The Human Diagnosis Project (also referred to as "Human Dx" or "the Project") is a worldwide effort created with and led by the global medical community to build an online system that maps the best steps to help any patient. By combining collective intelligence with machine learning, Human Dx intends to enable more accurate, affordable, and accessible care for all.
The Project is structured as a partnership between the social, public, and private sectors. Its partners include many of the top medical boards, societies, and academic centers.
Already, the Project is impacting medical training as well as timely and effective diagnosis, particularly in underserved communities. Over time, the Project has the potential to considerably alter the cost of, access to, and effectiveness of health care globally.
The Project's creation and development was inspired by other civilization-scale scientific projects (e.g., The International Space Station, the Large Hadron Collider at CERN, the Human Genome Project) and open technology efforts (e.g., Wikipedia, Linux, the Internet Protocol Suite).
To nearly every person on Earth, the wellbeing of oneself and one's loved ones is the most important concern. However, perhaps the most essential question of human wellbeing is still very difficult to answer: when someone has a health problem, what should be done?
Patients, doctors, insurers, foundations, governments, and many other health stakeholders struggle with this question. The information we seek is siloed in medical research, academic handbooks, disparate closed systems, and the minds of individual practitioners. However, the answers are not readily available in a single, open system that is universally accessible.
Diagnosis is a crucial early step in assessing, treating, managing, and ultimately curing disease. By improving our fundamental understanding of diagnosis, the Project could elevate wellbeing for current and future generations.
The Mission of the Human Diagnosis Project is to empower anyone with the world's collective medical insight. Applications range from helping an individual recognize when and where to seek professional medical attention, to helping a world-class specialist physician consider options for a complex patient presenting with a rare illness.
The Project does this by enabling the collaborative development of a fundamental data structure that can enable the potential of other health data sources. Ultimately, any patient, doctor, other individual, organization, device, or application will be able to access this information to make more informed decisions.
The Project intends to include all relevant categories of data, as well as the relevant details and relationships among them, to encode the richness of effective medical diagnosis. This includes: symptoms; physical exams; medical, social, and family histories; medical sensor and device data; diagnostic and laboratory tests; medical imaging; genomics; epigenomics; proteomics; published medical research; and health outcomes data.
The Project is building a common frame of reference for all stakeholders in the health system. This is conceptually similar to an online map. Just as an online map helps you get from one location to another, the Project gives patients, family members, doctors, hospitals, and others a shared path to helping any person.
Today, online maps make navigating the world faster and more efficient. They help us move seamlessly from origin to destination by synthesizing valuable insight from information sources such as speed limit, traffic, accident, weather, event, and topography data. Without geographic locations, all of the other information sources are not nearly as useful. As an example, using traffic data alone to determine how to get from one place to another is difficult without knowing the geographic locations of where the traffic is located.
In health care, many disparate sources of data exist, but we lack a comprehensive "map" to guide patients from an initial health problem to triage, diagnosis, treatment, discharge, ongoing care, and prevention. The Project hopes to facilitate the construction and broad distribution of such a system.
Our Vision is a world in which each person benefits from and contributes to the collective insights of each other (including doctors, patients and organizations). Such a system of collective intelligence would underlie all public health and health care systems so that information would be quickly and cheaply distributed to everyone in need. It would empower each stakeholder so that he or she could use the information to make the best decisions for themselves, their loved ones and their patients.
Over time, humans (e.g., patients, nurses, and physicians), machines (e.g., wearable devices, mobile phones, medical sensors/devices), and organizations (e.g., governments, multilaterals, health systems, insurers, employers) will work in concert to collect, analyze, and interpret real-time medical data. In addition to improving a patient's daily behaviors and habits, these parties will enable connected, distributed systems to effectively identify, treat, and prevent the development of disease.
For example, an integrated system could recognize patterns in a patient's physical activity, sleep patterns, mobile phone use, etc., to provide constructive recommendations on how the patient could effectively sustain or improve his or her general health and fitness. The system would encourage healthy practices by providing personalized information on how the patient's lifestyle influences his or her overall wellbeing and the next best sets of actions to potentially improve the patient's health status (e.g., going for a walk, drinking more water, getting more sleep, etc.).
Furthermore, the patient's desired doctors and other medical providers could access this data and the system's personalized recommendations at any time from anywhere in the world. If the system detected a potentially serious health anomaly, the patient's providers could receive automated alerts and related details. With the permission of the patient, the identification, prioritization, and coordination of care could potentially begin prior to an immediate problem. Thus, a patient who formerly would have received medical care during the latter stages of a given health problem's development could now receive accurate, preventative care well before.
The Human Diagnosis Project intends to facilitate many great organizations and individuals that are interested in making this Vision possible. When such a system is ultimately built, it must be readily accessible and open to all.
The need is urgent. Currently, one billion people lack access to health care, fewer than two physicians are available per 1,000 patients globally, and 100 million people are pushed into poverty each year because of their medical costs (Source: World Health Organization). The challenges and opportunities presented by these growing trends are immense, but not unsolvable. Novel approaches must be implemented to make health care affordable, accessible, and accurate.
Over the next several years, global mobile device penetration is likely to more than double in scale. Patients, doctors, and other medical providers who lack sufficient access to appropriate care and cost-efficient resources represent a significant portion of this growth trend. Additionally, many other types of connected devices (e.g., wearables, sensors, appliances, etc.) are coming online in the next few decades and are enabling us to collect useful information relevant to wellbeing that was never previously possible.
With these advances we have the ability, and the responsibility, to tackle humankind's most challenging problems.
As a synthesis of the global medical community's collective insight, the Project will initially support medical professionals in medical training, diagnosis, and effective decision-making. It also has applications in numerous other fields, including: medical training, patient medical information, online search, web content sites, appointment booking, telemedicine visits, insurance fraud detection, clinical trial enrollment, research funding allocation, continuing physician education, medical sensors, and the internet of things.
However, there are many applications of the Project that have not yet been conceptualized, and will only become apparent over time through the novel ideas of doctors, patients, engineers, and organizations from around the world.
The Project has no definite endpoint, and will continue to collect and interpret medical data as long as it has the support and resources to do so. The Project's first goal is to map any health problem to all of the diagnoses listed in the World Health Organization's International Classification of Diseases. Though it is difficult to predict how long this phase will take, the current expectation is years, not decades.
Additionally, unlike other civilization-scale scientific projects, the Human Diagnosis Project will not take decades for its potential applications to be of widespread use. Its efforts will also be initially applicable to people in many geographies and socioeconomic classes, not just a subset of the most fortunate.
The Project is already being pilot-tested to impact point-of-care clinical decision making among underserved patients in the United States. Other initial applications of the Project's data will hopefully be available in the next few years.
The Project is organized and led by the global medical community. In the recent past, an increasing number of parties have inserted themselves between patients and the medical community, marginalizing the doctor-patient relationship that serves as the foundation of medicine.
The Project empowers the global medical community to help tackle one of the most challenging problems in modern health on behalf of patients and themselves. While contributing to the Mission, health professionals and students can enhance their clinical abilities and benefit from the community's collective insights.
Unlike other projects at the intersection of medicine and machine learning, many of which state their intention to "replace the doctor," the Human Diagnosis Project is created with, designed for, and led by the medical community. The combination of collective human intelligence with machine learning has the potential to be vastly more powerful than either on its own.
Similar to how people around the world contribute encyclopedia articles to Wikipedia, or engineers contribute code to open source software projects like Linux, the global medical community submits clinical case contributions to Human Dx. The Project extends upon the ideas of open technology efforts like Wikipedia and Linux by bringing together multiple other communities (including the patient, scientific, and technology) with the medical community to ensure the creation and validation of clinical case data.
This can be done from any connected device. As medical practitioners, residents, and students give and receive input on clinical cases within Human Dx, the open system automatically contextualizes their decisions and individual clinical insights.
The primary reason why contributors from the medical community lend their efforts to the Project is a shared interest in the Project's Mission, with its potential to elevate wellbeing for all.
Within minutes, any member of the medical community can collaborate on clinical cases from around the world. Practical benefits to contributors include:
The Project's data structure is generated when the medical community's case contributions intersect with the Project's technologies. Contributions currently take place via web and smartphone, though longer-term from any connected device. This enables the fast, scalable, and distributed creation of one of the most highly structured and high-fidelity medical data sets that currently exists.
As contributors interact with the Project, the system encodes structured data based on their thought processes and decisions. To do so, Human Dx employs medical ontologies, information retrieval, semantic text extraction, natural language processing, computer vision, machine learning, data mining, and other techniques. Though there are many ways for contributors to interact with the Project, the primary ones are creating and solving case contributions.
When contributors create a case contribution, they initially receive a prompt based on their preferences and capabilities, as well as the Project's immediate requirements. They then begin creating around this prompt, which could include constraints like the patient's diagnosis, background (e.g., age, gender, race, etc.), symptoms, medical history, etc. They can draw upon an actual case they have seen in the field, develop a plausible simulation of the patient, or something in-between. Contributors start by adding medical findings that describe the given case. These findings could be any medical information described in any form of content (e.g., a text description of a physical exam result, an image of an X-Ray, or an audio recording of lung sounds). As findings are added, their details can also be specified as appropriate. The created case is then sent to the Human Dx central repository for other contributors to solve.
When contributors solve a case contribution, they first receive a partially revealed set of findings associated with that case. They then create a list of possible diagnoses (referred to in medicine as a "differential diagnosis" or "differential") based on the information they currently have. Contributors update their differentials as additional findings are subsequently revealed, and the case is only verified if a contributor can successfully solve the case. Because the contributor who created the contribution is not revealed to the contributors who are attempting to solve the contribution, the system learns whether the case appears to be reliable. Both correct and incorrect attempts to solve help the model become more accurate, as long as they are identified as credible attempts to be correct.
Human Dx verifies quality and accuracy using the collective insights of the medical community and comparing this against the Project's stored data and algorithms.
Contributions are not validated until a member of the Project community solves the contribution without knowing the identity of the contributor. This ensures that one or more objective, independent parties have arrived at similar conclusions with the data available. Moreover, any member of the Project community can dispute a contribution as invalid, which then triggers a review by more senior contributors in the system.
Such a system allows for a distributed community of contributors to provide input into the Project, while still maintaining the highest levels of quality. Thus, longer-term the Project will not only learn from traditional allopathic medicine perspectives, but also from the perspective of other important health stakeholders, including scientists, public health professionals, osteopathic, complementary, and alternative healers, as well as patients and families contending day-to-day with the health problems considered.
Additionally, by using its stored data and algorithms, Human Dx can automatically flag potentially false, specious, or incorrect contributions for manual review by a sufficiently senior contributor.
The Project is first focused on diagnosis as the principal step towards identifying, treating, curing, and ultimately preventing disease. However, while the Project's name suggests diagnosis is the sole focus, we are simultaneously tackling the rest of the care continuum.
Traditional diagnostic solutions can be broadly segmented into two categories. Primarily human-based approaches (e.g., static content providers, clinical decision support systems, symptom checkers, and collective intelligence) have not scaled well because of their rigid, inflexible, and archaic architecture. Primarily machine-based approaches (e.g., unstructured machine learning, electronic medical/health record intelligence, health analytics, and patient simulators) have difficulty sifting through and creating valid recommendations from unstructured data.
Furthermore, the shortcomings of existing diagnostic solutions are perpetuated not just by their methodologies, but the data itself. Currently, medical data is not collected in a highly structured format that can be effectively utilized by both humans and machines.
Human Dx focuses on the ideal intersection of human and machine intelligence. The system maximizes the capabilities of both medical practitioners and machines, allowing the two to work effectively in concert.
The data captured by Human Dx addresses four fundamental problems in existing health data sources by:
When members of the global medical community contribute cases to Human Dx, the aggregated & anonymized clinical insights are made available via the website, and can be accessed by searching diagnoses and symptoms on the homepage. Additionally, this insight will later be available directly to stakeholders via API using a Creative Commons License. This gives the Project community the ability to share, use and build upon work that other contributors have created. The specific license is available for review here: CC BY-SA-NC 4.0.
Over time, as more of the Project's data reaches a level of fidelity that is practically useful, we intend to make much of the Project's aggregated case insights, metrics, source code, services, and applications available for academic, research, social, and other noncommercial uses.
Potential partners should contact email@example.com to request further access.
The Project is highly sensitive to patient and contributor privacy. Our focus on this important concern manifests in multiple ways:
Existing open health initiatives primarily seek to collect and aggregate health data from multiple sources. Human Dx's goal is to enable these and other platforms to generate actionable insight that will help them change the cost of, access to, and effectiveness of care globally.
Human Dx is architected to be both human-interpretable and machine-processable. This means that the system will be compatible with existing offline workflows and online systems.
Much of the project's aggregated & anonymized case insights will ultimately become available to any individual or organization for academic, research, social, and other noncommercial uses. Paid commercial usage through software-as-a-service contracts, API calls, and licensing/royalty agreements, will help to support and resource the Project over the long-term. Reimbursement and regulatory requirements will be dealt with by the party that uses Human Dx's data for their own processes, as the Project will solely provide information (and not advice).
Human Dx draws inspiration from many different sectors and technologies, including human-based computation, active machine learning, collective intelligence, and games with a purpose. For current applications of related approaches, please reference reCAPTCHA and Foldit.
Though the Project's code base is not currently open source, we intend to make many parts of it open source in the future. If you are interested in contributing to several major open source projects we are thinking about, please email us at firstname.lastname@example.org.
Human Dx is not currently a free software project. We are closely aligned with the intentions of free software, but also want to ensure we can make the Project both highly useful for the world while also self-sustainable financially. We are not ruling out that this will become possible with time.
Please email us at email@example.com if you would like to be notified when Human Dx APIs become publicly available.
The Project is developed by the global medical, technology, research, and patient communities as well as the Human Dx Team (including its full-time staff, partners, supporters, and board of directors).
Human Dx (the Institution) is structured as a partnership between the social, public, and private sectors, and will ultimately interface with a wide variety of stakeholders. These include: nonprofits, foundations, governments, multilateral organizations, nongovernmental organizations, individual benefactors, private companies, and the Project's relevant communities (i.e. medical, technology, research, and patient).
The Human Diagnosis Project, Ltd. (the Project) and Human Dx, Ltd. (the Organization) are entities that interface with different stakeholders for different purposes. The Project is a nonprofit that interacts with academic, research, social, public and other noncommercial stakeholders. The Organization, its commercial counterpart, is structured as a Public Benefit Corporation and regulates all commercial uses of the Project's intellectual property (similar to an academic research institution's technology licensing arm). It may also build its own products, applications, and services with the Project's data as well as enable commercial implementations of Human Dx.
The two entities are organized in a hybrid structure. A few representative models of such structures include: Mozilla Foundation and Mozilla Corporation; the J. Craig Venter Institute and Synthetic Genomics; Wordpress and Automattic. Other open software projects and the companies that support them also have interesting analogues to the Project. Examples include: Linux and Red Hat; MongoDB and MongoDB, Inc.; Git and GitHub.
Human Dx aims to ultimately be self-sustaining, so that the Project can support and resource its Mission and make itself widely available and accessible for academic, research, social and other noncommercial uses. Over time, the Project will be funded by products, services, applications, implementations, royalty, and licensing agreements related to commercial uses of the Project.
Human Dx is currently supported by some of the world's top thinkers in medicine, computer science, data, design, business, and social impact. A sample of the Project's stakeholders and supporters are below:
Additionally, the Project will also likely be funded by philanthropic sources (including foundations, governments, multilateral organizations, research institutes, and individual benefactors) who share an interest in the Project's Mission.
If you are interested in learning more about how you or your organization can support the Project, email the Team at firstname.lastname@example.org.
The Project is uniting talented individuals to tackle one of humankind's most challenging problems. The Team includes diverse backgrounds at the intersection of medicine, global health, computer science, information science, mathematics, software engineering, design, business, and public policy. The Team has included alumni from leading organizations (including the World Health Organization, NASA Jet Propulsion Laboratory, Lawrence Livermore Labs, Lincoln Laboratory, Facebook, Amazon, McKinsey & Company, Goldman Sachs, the Blackstone Group, and IDEO) and academic institutions (including Oxford, Cambridge, Harvard, MIT, Stanford, Berkeley, Yale, Penn, Dartmouth, and Cornell).
Applicants must be passionate about the Project's Purpose of elevating wellbeing for all and its Mission to empower anyone with the world's collective medical insight.
Please visit the Human Dx Team page to learn more about joining us.
Future updates, articles, and news will be available on an upcoming Human Dx blog. Please sign up on our home page (www.humandx.org) to receive occasional emails from us.
The Project is global in nature, but currently has permanent presences in the Bay Area and Northeast regions of the United States. Over time, the Project intends to have regional presences throughout the world.
The named Human Dx has multiple meanings. The first is literal: "Dx" is the medical shorthand notation for diagnosis, thus translating to Human Diagnosis.
The second is less apparent, and is related to mathematics. A definite integral can be used to express the summated area formed by a function and an axis over an interval. The general notation for an integral is ∫f(x)dx. In this case, the "function" being integrated, or integrand, is "human." Thus, it metaphorically describes the summated potential of many humans working together to tackle a complex problem.
The third meaning is a connection between mathematics and medicine. Dx in integration is literally translated as the "differential of x." Differentials are the list of possible diagnoses that medical practitioners, residents, and students apply when thinking about what a given patient (x) may have.
Please feel free to reach us through our contact page with any general inquiries. Otherwise, use the specific email addresses referenced elsewhere.
Dr. Shantanu Nundy is the Director of the Human Diagnosis Project nonprofit. He also practices primary care at Mary's Center, a federally qualified health center for low-income and uninsured individuals in Washington, D.C., and is a lecturer in health policy.
Dr. Nundy's career spans medicine, global health, academia and technology. Previously, he founded a medical clinic in rural India and served as a consultant to the World Health Organization (WHO) on patient safety, the Results for Development (R4D) Institute on noncommunicable diseases, and SNEHA, a women's and children's health non-governmental organization in Mumbai, India. He has extensive field experience working with the United Nations Development Programme (UNDP) in Ruhiira, Uganda; the Centers for Disease Control and Prevention (CDC) in Lima, Peru; and multiple non-governmental organizations across India. Dr. Nundy has conducted research at Harvard Medical School, Johns Hopkins and the University of Chicago and has published research on connected health, health informatics, and health disparities in Health Affairs, JAMA Internal Medicine, and the Journal of Medical Internet Research. Prior to Human Dx, Dr. Nundy cofounded SMS-DMCare, a mobile messaging system for low-income patients with chronic disease, which was adopted by the WHO eHealth Compendium in 2013, and led clinical innovation at Evolent Health, a population health management organization.
Dr. Nundy received his Bachelor of Science from the Massachusetts Institute of Technology, his Doctor of Medicine with Alpha Omega Alpha distinction from The Johns Hopkins University School of Medicine, and a Master of Business Administration from the University of Chicago Booth School of Business. He trained in internal medicine at the University of Chicago Medical Center, where he also completed an Agency for Healthcare Research and Quality (AHRQ) research fellowship in health disparities and served as a principal investigator. His book on preventive medicine, "Stay Healthy at Every Age: What Your Doctor Wants You to Know," was chosen as one of the top five health books of 2010 by Wall Street Journal Online.
Dr. Nundy and other relevant members of our staff can be reached via our contact page.