Sử dụng quang phổ AI và Raman để đo glucose Using ai and raman spectroscopy to measure glucose Sử dụng quang phổ AI và Raman để đo glucose Using ai and raman spectroscopy to measure glucose
Overview
Diabetes mellitus, often referred to as diabetes, is a metabolic disorder characterized by the body's inability to properly manage glucose, resulting in elevated blood sugar levels Glucose is essential for the body's health, serving as the main energy source for cells, especially in the brain The causes of diabetes are complex and vary by type, but all forms of the condition lead to increased blood glucose levels, which can result in serious health complications.
Type 2 diabetes, once primarily affecting adults, is increasingly being diagnosed in young individuals, who often experience complications like vascular diseases shortly after onset Many of these cases go unnoticed or are overlooked due to mild or absent symptoms.
Prompt identification and timely intervention are crucial for individuals with diabetes
By effectively managing type 1 diabetes, one can prevent the development of type 2 diabetes, hence reducing the potential difficulties associated with the condition
Figure 1: Prevalence of diabetes between 2010-2030
Currently, over seven million people in Vietnam are diagnosed with diabetes, with more than 55% experiencing complications Among these, 34% suffer from cardiovascular issues, 39.5% face eye and nerve problems, and 24% deal with kidney complications These diabetes-related complications not only increase healthcare costs but also reduce the overall quality of life for patients.
According to the World Health Organization (WHO), approximately 422 million people worldwide suffer from this illness, predominantly in low- to middle-income countries The annual death toll exceeds 1.5 million, highlighting a concerning trend In recent decades, both the incidence and mortality rates have shown a steady and alarming increase.
Invasive blood testing remains the most common diagnostic method due to its accuracy; however, its high costs and lengthy result turnaround can lead to patient discomfort As a result, there is growing interest in non-invasive testing alternatives.
We are collaborating with Associate Professor Dr Nguyen Thanh Tung to develop a novel methodology that utilizes Raman spectroscopy and artificial intelligence (AI), enabling us to achieve rapid results that exceed the limitations of traditional methods.
Theory
Diabetes
Diabetes is a chronic medical condition characterized by the insufficient production of insulin by the pancreas or the body's inability to effectively utilize the insulin produced Insulin plays a crucial role in regulating blood glucose levels.
Diabetes leads to the body's inability to manage blood sugar levels, resulting in elevated blood sugar that can cause serious damage to various biological systems, especially the nerves and blood vessels over time.
Diabetes can be classified into three primary categories: type 1, type 2, and gestational diabetes The majority of patients receive a diagnosis of either type 1 or type 2
Figure 2: Type of Diabetes 2.1.3.1 Type 1 Diabetes
The onset phase of diabetes marks the body's inability to produce insulin, a crucial hormone for regulating blood glucose levels Approximately 5-10% of individuals are diagnosed with type 1 diabetes, which can now be identified at various ages due to advancements in the food industry The onset of this condition can be sudden and is accompanied by specific symptoms.
11 subtle, making them challenging to identify unless examined on a regular basis This is also the underlying factor responsible for the development of type 2 diabetes
Type 2 diabetes, accounting for 90-95% of diabetes cases, primarily affects adults but is increasingly seen in younger individuals due to high sugar intake from processed foods This condition occurs when the body cannot effectively use insulin to control blood sugar levels, often requiring medical treatments like medications and injections Type 2 diabetes develops gradually over several years and shares the absence of specific symptoms with type 1 diabetes, making regular screening essential However, making healthier dietary choices and incorporating physical activity can help delay its onset.
Gestational diabetes affects pregnant women without prior diabetes and usually resolves after childbirth It occurs when the placenta produces hormones that increase blood glucose levels, coupled with insufficient insulin production to regulate these levels Although it often subsides post-delivery, gestational diabetes can negatively impact the child's health and elevate the risk of developing type 2 diabetes later in life.
Artificial Intelligence (AI)
Artificial Intelligence (AI) is a branch of computer science dedicated to solving cognitive challenges similar to those faced by humans It has the ability to mimic human behavior in areas like reasoning, creativity, and perception The primary goal of AI is to create autonomous systems that can understand information and apply that knowledge to solve problems effectively.
12 akin to human beings Artificial intelligence (AI) constantly acquires new knowledge by leveraging accurate incoming data
ChatGPT is an artificial intelligence software that is continuously improving its ability to respond to extremely intricate questions This is the outcome of its ongoing process of acquiring knowledge
Artificial Intelligence (AI) can be utilized across a wide range of domains, including both everyday life and industrial settings, to streamline corporate operations, improve consumer satisfaction, and foster groundbreaking advancements
2.2.2 AI in the Medical Field
Artificial Intelligence (AI) is set to revolutionize the medical field, demonstrating significant progress in disease treatment With advanced algorithms and fast data analysis, AI facilitates early disease detection, leading to quicker treatments and better patient outcomes The considerable benefits of AI have prompted extensive research into its applications in diagnosis and therapy.
A recent study by experts from the University of Canterbury in New Zealand reveals that artificial intelligence (AI) can help healthcare practitioners develop more effective cancer treatment strategies, significantly improving patient survival rates Associate Professor Alex Gavryushkin, from the Mathematical Biology Research Centre, led a four-year investigation that supports this conclusion.
Experts developed algorithms to analyze complex biological data related to genetic disorders like cancer and gout, aiming to create therapy protocols based on genetic insights The research team trained AI to evaluate genetic and clinical data, linking individual patient conditions with established medical knowledge Following this, clinical trials utilized AI to provide crucial recommendations, including the combination of various medications during treatment.
Associate Professor Gavryushkin emphasized that traditional medical practices often rely on data from large groups of patients to establish treatment methods However, in diseases like cancer, individual patient conditions can differ greatly despite similar symptoms Therefore, a one-size-fits-all approach to medication and therapy is less effective than personalized treatments that consider each patient's unique genetic makeup.
This technology is set to empower healthcare professionals to create more effective cancer treatment protocols, manage the proliferation of drug-resistant cells, and improve overall healthcare convenience Associate Professor Gavryushkin has affirmed that this AI-driven approach can serve as a reliable medical assistant in cancer therapy.
He anticipates the widespread implementation of this approach, particularly in regions where clinicians may lack extensive training in genetics or the ability to thoroughly review all existing genetic literature
Machine learning, a key area of artificial intelligence and computer science, involves the study and creation of algorithms and techniques that enable computers to learn from data autonomously, without requiring explicit programming.
Machine learning involves programming computers to independently detect models and hidden patterns in data using specialized algorithms These algorithms leverage techniques from statistics, mathematics, and information theory to learn from data and enhance specific goals, including prediction, classification, and pattern recognition.
Machine learning enables computers to learn similarly to humans by processing data and gaining experience Through analyzing incoming information, computers identify key features and automatically adjust their model parameters to improve prediction and classification accuracy.
By repeatedly applying this procedure with new data, computers can continuously improve their performance, leading to more accurate predictions and decisions This advancement allows machine learning to be utilized in various fields, including financial market forecasting, medical diagnosis, and automation of production processes.
Machine learning is increasingly integrated into various applications and management systems, allowing businesses to quickly analyze data and present viable options and products to consumers This technology has the potential to significantly improve conversion rates, driving business success.
A machine learning-driven e-commerce platform can analyze customer behavior by leveraging algorithms to predict actions on the site By utilizing data from product views, user interactions, and search terms, the platform can significantly improve user experience, moving beyond a simple product display based on predetermined criteria or purchase history.
The system utilizes user data to analyze and understand individual behavioral patterns, such as a consistent interest in premium mobile phones, indicating a preference for advanced technology When users visit the website, the algorithm can provide personalized product recommendations based on their interests, purchase history, and reviews, ensuring the suggestions are highly relevant and engaging.
Implementing a personalized shopping approach can significantly improve the user experience on the website, making it easier for customers to discover products that match their unique interests and needs Additionally, leveraging machine learning allows for the analysis of customer shopping behaviors, enabling the company to adapt its strategies effectively.
In this project, I employ Convolutional Neural Networks (CNN), which are widely utilized deep neural networks, to assess diabetes
A Convolutional Neural Network (CNN) is a specialized type of artificial neural network designed for image recognition and processing, known for its ability to detect and analyze patterns in images Training a CNN necessitates a large dataset with millions of labeled data points, and to ensure efficient training, high-performance processors like Graphics Processing Units (GPUs) or Neural Processing Units (NPUs) are typically utilized.
Raman Spectroscopy
Raman Spectroscopy is a non-invasive analytical technique that offers insights into the chemical composition, crystalline structure, and molecular interactions of materials By examining the interaction between light and chemical bonds, this technology delivers detailed information about a substance's properties.
Raman Spectroscopy is a technique that analyzes the scattering of light from a high-intensity laser as it interacts with molecules While the majority of the scattered light, known as Rayleigh Scatter, maintains its original wavelength and offers little information, a tiny fraction—approximately 0.0000001%—scatters at different wavelengths, referred to as Raman Scatter This variation in wavelengths provides essential insights into the chemical composition of the molecules being studied.
A Raman spectrum features distinct peaks that indicate the intensity and wavelength of Raman scattered light, reflecting specific molecular bond vibrations These peaks can correspond to individual bonds, such as C-C or C=C, as well as collective bond groupings like benzene ring breathing modes or polymer chain vibrations The intricate pattern of these peaks provides valuable insights into the molecular structure and composition of the substance being analyzed.
Raman spectroscopy is an effective method for investigating the chemical composition of materials, providing significant information on several aspects
● Chemical structure and identity: Raman spectroscopy provides detailed information about the molecular composition and arrangement of atoms within a material, enabling the identification of chemical structures and compounds
Raman spectroscopy is essential for material characterization and quality control as it analyzes the characteristic Raman spectra of various phases and polymorphs, enabling the identification of phase transitions and the presence of different polymorphic forms in a material.
Intrinsic stress and strain can be identified through changes in the Raman spectra of materials, enabling the assessment of their mechanical properties and structural integrity.
Raman spectroscopy is an essential tool for identifying contaminants and impurities in materials, even at minimal concentrations Its ability to detect these unwanted substances makes it invaluable for quality assurance and purity assessment across various industries.
A Raman spectrum acts as a distinct chemical fingerprint for specific molecules or materials, enabling rapid comparison with extensive spectral libraries that house thousands of reference spectra This capability significantly enhances the efficiency and accuracy of material identification, making Raman spectroscopy a valuable analytical tool across diverse fields such as pharmaceuticals, forensics, and materials science.
Raman spectroscopy originated in 1928 when Dr Raman in India discovered the Raman effect, earning him a Nobel Prize Despite this early breakthrough, the practical use of Raman spectroscopy was limited due to the weak scattered light compared to the excitation light The introduction of lasers in the 1960s provided an effective light source, significantly enhancing the capabilities of Raman spectroscopy and allowing it to reach its full potential.
In the late 1970s, the advent of microscopic Raman spectroscopy, which combined an optical microscope with a Raman spectrometer, transformed local analysis across multiple disciplines However, during the 1980s, the swift progress of the Fourier transform infrared absorption method eclipsed Raman spectroscopy, primarily due to the latter's reputation for measurement complexity.
The evolution of charge-coupled devices (CCDs) in digital and video cameras has transformed the landscape of spectral analysis by enhancing detector performance and enabling quicker results This technological advancement has led to the development of more compact spectrometers with improved sensitivity, simplified maintenance, and superior performance Additionally, advanced filters designed to eliminate Rayleigh scattered light have further increased measurement accuracy.
Recent advancements have revitalized Raman spectroscopy as a preferred analytical method among researchers in diverse fields Similar to absorption spectroscopy, it has emerged as a key focus in research, providing critical insights into molecular structure, composition, and interactions.
The Raman effect is fundamental to Raman spectroscopy and involves the interaction between a sample's electron cloud and the electric field of monochromatic light, resulting in an induced dipole moment based on the molecule's polarizability Unlike fluorescence or phosphorescence, which involve transitions between electronic states, Raman scattering is characterized by inelastic light scattering, where incident photons excite the sample without elevating it to higher energy levels.
Raman scattering occurs when a photon excites a molecule to a virtual energy state, resulting in a scattered photon whose energy may differ from that of the incident photon This energy difference, or frequency shift, reflects the transition between the initial and final rovibronic states of the molecule If the final state has a higher energy level, the scattered photon will exhibit a specific frequency shift.
27 photon experiences a downshift (Stokes shift), while a lower energy final state results in an upshift (anti-Stokes shift)
Raman spectroscopy is based on the variations in electric dipole-electric dipole polarizability related to vibrational coordinates, which influences the intensity of Raman scattering This intensity correlates directly with the changes in polarizability, making the Raman spectrum a representation of the rovibronic states of the molecule.
Raman spectroscopy differs from infrared absorption by relying on polarizability derivatives instead of dipole moment derivatives, enabling it to study rovibronic transitions that are often inactive in IR spectroscopy This capability is particularly significant for centrosymmetric molecules, where mutual exclusion rules limit IR activity.
Furthermore, Raman spectroscopy complements other vibrational spectroscopy techniques like IR absorption and inelastic incoherent neutron scattering (IINS) While
Result
Data input
In my study, I gathered authentication data from human bodies using Raman spectroscopy, with explicit consent from both my teacher and a friend The data collection involved illuminating specific regions of the human body, focusing primarily on the ear, which has consistently produced accurate results in previous experiments After the illumination process, I generated a sequence of Raman data and employed Machine Learning techniques, specifically a Convolutional Neural Network (CNN) model, to detect diabetes by analyzing labels such as pre-breakfast and post-meal The accuracy of each run was evaluated using the val_accuracy result obtained through this methodology.
For this research, I underwent training using two datasets: one with two labels and another with three labels
In the dataset with two labels, there are a total of 20 samples that are divided into two unique categories, labeled as 0 and 1
Table 1 Meaning of binary Classification
The three-label dataset comprises 30 samples, each associated with one of three labels:
Table 2 Meaning of three-labels Classification
Both datasets feature a uniform length of 2048 units for each measurement sample's signal They are organized into two CSV files: one containing the signal data and the other containing the corresponding signal labels.
The complexity of the dataset with three labels is significantly higher than that of the two-label dataset, leading to differences in their respective data processing models.
Training model
Selecting a convolutional neural network (CNN) model for this project offers numerous benefits, particularly its ability to automatically learn hierarchical features from data This reduces the need for manual feature engineering, streamlining the model development process and increasing its adaptability to different datasets.
CNNs are highly effective at recognizing spatial dependencies in input data, which makes them ideal for tasks that involve spatial recognition and sequential data analysis This capability is crucial in areas such as medical diagnostics and natural language processing, where grasping the complex relationships between data points is essential for making accurate predictions.
CNNs exhibit impressive scalability and efficiency, enabling them to effortlessly manage large datasets This scalability is crucial for real-world applications that face challenges with complex datasets and computational limitations.
In essence, opting for a CNN model signifies a strategic embrace of cutting-edge machine learning techniques, ensuring optimal performance and resilience in tackling the project's challenges
The libraries listed above are the ones I utilize in the project
NumPy, imported as `import numpy as np`, is an essential library for scientific computing in Python It offers robust support for arrays and matrices, along with a variety of mathematical functions designed to efficiently handle these data structures In many projects, NumPy is particularly valuable for performing numerical operations on large datasets.
Pandas, imported as `import pandas as pd`, is a robust library designed for data manipulation and analysis It provides essential data structures like DataFrame and Series, making it ideal for managing structured data such as CSV files In your project, you are utilizing Pandas for data processing tasks, including reading CSV files and executing a variety of data operations.
Matplotlib, imported as `import matplotlib.pyplot as plt`, is a versatile library in Python designed for creating static, animated, and interactive visualizations With a MATLAB-like interface, it offers a diverse range of plotting options and extensive customization capabilities In your project, it serves as a crucial tool for visualizing data and analyzing model performance effectively.
TensorFlow, an open-source machine learning framework developed by Google, is extensively used for constructing and training machine learning and deep learning models In your project, you are leveraging TensorFlow's high-level API, Keras, which streamlines the development and training of neural networks.
Keras, part of TensorFlow, is an API that facilitates user-friendly and rapid experimentation with deep neural networks It offers a high-level interface for constructing and training neural networks, streamlining the prototyping and iterative process of model development In your project, Keras plays a crucial role in building and training your models effectively.
Early stopping is a valuable technique in machine learning that helps prevent overfitting by halting the training process when a monitored metric, like validation loss, ceases to improve By implementing early stopping in your project, you can avoid excessive training epochs, thereby conserving both time and computational resources.
ModelCheckpoint, a callback from tensorflow.keras.callbacks, is essential for saving a model's weights at specific intervals during training This feature enables you to preserve the best-performing model based on a chosen metric, ensuring that you can easily access and evaluate the optimal version of your model for future use.
The Adam optimizer, imported from tensorflow.keras.optimizers, is a widely-used algorithm for training deep learning models It effectively merges the benefits of two advanced stochastic gradient descent variants: AdaGrad and RMSProp In your project, the Adam optimizer plays a crucial role in updating the network weights throughout the training process.
L2 regularization, commonly referred to as weight decay, is a crucial technique in machine learning that helps mitigate overfitting by incorporating a penalty term into the loss function to discourage excessively large weights In your project, you are utilizing L2 regularization by importing it from the tensorflow.keras.regularizers module.
38 regularization, which can be applied to the network's layers to improve generalization performance
The random module in Python is essential for generating random numbers and sampling from sequences, making it useful for tasks like shuffling datasets and creating random seeds to ensure reproducibility in your projects.
The shuffle_data function is used to randomize the order of data and their corresponding labels within arrays, which is essential for machine learning and deep learning tasks This randomization ensures that data and labels are not arranged in a fixed sequence, allowing the model to learn more effectively and preventing it from making associations based on the order of the data.
This function requires two specific arguments: data, which is an array that contains the data, and label, which is an array that has the associated labels Subsequently, the
39 program outputs the configuration of the initial data and labels, together with the unaltered sequence of samples and labels
The process involves generating a new set of randomized indices by shuffling the original sample indices These newly created indices are then used to produce a rearranged duplicate of the data and labels.
Result
The result obtained by this model is quite high, typically around from 75% to 100% This is a promising outcome in diabetes diagnosis research without invasive measures
Furthermore, after executing 10 iterations, all of them have a consistent outcome of 95%
Table 3 2-labels training data results
Figure 22: Plot Accuracy and Loss for 2-data labels
Thus, the average accuracy across runs is 95% with 2 data labels
Achieving 95% accuracy with a dataset containing two labels highlights the impressive capability of the AI system in accurately classifying patients with and without diabetes This result underscores the potential of AI in analyzing complex medical data, suggesting that such technology could effectively assist medical professionals in diagnosing and treating diabetes in real-world applications However, it's important to recognize that this high accuracy is attained only in a binary classification environment, indicating a simpler data structure.
The results from the 3-label dataset are promising, achieving an accuracy of 83.33%, indicating that the model has effectively learned key features for accurate classification To fully evaluate the model's performance, it is essential to consider factors such as data balance, generalization ability, and potential enhancements through model architecture adjustments, parameter tuning, or data augmentation techniques Overall, this outcome marks a significant advancement in developing an effective classification model for 3-label data.
In general, each run of the machine learning model leads to an accuracy improvement ranging from 66.7% to 100% However, when utilizing 3-label data, the results exhibit greater instability compared to 2-label data, largely due to fluctuations observed between different runs.
Below are the recorded results from the last 10 runs
Table 4 3-labels training data results
Figure 24: Plot Accuracy and Loss for 3-data labels
Thus, the average accuracy across runs is 84.4% Currently, this is the best result with the 3-label data type, as its greater complexity impacts the accuracy of predictions
The AI system achieved a significant accuracy of 84.4% using three-label data, despite being lower than the two-label data results This complexity arises from classifying patients into categories such as without diabetes, at risk for diabetes, and with diabetes The reduced accuracy highlights the challenges of managing intricate classification scenarios, emphasizing the necessity for ongoing research and development to enhance the model and input data for improved accuracy.
During my graduate project, I systematically documented my work, focusing on comprehensive research and a robust theoretical foundation The primary objective was to create an artificial intelligence model capable of accurately assessing diabetes using Raman input data To improve the system's accuracy, I carefully fine-tuned the model by adjusting parameters, including the number of layers, layer size, and the volume of input samples in each training iteration.
I developed a highly accurate AI model that achieved an impressive 84.33% accuracy with three-label data and a perfect 95% accuracy with two-label data Despite these successes, challenges like overfitting remain unresolved.
This project deepened my understanding of health, artificial intelligence, and big data, leading to the creation of disease diagnostic models As technology rapidly evolves, sectors like healthcare are increasingly adopting these advancements across various activities However, inherent limitations persist, necessitating careful attention to risks and ethical challenges associated with machine use These challenges highlight the ongoing need for human experts in specialized fields.
Recent literature highlights the significant role of AI in non-invasive diabetes identification, emphasizing its ability to swiftly and accurately analyze large data sets for earlier detection and intervention Non-invasive methods, such as optical sensors and infrared imaging, improve patient comfort by avoiding blood draws, thereby enhancing compliance and facilitating routine screenings Additionally, AI-driven solutions can be scaled and tailored to diverse healthcare settings, potentially expanding access to diabetes screening in underserved regions Studies also underscore the cost-effectiveness of these innovative approaches, making them a promising option for broader diabetes detection.
57 effectiveness of AI technology in the long term, due to reduced need for consumables and the potential for automated monitoring
While AI-based diabetes detection systems show promise, they still face notable weaknesses, including accuracy challenges, as evidenced by an 84.4% accuracy rate with 3-label data, indicating a need for improvement in diverse clinical scenarios The performance of these systems heavily relies on high-quality input data, with poor data quality potentially hindering effectiveness Furthermore, high initial costs and technical complexity pose significant barriers to adoption, especially in resource-limited environments The integration of AI into existing healthcare workflows also demands extensive training and adjustments in clinical practices, making the transition a slow and challenging process.
Current literature highlights significant gaps in realizing the potential of AI for non-invasive diabetes detection Extensive validation studies across diverse populations are essential to ensure the generalizability and robustness of AI models, as many existing studies focus on homogenous datasets that fail to capture the full variability of the general population Furthermore, the absence of standardized protocols for data collection and analysis leads to inconsistent results, complicating the comparison of findings across studies Additionally, interdisciplinary research is crucial to integrate expertise from AI, medicine, and patient care, fostering the development of more comprehensive and user-friendly solutions.
Recent advancements in machine learning, especially deep learning techniques, have significantly improved accuracy by identifying complex data patterns Convolutional neural networks (CNNs) have been effectively utilized in image-based diagnostics, notably enhancing the detection of diabetic symptoms through retinal images and skin scans Furthermore, innovations in sensor technology have resulted in the creation of more sensitive and accurate non-invasive devices, like continuous glucose monitors that analyze interstitial fluid Collaborative initiatives between technology companies and healthcare providers are also fostering innovation in this field.
58 with initiatives aimed at integrating AI systems into electronic health records (EHRs) for seamless data flow and real-time monitoring
The current literature emphasizes the promising potential of AI for non-invasive diabetes detection, while also underscoring the necessity for ongoing research and development to address existing limitations and maximize the benefits of this technology.
8 HPLC High-performance liquid chromatography
13 SORS Spatially offset Raman spectroscopy
Centers for Disease Control and Prevention, 2014 National Diabetes Statistics Report:
According to the 2014 report by the U.S Department of Health and Human Services, the prevalence and impact of diabetes in the United States are significant The report provides comprehensive statistics on the national diabetes burden, highlighting the need for awareness and action For more detailed information, the full report can be accessed at the CDC's official website.
Lee, J.W., Brancati, F.L and Yeh, H.C., 2011 Trends in the prevalence of type 2 diabetes in Asians versus whites: results from the United States National Health Interview Survey, 1997-
2008 Diabetes Care, 34(2), pp.353-357 [PMC free article] [PubMed]
Nichols, G.A., Schroeder, E.B., Karter, A.J., et al., 2015 Trends in diabetes incidence among 7 million insured adults, 2006-2011: the SUPREME-DM project American Journal of
Epidemiology, 181(1), pp.32-39 [PMC free article] [PubMed]
Maruthur, N.M., 2013 The growing prevalence of type 2 diabetes: increased incidence or improved survival? Current Diabetes Reports, 13(6), pp.786-794 [PubMed]
Centers for Disease Control and Prevention, 2013 Diabetes Public Health Resource:
Incidence and Age at Diagnosis Available at: http://www.cdc.gov/diabetes/statistics/incidence_national.htm [Accessed 27 January 2015]
Anon, 2008 Economic costs of diabetes in the U.S in 2007 Diabetes Care, 31(3), pp.596-615 [PubMed]
Anon, 2014 Standards of medical care in diabetes-2014 Diabetes Care, 37(SUPPL.1), pp.S14- S80 [PubMed]
Inzucchi, S.E., Bergenstal, R.M., Buse, J.B., et al., 2012 Management of hyperglycemia in type
2 diabetes: A patient-centered approach: Position statement of the American Diabetes
Association (ADA) and the European Association for the Study of Diabetes (EASD) Diabetes Spectrum, 25(3), pp.154-171 [PMC free article] [PubMed]
UK Prospective Diabetes Study (UKPDS) Group, 1998 Effect of intensive blood-glucose control with metformin on complications in overweight patients with type 2 diabetes (UKPDS
The UK Prospective Diabetes Study (UKPDS) Group conducted a landmark study in 1998, comparing intensive blood-glucose control using sulphonylureas or insulin with conventional treatment in patients with type 2 diabetes Published in The Lancet, the study, known as UKPDS 33, revealed significant findings on the risk of complications in type 2 diabetes patients The research aimed to investigate the effects of intensive blood-glucose control on the development of complications in type 2 diabetes patients The study's results, published in volume 352, issue 9131, of The Lancet, provided valuable insights into the management of type 2 diabetes.
Duckworth, W., Abraira, C., Moritz, T., et al., 2009 Glucose control and vascular complications in veterans with type 2 diabetes The New England Journal of Medicine, 360(2), pp.129-139 [PubMed]
Gerstein, H.C., Miller, M.E., Byington, R.P., et al., 2008 Effects of intensive glucose lowering in type 2 diabetes The New England Journal of Medicine, 358(24), pp.2545-2559 [PMC free article] [PubMed]
Papademetriou, V., Lovato, L., Doumas, M., et al., 2014 Chronic kidney disease and intensive glycemic control increase cardiovascular risk in patients with type 2 diabetes Kidney
International, [online] Available at: https://www.kidney-international.org [Accessed 17
Lewis, I.R and Edwards, H.G.M., 2001 Handbook of Raman Spectroscopy: From the Research Laboratory to the Process Line New York, New York: CRC Press
Bazin, C., Cassard, B., Caudron, E., Prognon, P and Havard, L., 2015 Analysis of pharmaceuticals using Raman spectroscopy International Journal of Pharmaceutics, 494(1), pp.329-336
Mazurek, S and Szostak, R.J., 2006 Application of Raman spectroscopy for pharmaceutical analysis Journal of Pharmaceutical and Biomedical Analysis, 40(5), pp.1235-1242
Bourget, P., Amin, A., Vidal, F., Merlette, C and Lagarce, F., 2014 Pharmaceutical applications of Raman spectroscopy Journal of Pharmaceutical and Biomedical Analysis, 91, pp.176-184
Buckley, K., Atkins, C.G., Chen, D., Schulze, H.G., Devine, D.V., Blades, M.W and Turner, R.F.B., 2016 Advances in Raman spectroscopy for clinical diagnostics The Analyst, 141(5), pp.1678-1685
Weaver, J., 2011 Review of anesthesia monitoring technologies Anesthesia Progress, 58(3), pp.111-112
Van Wagenen, R.A., Westenskow, D.R., Benner, R.E., Gregonis, D.E and Coleman, D.L., 1986 Clinical monitoring using Raman spectroscopy Journal of Clinical Monitoring, 2(4), pp.215-
Lawson, D., Samanta, S., Magee, P.T and Gregonis, D.E., 1993 Raman spectroscopy in clinical monitoring Journal of Clinical Monitoring, 9(4), pp.241-251
Schlüter, S., Krischke, F., Popovska-Leipertz, N., Seeger, T., Breuer, G., Jeleazcov, C., Schüttler, J and Leipertz, A., 2015 Raman spectroscopic analysis in clinical diagnostics Journal of Raman Spectroscopy, 46(8), pp.708-715
Schlüter, S., Krischke, F., Popovska-Leipertz, N., Seeger, T., Breuer, G., Jeleazcov, C., Schüttler, J and Leipertz, A., 2014 Advances in laser applications for chemical and environmental analysis Laser Applications to Chemical, Security and Environmental Analysis
2014, LW4D.6, OSA, Seattle, Washington, p LW4D.6
Sung, H., Ferlay, J., Siegel, R.L., Laversanne, M., Soerjomataram, I., Jemal, A et al., 2021 Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for
36 cancers in 185 countries CA: A Cancer Journal for Clinicians, 71, pp.209-249
Harbeck, N., Penault-Llorca, F., Cortes, J., Gnant, M., Houssami, N., Poortmans, P et al.,
2019 Breast cancer review Nature Reviews Disease Primers, 5, pp.1-31
Siegel, R.L., Miller, K.D and Jemal, A., 2019 Cancer statistics, 2019 CA: A Cancer Journal for Clinicians, 69, pp.7-34
Milosevic, M., Jankovic, D., Milenkovic, A and Stojanov, D., 2018 Early diagnosis and detection of breast cancer Technology and Health Care, 26, pp.729-759
Sahu, R., 2016 Curcumin: A boon as antidiabetic International Journal of Green Pharmacy (IJGP) Available at: http://greenpharmacy.info/index.php/ijgp/article/view/3373
A network meta-analysis conducted by Lozano-Ortega et al (2016) evaluates various treatment options for type 2 diabetes mellitus in patients who have not responded adequately to a combination of metformin and sulfonylurea The findings, published in Current Medical Research and Opinion, provide valuable insights into alternative therapies for managing this condition For more information, the full study can be accessed at https://doi.org/10.1185/03007995.2015.1135110.
Ben’s Natural Health, 2021 The Role of Blood Glucose (Blood Sugar) in Diabetes Management Available at: https://www.bensnaturalhealth.com/blog/diabetes-health/glucose/.