Your browser doesn't support javascript.
Identification of the Types of Skin Cancers from Skin Cancer Images and Covid-19 Detection on Chest X-Ray Images using Deep Learning
i-Manager's Journal on Software Engineering ; 15(3):5-20, 2021.
Article in English | ProQuest Central | ID: covidwho-1630418
ABSTRACT
COVID-19 is a very deadly disease, which has killed thousands and infected millions of people worldwide. More recently in the year 2021, one of its mutants known as "The Delta Variant" has ravaged our country. It is also currently the chief cause of increasing cases in some North-Eastern states like Manipur and Arunachal Pradesh. Different measures have been adopted by the Government in collaboration with local social bodies to identify the infected individuals, detect the level of infection and also vaccinating individuals to shield them from this deadly disease. The current paper is also focused on one such stage, which is quite critical at this juncture, and will use the power of Artificial Intelligence to appropriately identify COVID-19 affected individuals using chest X-Ray images. When implemented, it will make it easier to identify the infection of the lungs by COVID-19. More specifically, the proposed methodology seeks to establish a chain of processes that can help in detecting the infection in the lungs using an advanced and novel image pre-processing with a prediction fusion-based deep learning-based identification system. The image pre-processing technique will initially improve the raw images by selectively optimizing the chromatic intensity and brightness of needy pixels using a deep learning-based Conditional Random Field (CRF) that uses the sigmoidal function. The enhanced image samples are made to undergo training with GoogLeNet and MobileNet deep learning models so that during the testing phase a prediction-fusion approach can be implemented to generate more robust prediction results. An exhaustive implementation with a standard dataset has revealed that the proposed approach can provide a mean accuracy of 98.63%, with the Covid and Normal classes showing 97.17% and 99.22% accuracies respectively. Another deadly disease that has infected thousands of people worldwide is skin cancer. Using the similar technical approach described above, a technique for identifying the type of skin cancer has been developed and experimented by using a standard dataset. Good accuracy of 85.42% has been achieved despite some classes having a comparatively lesser number of image samples. Finally, a Graphical User Interface (GUI) has also been developed by using the trained deep learning files of GoogLeNet and MobileNet so that a user can simply enter the desired image and check the type of prediction/class.
Keywords

Full text: Available Collection: Databases of international organizations Database: ProQuest Central Type of study: Prognostic study / Randomized controlled trials Topics: Vaccines / Variants Language: English Journal: I-Manager's Journal on Software Engineering Year: 2021 Document Type: Article

Similar

MEDLINE

...
LILACS

LIS


Full text: Available Collection: Databases of international organizations Database: ProQuest Central Type of study: Prognostic study / Randomized controlled trials Topics: Vaccines / Variants Language: English Journal: I-Manager's Journal on Software Engineering Year: 2021 Document Type: Article