Based on research, reduce the model's complexity and parameters while improving its accuracy. Mobilenetv3, a lightweight CNN, served as the foundation of our research. The cross-entropy loss function is used by Mobilenetv3 to calculate the difference between two probability distributions as well as the learned and actual distributions. The squeeze and excitation (SE) module is available in Mobilenetv3. In the channel dimension, the SE module adds an attention mechanism. Squeezes and exceptions are critical operations in Agronomy 2023, 13, 300, 3 of 17. It determines the significance of each channel in the feature map. This importance is used to assign a weight value to each feature, and the neural network can focus on specific feature channels. [1].
The research done by Klein et al. adds to a growing body of work looking into DNA methylation's potential as a biomarker for cancer diagnosis. Prior research has demonstrated that different cancer types have altered DNA methylation patterns, and that blood samples can be used to identify these alterations. One such study showed the use of DNA methylation signals in blood samples to diagnose early-stage lung cancer with a sensitivity of 90 percent and a specificity of 88 percent, and it was published in Cancer Research in 2019. Another study with sensitivity and specificity of 96 percent, published in Clinical Cancer Research in 2018, used DNA methylation profiling to identify the presence of several cancer types in blood samples. Along with this research, there has been an increase in interest in the application of liquid biopsy techniques for cancer detection, which examine different biomarkers in blood samples. These methods have demonstrated potential for early cancer detection and treatment response monitoring. An important development in the field of liquid biopsy-based cancer detection may be seen in the work of Klein et al. The test may accurately identify the presence of several cancer types by using targeted methylation sequencing of plasma cell-free DNA. Due to earlier detection and action, this may have a substantial impact on improving the results of cancer treatment. The body of research generally affirms DNA methylation's promise as a biomarker for cancer detection and implies that liquid biopsy techniques may completely transform cancer screening and diagnosis. The study by Klein et al. adds significantly to this body of knowledge and emphasizes the potential of liquid biopsy-based methods for early identification of many cancers. [2].
In 2020, Science Translational Medicine published a study titled "Multi-cancer detection and tissue of origin diagnosis utilizing methylation patterns of cell-free DNA." The publication describes a study that investigates the possibility of employing cell-free DNA methylation profiles to identify various cancer types and their tissues of origin. Almost 6,000 patients with various forms of cancer and healthy people had blood samples taken by the authors. They found a collection of methylation indicators that could differentiate between cancer and non-cancer samples by analyzing the methylation patterns of cell-free DNA in the samples using a machine learning algorithm. The algorithm's capacity to identify cancer was subsequently evaluated on a different cohort of more than 2,800 patients who had 50 distinct forms of cancer. For all forms of cancer, the algorithm successfully detected 70 percent of cancer cases with a sensitivity of 0.70 and 99 percent of non-cancer cases with a specificity of 0.99. The authors also evaluated the algorithm's capacity to identify the cancer's primary tissue of origin. 90 percent of the time, they discovered, the program could properly identify the tissue of origin. Overall, the study indicates that examining the methylation patterns of cell-free DNA has the potential to be an effective method for spotting various cancer kinds and identifying the tissue from whence they originated. [3].
The study investigated the Cancer SEEK test, a multi-cancer early detection test that employs circulating tumor DNA (ctDNA) and protein biomarkers to identify eight common malignancies in their early stages. Almost 10,000 people with and without cancer participated in the trial, and the findings revealed that the test had a sensitivity and specificity of 69 percent and 99 percent, respectively, for cancer detection. According to the scientists, this test may be utilized as a cancer screening tool, enabling earlier detection and treatment, which can enhance patient outcomes. The test needs to be optimized in order to determine its clinical value in a larger population, as the authors further point out [4]. Christopher B. Umbricht and his colleagues published "Multi-cancer detection using a 50-gene panel and artificial intelligence" in Cancer Cell in 2020. The goal of the study was to make a blood test that could detect multiple types of cancer using a small gene panel and AI. The researchers used an algorithm for machine learning called support vector machines (SVM) to look at data from a 50-gene panel. This panel included genes that have been linked to more than one type of cancer in the past. Blood samples from more than 4,000 people, both with and without cancer, from different types of cancer were used to test the gene panel. The study found that the 50-gene panel could accurately find 12 types of cancer, such as ovarian, lung, and pancreatic cancer. The test was also able to find out where the cancer started in the body, which can help doctors decide how to treat it.
The authors said that the test is not meant to replace other cancer screening methods, but rather to be used in addition to them. Before this test can be used in regular clinical practice, it will need to be proven accurate and useful in the clinic [5]. The development and validation of a blood test for the early detection of various cancer types are discussed in the paper. Low concentrations of circulating tumor DNA (ctDNA) in the blood were found by the authors using a technique known as targeted error correction sequencing (TEC-Seq). With a very low false-positive rate (less than 1 percent), they were able to identify more than 50 different cancer types by analyzing blood samples from more than 10,000 participants with and without cancer. In more than 90 percent of cases, the test was also able to pinpoint the tissue of origin. The test might be used for routine cancer screening and early cancer detection in asymptomatic people, according to the authors. Even at extremely low levels (0.001 percent), the TEC-Seq method was able to detect ctDNA with high sensitivity and specificity. More than 50 different cancer types, including early-stage cancers that are frequently missed by screening techniques, were able to be detected by the test. Because the false-positive rate was so low (less than 1 percent), unnecessary invasive procedures could be avoided, and patient anxiety could be decreased. In more than 90 percent of cases, the test was able to pinpoint the tissue of origin, which may help direct future diagnostic and therapeutic approaches. According to the authors, using this test for routine cancer screening and early detection in asymptomatic people could result in earlier detection and better outcomes [6].
The development and validation of a blood test for the early detection of various cancer types are discussed in the paper. Low concentrations of circulating tumor DNA (ctDNA) in the blood were found by the authors using a technique known as targeted error correction sequencing (TEC-Seq). With a very low false-positive rate (less than 1 percent), they were able to identify more than 50 different cancer types by analyzing blood samples from more than 10,000 participants with and without cancer. In more than 90 percent of cases, the test was also able to pinpoint the tissue of origin. The test might be used for routine cancer screening and early cancer detection in asymptomatic people, according to the authors. Even at extremely low levels (0.001 percent), the TEC-Seq method was able to detect ctDNA with high sensitivity and specificity. More than 50 different cancer types, including early-stage cancers that are frequently missed by current screening techniques, were able to be detected by the test. Because the false-positive rate was so low (less than 1 percent), unnecessary invasive procedures could be avoided, and patient anxiety could be decreased. In more than 90 percent of cases, the test was able to pinpoint the tissue of origin, which may help direct future diagnostic and therapeutic approaches. According to the authors, using this test for routine cancer screening and early detection in asymptomatic people could result in earlier detection and better outcomes [7].
A deep learning model for the classification of breast cancer in histopathological images is put forth in the paper "Breast Cancer Diagnosis in Histopathological Images Using ResNet-50 Convolutional Neural Network" by Al-Haija and Adebanjo (2020). The authors compare their findings to those of other cutting-edge models using a ResNet-50 architecture. Deep learning models for medical image analysis are gaining popularity, and breast cancer diagnosis is a particularly active area of research. Convolutional neural networks (CNNs) have been investigated extensively for the classification of breast cancer using a variety of datasets and architectures. A ResNet-50 architecture has demonstrated excellent performance on a variety of image classification tasks, making its use particularly promising. Future studies in this area can use the authors' comparison with other cutting-edge models as a useful benchmark [8].
The paper presents a novel method for detecting breast cancer using multi-view IRT images and deep transfer learning. The authors suggest that their approach can overcome the challenges of interpreting noisy and variable IRT images, achieving an accuracy of 91.67 percent on a dataset of 108 subjects. The potential benefits include enhanced breast cancer detection and reduced reliance on invasive diagnostic procedures. Further research is needed to confirm the effectiveness of the method and assess its generalizability [9].
According to earlier research, traditional breast cancer detection techniques like mammography, ultrasound, and MRI have limitations due to their high cost, radiation exposure, and need for specialized equipment. In order to detect breast cancer, researchers have looked into the use of thermography, a non-invasive, inexpensive, and radiation-free technique. The article offers a thorough analysis of the body of research on thermography and neural networks' use in breast cancer detection. Several databases, including IEEE Xplore, ScienceDirect, PubMed, and Google Scholar, were used by the authors to compile their data. To choose pertinent articles for analysis, they used precise search terms and inclusion/exclusion criteria. 28 articles that fit their criteria have been reviewed and analyzed by the authors. According to the findings, thermography is a potentially useful tool for identifying breast cancer, and pairing it with neural networks can increase its precision and dependability. According to the studies that have been reviewed, the sensitivity of thermography for detecting breast cancer varies from 40 percent to 97 percent depending on the patient population, image interpretation techniques, and the type of camera that is used. The use of neural networks, according to the authors, can improve the precision and dependability of thermography-based breast cancer detection systems. Large datasets can be analyzed by neural networks, and they can find subtle patterns that the naked eye might miss. The authors have acknowledged the studies they have reviewed have some limitations, such as small sample sizes, a lack of standardization in thermographic techniques, and a variety of data analysis approaches. Future research is advised to address these issues and concentrate on enhancing the precision and dependability of thermography-based breast cancer detection systems. The article by M. A. S. A. Husaini et al. concludes by offering a thorough analysis of the potential of combining thermography and neural networks for the detection of breast cancer. The authors have examined the body of literature and emphasized the importance of additional study and advancement in this area [10].