of the steady state temperature response data for each of the breasts; and assign a fuzzy value based on each of the absolute differences between the minimum value of the steady state temperature response data for each of the breasts, the maximum values of the steady state temperature response data for each of the breasts, and the mean values of the steady state temperature response data for each of the breasts.
15. The system (100) as claimed in claim 9, wherein the analysis engine (102) is to perform principal component analysis on the set of features to reduce a dimension of the set of features.
PCT/IB2021/061907 2020-12-18 2021-12-17 Techniques for cancer determination WO2022130308A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202180085873.5A CN116648186A (en) 2020-12-18 2021-12-17 Techniques for cancer determination

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
IN202011055304 2020-12-18
IN202011055304 2020-12-18

Publications (1)

Publication Number Publication Date
WO2022130308A1 true WO2022130308A1 (en) 2022-06-23

Family

ID=82058985

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/IB2021/061907 WO2022130308A1 (en) 2020-12-18 2021-12-17 Techniques for cancer determination

Country Status (2)

Country Link
CN (1) CN116648186A (en)
WO (1) WO2022130308A1 (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8231542B2 (en) * 2008-08-27 2012-07-31 Lifeline Biotechnologies, Inc. System for analyzing thermal data based on breast surface temperature to determine suspect conditions
WO2017066800A1 (en) * 2015-10-15 2017-04-20 University Of Louisville Research Foundation, Inc Methods of characterizing and/or predicting risk associated with a biological sample using thermal stability profiles
CN108078548A (en) * 2018-01-22 2018-05-29 哈尔滨海外电气系统工程总包有限公司 Physiological parameter detecting device and brassiere
CN108703746A (en) * 2018-08-13 2018-10-26 脱浩东 A kind of wearable device for detecting breast disease
US10111632B2 (en) * 2017-01-31 2018-10-30 Siemens Healthcare Gmbh System and method for breast cancer detection in X-ray images
US20200069191A1 (en) * 2018-08-07 2020-03-05 Higia, Inc. Temperature-based density-agnostic system and method for breast cancer diagnosis

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8231542B2 (en) * 2008-08-27 2012-07-31 Lifeline Biotechnologies, Inc. System for analyzing thermal data based on breast surface temperature to determine suspect conditions
WO2017066800A1 (en) * 2015-10-15 2017-04-20 University Of Louisville Research Foundation, Inc Methods of characterizing and/or predicting risk associated with a biological sample using thermal stability profiles
US10111632B2 (en) * 2017-01-31 2018-10-30 Siemens Healthcare Gmbh System and method for breast cancer detection in X-ray images
CN108078548A (en) * 2018-01-22 2018-05-29 哈尔滨海外电气系统工程总包有限公司 Physiological parameter detecting device and brassiere
US20200069191A1 (en) * 2018-08-07 2020-03-05 Higia, Inc. Temperature-based density-agnostic system and method for breast cancer diagnosis
CN108703746A (en) * 2018-08-13 2018-10-26 脱浩东 A kind of wearable device for detecting breast disease

Also Published As

Publication number Publication date
CN116648186A (en) 2023-08-25

Similar Documents

Publication Publication Date Title
Balasubramaniam Artificial intelligence algorithm with SVM classification using dermascopic images for melanoma diagnosis
EtehadTavakol et al. Breast cancer detection from thermal images using bispectral invariant features
Huang et al. A robust hair segmentation and removal approach for clinical images of skin lesions
Gogoi et al. Discriminative feature selection for breast abnormality detection and accurate classification of thermograms
Jafari et al. Automatic detection of melanoma using broad extraction of features from digital images
WO2019209680A1 (en) System and method for human temperature regression using multiple structures
Wang et al. Automatic digital ECG signal extraction and normal QRS recognition from real scene ECG images
Wang et al. A hybrid method for heartbeat classification via convolutional neural networks, multilayer perceptrons and focal loss
Vocaturo et al. Dangerousness of dysplastic nevi: A multiple instance learning solution for early diagnosis
Babu et al. Skin cancer detection using support vector machine with histogram of oriented gradients features
Raval et al. A Comprehensive assessment of Convolutional Neural Networks for skin and oral cancer detection using medical images
Chen et al. SE-MSCNN: a lightweight multi-scaled fusion network for sleep apnea detection using single-lead ECG signals
Zhou et al. Phase space reconstruction, geometric filtering based Fisher discriminant analysis and minimum distance to the Riemannian means algorithm for epileptic seizure classification
Dey et al. Red-plane asymmetry analysis of breast thermograms for cancer detection
Pious et al. A review on early diagnosis of skin cancer detection using deep learning techniques
Gamal et al. Early breast cancer screening from thermography via deep pre-trained edge detection with extreme gradient boosting
WO2022130308A1 (en) Techniques for cancer determination
CN114861738B (en) Electroencephalogram tracing and dipole selection-based motor imagery classification method
Navarro et al. Skin Disease Analysis using Digital Image processing
Pullagura et al. Recognition of Fetal heart diseases through machine learning techniques
Abdullah et al. Parkinson’s Disease Symptom Detection using Hybrid Feature Extraction and Classification Model
Takruri et al. Bayesian decision fusion for enhancing melanoma recognition accuracy
Dubey et al. Deconstructive human Confront Acknowledgment Utilizing Profound Neural Arrange
Lashgari et al. Manifold learning for ECG arrhythmia recognition
Chen et al. Camera-based peripheral edema measurement using machine learning

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21905963

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 202180085873.5

Country of ref document: CN

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21905963

Country of ref document: EP

Kind code of ref document: A1