![]() | Only 14 pages are availabe for public view |
Abstract Cancer is the leading cause of deaths worldwide. Both researchers and doctors are facing the challenges of fighting cancer. According to the American cancer society, 96,480 deaths are expected due to skin cancer, 142,670 from lung cancer, 42,260 from breast cancer, 31,620 from prostate cancer, and17,760 deaths from brain cancer in 2019 (American Cancer Society, new cancer release report 2019).Early detection of cancer is the top priority for saving the lives of many. Typically, visual examination and manual techniques are used for these types of a cancer diagnosis. The E-Nose technology has been widely used in many fields .Representative advances are introduced in the following parts. E-Noses have been successfully used for diagnosis of health conditions via detection and classification of volatile organic compounds (VOCs) into one or combination of body fluids released as a result of a disease evolvement in the human body. Since ANNs are undoubtedly powerful tools for classification, clustering and pattern recognition. They can be applied to data with almost any number of inputs and outputs, and are well supported in different programming languages and software suites. Through manual modification of weights prior to training, and through imposing custom limitations on their modification during training, existing expert knowledge can be incorporated into their design and construction. Additionally, neural networks are usually computationally inexpensive to use after they have been trained, making them ideal for real-time applications where immediate output is desirable. Due to the lack of technology and the existing debates regarding the costeffectiveness of cancer detection .We have recently shown the potential use of an electronic nose (E-Nose) based on artificial neural network (ANN) for detecting volatile organic compounds (VOCs) in biofluids of cancer patients, VOCs, which are carbon-based chemicals classified on basis of their retention time and boiling point (WHO, 1989), emitted in the headspace over collected body fluids ex-vivo (i.e. blood, urine and tissue biopsy ) are believed to reflect the diagnosis of type of cancer ,either lung, brain or breast . In this work, an optimization architecture model of the network, design a deep learning classifier was donefor the identification of various types of cancer (lung cancer, breast cancer and brain cancer, based on odors of 386 patients diagnosed with cancer and 212 healthy controls samples with no underlying diseases matched for sex, and the socioeconomic level. These samples of different biological fluids (in case of LC Tissue biopsy was taken from suspicious lung mass for histopathological evaluation and blood, urine, breath and tissue biopsy if present) were collected and processed using E-Nose. Odor-print patterns were further analyzed using the principal component analysis (PCA) and artificial neural network (ANN) analysis. Moreover, ANN technique can differentially diagnosis three types of cancer (lung ,breast, and brain)in case of using blood ,urine and biopsy samples with accuracy 91, 98, and 95%; respectively. Conclusions 65 5. CONCLUSIONS • PCA cluster plots for chronic lymphocytic leukemia, breast, brain, and lung cancer patients and for healthy control, biopsy, blood and urine samples measurements, where the variance is 98.55, 91.18 and 97.62%, respectively. • ANN technique can differentially diagnose chronic lymphocytic leukemia, breast, brain, and lung cancer patients and for healthy controls using biopsy, blood and urine samples with accuracy 91, 98, and 96%; respectively. • The progress parameter values of ANN successful training process for biopsy, blood and urine samples. It shows the high performance (0.00893) of the ANN topology for classifying and predicting the type of cancer lung, breast and lung caner and healthy controls from urine samples, in less than a second. • On comparing ANN regression and PCA regression, we appreciate the good performance and high accuracy of the deep learning and the capability of ANN topology to deal with biological samples. • We noticed also as high accuracy as 99% for predicting the type of cancer breast, lung, brain, or healthy control from urine samples, as compared to biopsy and blood samples. |