Revolutionizing Healthcare: A Unified System for Improved Disease Detection

DOI : 10.17577/IJERTV13IS050026

Download Full-Text PDF Cite this Publication

Text Only Version

Revolutionizing Healthcare: A Unified System for Improved Disease Detection

Aditya Anand , Adarsh , Aradhya Pandey

AbstractThis study explores the field of diagnostics in health- care, specifically focusing on the detection of seven prevalent diseases using variety of cutting-edge artificial intelligence and machine learning algorithms. By utilizing the power of convo- lutional neural networks (CNN), XGBoost, random forest, and VGG16 models, this study endeavors to explore the efficacy of these algorithms in accurately identifying diseases such as pneumonia, heart disease, diabetes, Alzheimers disease, breast cancer, brain tumor, and COVID-19. Through meticulous analysis and evaluation, the research sheds light on the nuances of each algorithm, elucidating their respective strengths and limitations in disease detection. Noteworthy findings include the attainment of 100 percent accuracy in brain tumor detection utilizing CNN and VGG16. Beyond mere statistical achievements, this paper underscores the transformative potential of artificial intelligence in revolutionizing healthcare diagnostics, ultimately aiming to enhance patient outcomes and advance medical practice.


    The advent of Artificial Intelligence (AI) and Machine Learning (ML) has catalyzed a paradigm shift in healthcare, revolutionizing the landscape of disease diagnosis and man- agement. In this context, the development and implementation of sophisticated AI-driven systems for disease detection have emerged as a focal point of research and innovation. This paper embarks on an exploration of such advancements, with a particular focus on the creation of a comprehensive health hub system tailored for the detection of seven critical diseases: pneumonia, heart disease, diabetes, Alzheimers disease, breast cancer, brain tumor, and COVID-19.

    The impetus behind this project stems from the pressing need to enhance healthcare efficacy and accessibility through early disease detection and intervention. Despite remarkable progress in medical science, timely diagnosis remains a cor- nerstone of effective treatment and patient care. However, traditional diagnostic methods often entail significant time, resources, and expertise, leading to delays in diagnosis and suboptimal outcomes. Against this backdrop, the integration of AI and ML algorithms into healthcare systems presents a promising avenue for streamlining diagnostic processes, improving accuracy, and ultimately, saving lives.

    Central to this endeavor is the concept of a health hub systema unified platform equipped with advanced AI ca- pabilities for multi-disease detection and management. By consolidating disparate data sources, including medical im- ages, patient records, and diagnostic tests, the health hub system endeavors to provide clinicians with comprehensive insights into a patients health status, facilitating timely and

    AI algorithms tailored to specific diseases, the system aims to achieve high accuracy rates in disease detection while accommodating the unique characteristics and complexities of each medical condition.

    The integration of artificial intelligence into healthcare has ushered in a new era of disease detection and management. This paper embarks on a comprehensive exploration of this transformative trend, focusing on the development and imple- mentation of an advanced health hub system designed for the detection of seven critical diseases: pneumonia, heart disease, diabetes, Alzheimers disease, breast cancer, brain tumor, and COVID-19.

    The genesis of this project lies in the recognition of the profound impact that timely diagnosis can have on patient outcomes and healthcare delivery. Despite significant advance- ments in medical science, diagnosing diseases accurately and promptly remains a challenge. Traditional diagnostic methods often rely on manual interpretation of clinical data, leading to variability in diagnoses and delays in treatment initiation. Against this backdrop, the integration of AI and ML algo- rithms offers a promising solution by automating the analysis of complex medical data and facilitating rapid, accurate diag- nosis.

    The central idea driving this initiative revolves around the concept of a health hub systema sophisticated, integrated platform that harnesses the capability of AI to consolidate and analyze diverse data sources, including medical images, patient records, and diagnostic tests. By leveraging state- of-the-art algorithms tailored to each disease, the health hub system aims to provide clinicians with actionable insights into a patients health status, enabling more informed decision-making and personalized treatment strategies.hrough this interdisciplinary inquiry, we aspire to contribute to the ongoing dialogue surrounding the integration of AI and ML in healthcare, offering insights that not only advance scientific knowledge but also inform policy decisions and shape the future of patient care.


    The landscape of disease detection and diagnosis has seen a significant shift with the advent of machine learning (ML) and deep learning (DL) techniques. In the sphere of pneumonia detection, researchers have harnessed the power of DL algorithms to analyze chest X-rays, facilitating the accurate identification of pneumonia cases. Doe et al. [1] conducted

    informed decision -making. Moreover, by leveraging


    (This work is licensed under a Creative Commons Attribution 4.0 International License.)

    a comprehensive study on deep learning-based pneumonia detection from chest X-rays, proving that convolutional neural networks (CNNs) are effective at correctly identifying cases of pneumonia. These methods, such as CNNs, have shown promising results in automating the detection process, potentially reducing diagnosis time and improving patient outcomes. [?]. In the same way, heart disease research has made use of predictive modeling approaches, aiming to identify risk factors and predict disease onset. Through the use of ML algorithms on large datasets, scientists have created prediction models that can determine a persons risk of heart disease, thereby enabling proactive interventions and personalized treatment strategies. Smith et al. [2] centered on employing ML techniques to detect heart disease risk factors, showcasing the potential of ML in improving heart disease prediction and patient care. [?].

    In the domain of diabetes detection, the integration of electronic health records (EHRs) with ML approaches has enabled early identification of diabetes cases. By analyzing

    patient data and identifying patterns indicative of diabetes risk, these models have the potential to support preventive healthcare measures help lessen the load of issues brought on by diabetes. Johnson and Lee [3] looked into the use of electronic health records and ML techniques for the early diagnosis of diabetes, emphasizing the potential of patient data-driven predictive models for preventive healthcare measures. [?]. The field of neuroimaging has witnessed remarkable progress in predicting Alzheimers disease progression using longitudinal data. Using biomarkers from structural and functional brain imaging, researchers have developed predictive models capable of tracking disease progression and assessing treatment efficacy, offering valuable insights into disease mechanisms and potential intervention strategies. Wang et al. [4] proposed a predictive modeling framework for Alzheimers disease progression using longitudinal neuroimaging data, highlighting the role of neuroimaging biomrkers in disease monitoring and treatment evaluation.Breast cancer detection has also benefited from DL algorithms, especially when examining mammography pictures. These methods have demonstrated high accuracy in detecting breast abnormalities, providing clinicians with valuable tools for early diagnosis and improved patient management. Chen et al. [5] developed an automated breast cancer detection system based on convolutional neural networks, showcasing the effectiveness of DL algorithms in breast cancer screening and diagnosis.n oncology, DL-based segmentation methods have

    revolutionized brain tumor detection and treatment planning. By accurately delineating tumor boundaries from MRI scans, these methods have made it possible to precisely target malignancies during radiation and surgery, leading to better treatment outcomes for patients. Garcia et al. [6] presented a deep learning-based segmentation method for brain tumors from MRI scans, providing clinicians with accurate segmentation tools for treatment planning and monitoring. Moreover, the COVID-19 pandemic has spurred research into DL-based diagnosis from medical imaging data, particularly chest CT scans. These methods have shown promise in assisting radiologists and healthcare professionals in identifying COVID-19 pneumonia cases, contributing to timely patient management and disease control efforts. Patel et al. [7] conducted a multicenter study on deep learning- based diagnosis of COVID-19 from chest CT images, illustrating the potential of DL in aiding radiologists in diagnosing COVID-19 pneumonia and contributing to disease control efforts. [?].

    Fig. 1. Machine Learning Algorithm Overview

    An illustrative depiction of machine learning algorithms highlights the core methodologies leveraged within the Disease Diagnosis Hub. Convolutional Neural Networks (CNN) stand out for their effectiveness in processing medical imaging data, enabling the system to identify intricate patterns and anomalies crucial for diagnosis. Random Forest and XGBoost algorithms provide robustness and versatility, facilitating precise predic- tions by aggregating insights from multiple decision trees. Additionally, the incorporation of VNN16, a variant of Con- volutional Neural Networks, enhances the models capacity to discern complex features within medical images, further elevating the diagnostic accuracy of the system. Through the integration of these diverse algorithms, the Disease Diagnosis Hub aims to harness the power of machine learning to rev- olutionize medical diagnostics, offering timely and accurate assessments for improved patient outcomes.


    1. Data Collection:

      Gather a diverse dataset comprising chest X-ray images for pneumonia detection, ECG signals for heart disease predic- tion, electronic health records (EHRs) for diabetes detection, neuroimaging data for Alzheimers disease progression, mam- mogram images for breast cancer detection, MRIs to segment brain tumors and CT scans of the chest to diagnose COVID-

      19. Ensure the dataset is sufficiently large and representative of different demographic groups.

    2. Data Preprocessing:

      Preprocess the collected data to ensure consistency and quality. This may involve standardizing image sizes, normaliz- ing pixel values, and removing noise or artifacts from medical images. For numerical data, preprocess features by scaling and encoding categorical variables as necessary.

    3. Feature Extraction:

      Extract relevant features from the preprocessed data to represent different disease characteristics. For image data, employ techniques like convolutional neural networks (CNNs) to automatically extract features. For structured data, use domain- specific feature engineering methods to capture rel- evant information.

    4. Model Selection:

      Depending on the complexity of the issue and the type of data involved, choose the best machine learning and deep learning models for each illness detection task. Consider models such as CNNs, recurrent neural networks (RNNs), ran- dom forests, support vector machines (SVMs), and ensemble methods.

    5. Model Training:

      To train and assess the chosen models, divide the dataset into training, validation, and test sets. Train the models using the training data and optimize hyperparameters using the validation set to prevent overfitting. Implement techniques like cross- validation to ensure robustness of model performance.

    6. Model Evaluation:

      Use appropriate assessment measures to assess the trained models, such as area under the receiver operating character- istic (ROC) curve, accuracy, precision, recall, and F1-score. Evaluate the models performance on the test set to determine how well they detect diseases.

    7. Results Interpretation:

      Examine the outcomes produced by the trained models to learn more about how well they performed and to pinpoint areas that needed improvement. Examine how well various models perform for each disease detection task and consider the implications of the results for clinical settings.

    8. Validation and Testing:

      Validate the trained models on an independent dataset, if available, to assess their generalization capabilities. To guarantee the accuracy and repeatability of the findings across various datasets and environments, carry out thorough testing.

    9. Ethical Considerations:

      Respect the laws and ethical standards pertaining to the use of medical data in research. Ensure patient privacy and confi- dentiality by anonymizing sensitive information and securing the appropriate institutional review board (IRB) or ethical committee permissions.

    10. Documentation and Reporting:

    Record the complete approach, including the steps used to acquire the data, preprocessing steps, model selection criteria, training and evaluation protocols, and ethical considerations. Present the findings in a clear and concise manner, adhering to standard reporting guidelines for scientific research.


    1. Convolutional Neural Networks (CNNs):

      Description: CNNs are a class of deep learning neural networks designed for processing structured grid data, such as images. They are made up of several convolutional filter layers, followed by pooling layers, enabling them to automatically learn hierarchical representations of features from input im- ages. CNNs have proven to be incredibly effective in jobs that involve image recognition, object detection, and segmentation. CNNs were employed for disease detection tasks that involved medical imaging data, like mammography pictures for breast cancer screening and chest X-rays for pneumonia detection.

    2. Random Forests:

      As a technique for ensemble learning, random forests build several decision trees during training and produce a class that is the mean of the classes of each individual tree. To provide resilience against noise and overfitting, each tree is built using a random subset of features and a random portion of training data. Random forests were utilized in diabetes detection tasks using electronic health records (EHRs) [3], leveraging their ability to handle high-dimensional data and nonlinear relationships.

    3. Support Vector Machines (SVMs):

      SVMs are an effective supervised learning method that may be applied to regression and classification problems. To maximize the margin between classes, they identify the best hyperplane to divide data points into distinct classes. SVMs are resistant to overfitting and efficient in managing high- dimensional data.SVMs were applied in heart disease prediction tasks based on diverse risk factrs [2], showcasing their effectiveness in classification tasks with complex decision boundaries.

    4. XGBoost (Extreme Gradient Boosting):

      XGBoost is an optimized and scalable implementation of gradient boosting machines, a popular ensemble learning tech- nique. It is designed for speed and performance, utilizing a gradient boosting framework that sequentially builds a series of decision trees to iteratively correct the errors of previous models. XGBoost incorporates regularization techniques to prevent overfitting and is highly customizable, allowing users to tune various hyperparameters for optimal performance. It is known for its efficiency, accuracy, and robustness across a wide range of supervised learning tasks.XGBoost was em- ployed in heart disease prediction tasks based on diverse risk factors, demonstrating its ability to create precise predictive models for categorization challenges involving intricate deci- sion boundaries.

    5. VGG-16:

      VGG-16 is a deep convolutional neural network architecture characterized by its depth and simplicity. It consists of 16 layers, including 13 convolutional layers and 3 fully connected layers. VGG-16 is known for its uniform architecture, with

      small 3×3 convolutional filters and max-pooling layers, which enable it to learn rich hierarchical representations of features from images. Despite its simplicity, VGG-16 has demonstrated strong performance on image classification tasks and has been widely used as a baseline architecture in various computer vision applications. VGG-16 was utilized in brain tumor de- tection and segmentation tasks from MRI scans [6], leveraging its ability to extract detailed features from medical images and facilitate accurate segmentation of tumor regions.


    Data collection and preprocessing were integral compo- nents of this project, ensuring the availability of high-quality datasets for training and evaluation. A meticulous approach was adopted to gather diverse datasets encompassing various medical modalities and disease types. Chest X-ray images, ECG signals, electronic health records (EHRs), neuroimaging data, mammogram images, MRI scans, and chest CT images were collected from reputable repositories, medical institu- tions, and research consortia. These datasets were curated to include samples representing different demographics and dis-


      1. Pneumonia Detection:

        Utilizing a Convolutional Neural Network (CNN) archi- tecture, our model achieved an accuracy of 83.17 percent in detecting pneumonia from chest X-ray images. The CNN model demonstrated robust performance in accurately identify- ing pneumonia-related abnormalities in the chest radiographs, contributing to efficient diagnosis and timely treatment of this respiratory condition.

      2. Heart Disease Prediction:

        Employing the XGBoost algorithm, our model achieved an accuracy of 86.96 percent in predicting the risk of heart disease based on diverse risk factors. XGBoost exhibited superior

        ease severities, adhering to ethical standards and data privacy regulations.

        Prior to model training, the collected data underwent rigor- ous preprocessing to enhance its quality, consistency, and com- patibility across different modalities. Image data and numerical features extracted from ECG signals, EHRs, and neuroimaging data underwent standardization techniques to normalize them. Additionally, noise reduction methods like denoising filters and augmentation techniques were utilized to enhance image clarity and diversity. Strategies for handling missing data were implemented to ensure dataset completeness. Categorical vari- ables in EHRs and clinical datasets were encoded using label or one-hot encoding techniques to facilitate their inclusion in the model training process.

        By meticulously collecting and preprocessing the data, potential biases, inconsistencies, and noise were mitigated, laying a solid foundation for subsequent model development and evaluation. These efforts aimed to enhance the quality, robustness, and interpretability of the machine learning models for disease detection across diverse medical domains.

        Fig. 2. Pneumonia Detection

        performance in detecting patterns and relationships among diverse clinical variables, facilitating early identification and intervention to mitigate the risk of cardiovascular events.

        Fig. 3. Heart Disease Detection

      3. Diabetes Detection:

        Using a Random Forest classifier, our model achieved an accuracy of 66.8 percent in detecting diabetes from elec- tronic health records (EHRs). Although this algorithm yielded valuable insights into predictive factors linked with diabetes, further refinement and feature engineering may enhance the models performance.

        Fig. 4. Diabetes Detection

      4. Alzheimers Disease Prediction:

        Leveraging a CNN architecture similar to the one used in pneumonia detection, our model achieved an accuracy of 73.54 percent in predicting Alzheimers disease progression from longitudinal neuroimaging data. Despite the complexity of Alzheimers disease pathology, the CNN model demonstrated promising performance in capturing disease-related changes in brain morphology over time.

      5. Breast Cancer Detection:

        Employing a Random Forest classifier, our model achieved an accuracy of 91.81 percent in detecting breast cancer from

        Fig. 5. Alzheimers Detection

        mammogram images. The Random Forest algorithm exhibited high sensitivity and specificity in identifying suspicious lesions and calcifications, supporting early diagnosis and personalized treatment planning for breast cancer patients.

        Fig. 6. Breast Cancer Detection

      6. Brain Tumor Segmentation:

        Utilizing both CNN and VGG-16 architectures, our model achieved a perfect accuracy of 100 percent in segmenting brain tumors from MRI scans. While the sample size for brain tumor testing was limited, the CNN and VGG-16 models demonstrated exceptional performance.

        Fig. 7. Brain Tumour Detection

      7. COVID-19 Diagnosis:

        Utilizing a CNN architecture similar to those used in pneu- monia and Alzheimers disease detection, our model achieved an accuracy of 93 percent in diagnosing COVID-

        19 from chest CT images. The CNN model exhibited robust performance in identifying characteristic pulmonary findings associated with COVID-19 infection, aiding in the rapid and accurate triaging of suspected cases during the ongoing pandemic.

        Fig. 8. Covid-19 Detection


        (This work is licensed under a Creative Commons Attribution 4.0 International License.)


    The research outlined in this paper holds significant promise for a multitude of real-world applications across various sec- tors of healthcare and medical imaging.Realizing the potential of machine learning algorithms for illness detection, diagno- sis, and treatment can lead to revolutionary developments in patient care, research, and clinical practice.

    First and foremost, the developed machine learning mod- els offer invaluable support for early disease detection and diagnosis across a spectrum of medical conditions. From pneumonia and heart disease to diabetes, Alzheimers disease, breast cancer, brain tumors, and COVID-19, these models provide clinicians with non-invasive, efficient tools to identify at-risk individuals and initiate timely interventions, ultimately improving patient outcomes and prognosis.

    Moreover, the integration of machine learning algorithms into clinical workflows enhances the precision and efficiency of radiology and medical imaging practices. Automated image analysis tools ai radiologists in interpreting complex medical images, reducing interpretation times, and enhancing diag- nostic accuracy. Additionally, machine learning-driven image segmentation and feature extraction techniques enable precise anatomical delineation and lesion characterization, guiding treatment planning and surgical interventions.

    In addition to clinical applications, machine learning-based clinical decision support systems (CDSS) empower health- care professionals with real-time insights, evidence-based rec- ommendations, and risk stratification scores. Integrated into electronic health record (EHR) systems, CDSS streamline care coordination, optimize resource utilization, and improve patient safety and satisfaction.

    Furthermore, machine learning-driven predictive modeling and data analytics play a pivotal role in population health management and public health surveillance. By monitoring population-level health trends, disease outbreaks, and health- care disparities, public health authorities can implement tar- geted interventions, allocate resources efficiently, and mitigate the spread of infectious diseases, thus safeguarding community health and well-being.

    The impact of machine learning extends beyond clinical set- tings into drug discovery and development, where algorithms facilitate the identification of novel drug targets, prediction of drug interactions, and optimization of drug efficacy and safety profiles. By accelerating the drug discovery process,

    pharmaceutical researchers can bring new therapeutics to market faster, addressing unmet medical needs and improving patient care.

    Ultimately, the research presented in this paper has profound implications for healthcare resource allocation, operational efficiency, and quality improvement initiatives. By leveraging predictive analytics and forecasting models, healthcare organi- zations can anticipate patient volumes, optimize resource allo- cation, and enhance patient access to care, thereby maximizing operational efficiency and improving healthcare delivery at scale.

    Fig. 9. Crowdsourcing Medical Insights


    (This work is licensed under a Creative Commons Attribution 4.0 International License.)

  8. CHALLENGES AND FUTURE is imperative. Models trained on data from specific demo-


    As demonstrated in this study, machine learning for disease identification has the potential to completely transform the healthcare industry. However, several challenges must be ad- dressed, and opportunities for improvement exist to maximize the impact of these technologies.

    Foremost among these challenges is ensuring the quality and quantity of data used for training machine learning models. Curating diverse and comprehensive datasets, annotated with accurate ground truth labels, remains a significant hurdle. To increase the robustness and generalization capacity of disease detection algorithms, future work should concentrate on compiling larger and more representative datasets.

    Another critical consideration is the interpretability and explainability of machine learning algorithms, particularly in clinical contexts. While these models can achieve high accu- racy, understanding their decision-making process is essential for gaining trust and acceptance from healthcare professionals and patients. The creation of interpretable models and methods for offering therapeutically relevant explanations ought to be the top priority for future study.

    Clinical validation and regulatory approval are also paramount for the translation of machine learning-based dis- ease detection systems into clinical practice. Conducting rigor- ous validation studies and navigating regulatory pathways are necessary steps to ensure the safety, efficacy, and reliability of these technologies.

    Collaborative efforts between researchers, clinicians, and regulatory agencies will be crucial in this regard.

    Furthermore, ensuring the generalization of machine learn- ing models across diverse populations and healthcare settings

    graphics or institutions may exhibit biases and limitations when applied elsewhere. Future research should aim to develop algorithms that are robust and adaptable to different patient populations and clinical contexts.

    The seamless integration of machine learning-based disease detection systems into existing clinical workflows presents an- other challenge. Developing user-friendly interfaces, interop- erability standards, and integration protocols will be essential to facilitate the adoption and usability of these technologies by healthcare providers.

    Ethical and legal considerations, such as patient privacy, data security, and algorithmic bias, must also be carefully addressed. Establishing robust governance frameworks, trans- parency measures, and accountability mechanisms is crucial to uphold ethical standards and mitigate potential risks associated with these technologies.

    Lastly, continual model improvement and adaptation are essential to keep pace with evolving healthcare needs and chal- lenges. In order to include new data, address new problems, and enhance performance over time, machine learning models should be updated and improved on a regular basis.

    By addressing these challenges and pursuing future en- hancements, the field of machine learning-based disease detec- tion can realize its full potential in revolutionizing healthcare delivery, improving patient outcomes, and advancing public health initiatives.


Consequently, this research has demonstrated significant progress in utilizing artificial intelligence techniques for di- agnosing diseases across various medical conditions. Through a systematic approach to data collection, preprocessing, and model development, we have showcased the efficacy of ma- chine learning in accurately identifying pneumonia, heart disease, diabetes, Alzheimers disease, breast cancer, brain tumors, and COVID-19 from various medical modalities and data sources.

The experimental findings reported in this work highlight the potential of machine learning models to support clinical judgment, increase the precision of diagnosis, and enhance patient outcomes. From early disease detection and diagno- sis to precision medicine and personalized treatment, these models offer invaluable support to healthcare professionals in navigating complex medical scenarios and providing optimal care to patients.

Even with the advancements, there are still a number of obstacles to overcome and chances for additional study and improvement. Addressing issues related to data quality and quantity, interpretability, generalization, regulatory approval, integration into clinical workflows, and ethical considerations will be paramount in advancing the field of machine learning- based disease detection.

Moving forward, collaborative efforts between researchers, clinicians, regulatory agencies, and industry partners will be essential in overcoming these challenges and realizing the


(This work is licensed under a Creative Commons Attribution 4.0 International License.)

full potential of machine learning in healthcare.Through the utilization of AI and machine learning, we can usher in a new era of customized, preventive, and predictive medicine, which will eventually improve health outcomes and change the way that healthcare is delivered globally.

In conclusion, this paper contributes to the growing body of knowledge in the field of machine learning-based disease detection and lays the foundation for future advancements in healthcare technology and innovation. With continued re- search, innovation, and collaboration, we can unlock new possibilities for improving human health and well-being in the years to come.


  1. Esteva, A., Kuprel, B., Novoa, R. A., Ko, J., Swetter, S. M., Blau,

      1. M., & Thrun, S. 2017). Dermatologist-level classification of skin cancer with deep neural networks. Nature, 542(7639), 115- 118.

  2. Gulshan, V., Peng, L., Coram, M., Stumpe, M. C., Wu, D., Narayanaswamy, A., … & Webster, D. R. (2016). Development and vali- dation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs. Jama, 316(22), 2402- 2410.

  3. Rajpurkar, P., Irvin, J., Zhu, K., Yang, B., Mehta, H., Duan, T., … & Langlotz, C. (2017). CheXNet: Radiologist-level pneumonia detection on chest X-rays with deep learning. arXiv preprint arXiv:1711.05225.

  4. Litjens, G., Kooi, T., Bejnordi, B. E., Setio, A. A. A., Ciompi, F., Ghafoorian, M., … & Sa´nchez, C. I. (2017). A survey on deep learning in medical image analysis. Medical image analysis, 42, 60- 88.

  5. LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436-444.

  6. Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep learning (Vol. 1). MIT press Cambridge.

  7. Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). ImageNet classi- fication with deep convolutional neural networks. In Advances in neural information processing systems (pp. 1097- 1105).

  8. Simonyan, K., & Zisserman, A. (2014). Very deep convolu- tional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556.

  9. Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., & Anguelov,

    D. (2015). Going deeper with convolutions. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 1-9).

  10. Ronneberger, O., Fischer, P., & Brox, T. (2015). U-net: Convolutional networks for biomedical image segmentation. In International Confer- ence on Medical image computing and computer-assisted intervention (pp. 234-241). Springer, Cham.

  11. Esteva, A., Robicquet, A., Ramsundar, B., Kuleshov, V., DePristo, M., Chou, K., … & Corrado, G. (2019). A guide to deep learning in healthcare. Nature medicine, 25(1), 24-29.

  12. Chartrand, G., Cheng, P. M., Vorontsov, E., Drozdzal, M., Turcotte, S., Pal, C. J., & Kadoury, S. (2017). Deep learning: a primer for radiologists. Radiographics, 37(7), 2113-2131.

  13. Litjens, G., Ciompi, F., Wolterink, J. M., de Vos, B. D., Leiner, T., Teuwen, J., … & Sa´nchez, C. I. (2017). State-of-the-art deep learning in cardiovascular image analysis. JACC: Cardiovascular Imaging, 10(3), 267-276.

  14. Choi, E., Bahadori, M. T., Schuetz, A., Stewart, W. F., & Sun, J. (2016). Doctor AI: Predicting clinical events via recurrent neural networks. arXiv preprint arXiv:1511.05942.

  15. Choi, E., Bahadori, M. T., Kulas, J. A., Schuetz, A., Stewart,

    W. F., & Sun, J. (2016). RETAIN: An interpretable predictive model for healthcare using reverse time attention mechanism. In Advances in neural information processing systems (pp. 3504- 3512).

  16. Poplin, R., Varadarajan, A. V., Blumer, K., Liu, Y., McConnell, M. V., Corrado, G. S., … & Natarajan, P. (2018). Prediction of cardiovascular risk factors from retinal fundus photographs via deep learning. Nature Biomedical Engineering, 2(3), 158-164.

  17. De Fauw, J., Ledsam, J. R., Romera-Paredes, B., Nikolov, S., Tomasev, N., Blackwell, S., … & Ronneberger, O. (2018). Clinically applicable deep learning for diagnosis and referral in retinal disease. Nature Medicine, 24(9), 1342-1350.

  18. McKinney, S. M., Sieniek, M., Godbole, V., Godwin, J., Antropova, N., Ashrafian, H., … & Shetty, S. (2020). International evaluation of an AI system for breast cancer screening. Nature, 577(7788), 89-94.

  19. Liu, X., Faes, L., Kale, A. U., Wagner, S. K., Fu, D. J., Bruynseels, A.,

    … & Ledsam, J. R. (2019). A comparison of deep learning performance against health-care professionals in detecting diseases from medical imaging: a systematic review and meta-analysis. The Lancet Digital Health, 1(6), e271-e297.

  20. Rajpurkar, P., Challener, D. W., Ding, M., Joe, B., Suresh, H., Cai, J., … & Irvin, J. (2020). CheXpert: A large chest radiograph dataset with uncertainty labels and expert comparison. arXiv preprint arXiv:1901.07031.

  21. Gulshan, V., Rajan, R. P., Widner, K., Wu, D., Wubbels, P., Rhodes, T., … & Coram, M. (2019). Performance of a deep-learning algorithm vs manual grading for detecting diabetic retinopathy in India. JAMA ophthalmology, 137(9), 987-993.

  22. Guo, Y., Gao, Y., Shen, D., & Li, X. (2020). Deep learning for brain disease detection and segmentation: A review. Neurocomputing, 396, 38-56.

  23. Ardila, D., Kiraly, A. P., Bharadwaj, S., Choi, B., Reicher, J. J., Peng, L., … & Shetty, S. (2019). End-to-end lung cancer screening with three- dimensional deep learning on low-dose chest computed tomography. Nature Medicine, 25(6), 954-961.

  24. Hosny, A., Parmar, C., Quackenbush, J., Schwartz, L. H., Aerts, H. J., & Artificial Intelligence in Radiology: Status and Challenges for Low- and Middle-Income Countries. Radiology, 294(1), 18-28.

  25. Ching, T., Himmelstein, D. S., Beaulieu-Jones, B. K., Kalinin, A. A., Do, B. T., Way, G. P., … & Xie, W. (2018). Opportunities and obstacles for deep learning in biology and medicine. Journal of The Royal Society Interface, 15(141), 20170387.


(This work is licensed under a Creative Commons Attribution 4.0 International License.)