Early AI Detection of Pancreatic Cancer

We're tackling a longstanding challenge in healthcare: the late detection of pancreatic cancer. Pancreatic ductal adenocarcinoma (PDAC) is one of the deadliest cancers that exist. It is notorious for manifesting and being diagnosed in its advanced stages, drastically reducing the chances of survival for those affected. Globally, pancreatic cancer ranks as the 12th most common cancer with over 460,000 new cases each year. Alarmingly, it's on track to become the second leading cause of cancer-related deaths by 2030. The statistics paint a difficult picture, especially in the United States, where approximately 57,600 people will be diagnosed this year alone, and the five-year survival rate is only 9%.

The central hurdle in battling pancreatic cancer is its silent progression and lack of early symptoms. Symptoms are often vague and easily dismissed until the cancer has reached a critical, less treatable stage. At that point, the cancer has typically already metastasized. This pattern of late detection is primarily due to two things: (1) the absence of efficient, non-invasive screening methods, and (2) a reliance on recognizing symptoms that often appear too late.

OncoSight, our AI-driven solution, is designed to address this issue in pancreatic cancer care by effectively screening patients at scale. OncoSight leverages artificial intelligence to sift through electronic health records (EHRs), identifying subtle patterns that indicate the early stages of pancreatic cancer. Our innovative approach has demonstrated the potential to detect PDAC much earlier than current protocols allow. By creating autonomous systems that warn of potential cancer early on, the likelihood of successful treatment and, ultimately, survival significantly improves.

The universal challenge of early detection in pancreatic cancer, combined with its global prevalence and the ubiquity of electronic health records, underscores the profound impact OncoSight could have. OncoSight is deliberately designed to integrate seamlessly with existing healthcare IT frameworks and detect PDAC using only data that all patients will have on record. This means an accessible and scalable advance in cancer care that includes all patients, free of friction points or specialized tests. This paradigm of screening promises to enhance survival rates and provide patients with a fighting chance against a disease that has long been considered a death sentence.

In essence, our Ai confronts a dual challenge: the longstanding need for early pancreatic cancer detection and the current inadequacy of available diagnostic timelines. By harnessing AI to analyze routine health records, we're offering an inclusive and effortless path to early intervention, fundamentally transforming the fight against pancreatic cancer on a global scale.

Our platform, OncoSight, directly serves two main groups within the healthcare ecosystem: (1) patients at risk of pancreatic cancer, and (2) the healthcare professionals who care for them, including pancreatic oncologists and primary care physicians. These groups are currently underserved in the realm of early cancer detection due to the limitations of existing diagnostic methods.

Patients at Risk of Pancreatic Cancer:
This includes individuals with a family history of the disease, those with certain genetic mutations, and others in high-risk categories based on lifestyle and health factors. Currently, these individuals face a dire lack of non-invasive, efficient screening methods for early detection of pancreatic cancer. The late diagnosis of this disease often results in a prognosis that is poor and treatment options that are limited and less effective. Our AI-driven diagnostic tool aims to change this by providing a method for early detection, significantly improving the chances of successful treatment and survival.

Healthcare Professionals & Healthcare Systems:
Pancreatic oncologists and primary care physicians are on the frontline of diagnosing and treating pancreatic cancer. However, they are often handicapped by the current diagnostic tools that rely heavily on symptomatic presentation, which typically occurs in the later stages of the disease. OncoSight will empower these professionals with actionable insights derived from EHR data, enabling them to identify and begin treatment for pancreatic cancer much earlier than has been possible.

Impact for Each Group:

For patients, early detection of pancreatic cancer means a significantly better prognosis, more treatment options (including surgical resection, which is only possible in early stages), and a higher chance of survival. It transforms a diagnosis of pancreatic cancer from a likely death sentence into a manageable condition.

For healthcare professionals, OncoSight offers a powerful tool to enhance their diagnostic capabilities and treatment options. An early diagnosis makes it possible to intervene before PDAC becomes terminal and tailor treatments to the individual patient more effectively. This not only improves patient outcomes but also contributes to the overall efficiency of the healthcare system by potentially reducing the need for more invasive, prolonged treatments that required extended hospital stays. In this way, early detection relieves a significant financial and operational burden on healthcare systems. Late-stage diagnoses lead to expensive treatments requiring significant resources. The average direct costs alone in pancreatic cancer treatment is approximately $62,130 per patient each year. This excludes the cost of surgery, which is often necessary for these patients, and adds $29,000 to $175,000 to the yearly treatment costs. Early diagnosis greatly reduces the burden on providers and care networks alike.

In essence, OncoSight addresses the critical gap in early pancreatic cancer detection, serving those most at risk and the dedicated professionals who care for them. By providing a means to detect the disease at a much earlier stage, OncoSight has the potential to dramatically improve survival rates and quality of life for individuals diagnosed with pancreatic cancer, while also empowering healthcare providers to deliver more effective care.

Probably the most unique aspect of our startup is our founding team: Logan Nye and Kushagra Agarwal.

Logan, as both a physician and AI engineer, has a rare dual expertise in medicine and computer science, bridging the crucial gap between cutting-edge AI technology and its practical healthcare applications. It was Logan's clinical experiences in underserved regions like sub-Saharan Africa and the Himalayas that inspired the idea of using AI to democratize screening procedures and public health recommendations. Kushagra also has significant experience and interest in using AI to break down barriers to care. He comes from a rural part of India and seeks to use his experience in machine learning and healthcare research to help patient populations similar to those he grew up in. Kushagra's rich technical experience complements Logan’s clinical insights, forming a balanced and talented team that’s not just technically adept but deeply passionate about making a difference in cancer care. 

These combined strengths and vision are further bolstered by our proximity to UPMC Hillman Cancer Center. Access to leading cancer specialists grants us ample opportunity to collaborate with pancreatic oncologists and physician-scientists. We have already made use of this advantage, working directly with UPMC clinicians to develop and iterate on our MVP.

We are in the process of obtaining access to UPMC's patient data. Thus far, our AI models have been built using deidentified EHR data for proof-of-concept and design experimentation. We have also initiated research studies with the pancreatic oncologists at UPMC Hillman Cancer Center to validate these prototypes on actual patient records. However, we have not obtained access to actual patient data at UPMC yet (and will not for several weeks), so we most closely match the "Prototype" stage description at this time.

Our goal is to make early detection and mitigation of deadly diseases a standard practice that everyone can access. This is why we have developed our AI using commonplace EHR data for its predictive analytics. We want to develop a new paradigm of AI tools that ensure everyone, everywhere is protected by screening procedures and healthcare intelligence. MIT Solve is the sort of venue that attracts similarly audacious and forward-thinking people. It's a gathering place where we can connect and partner with other healthcare and technology innovators set on doing good and breaking down barriers. Funding can come from a variety of places; it's not why we are interested in MIT Solve. We hope to tap into the brilliant group of driven entrepreneurs and doers that can help us deploy life-saving AI at scale and expand into other dire healthcare challenges.

Our main innovation lies in the use of AI to analyze broadly-available, nonspecific EHR data as a proxy screening procedure for a deadly disease (i.e., pancreatic cancer). This method removes the need for specialized or invasive testing, and fundamentally creates a screening protocol where we have lacked one for decades. Unlike traditional diagnostic tools that rely on physical symptoms, imaging tests, or invasive procedures, our AI utilizes the existing patient data already on file to identify indicators suggestive of early stages of disease. It's a non-invasive, cost-effective, and scalable method of screening, overcoming many of the barriers associated with current detection methods. 

Furthermore, by demonstrating the efficacy of AI in early cancer detection, this hints at a broad shift in how the medical community approaches diagnostics across other diseases, too. This model of leveraging EHR data can be adapted and expanded to other cancers types, chronic conditions, etc. to encourage a more proactive and preventative approach to healthcare. If it works for detecting pancreatic cancer, this approach should also work for detecting various other diseases.

Our logical framework is constructed around the following components:

Activities:

• Development and refinement of the OncoSight AI model using vast datasets from electronic health records (EHR).
• Partnership with healthcare providers for input & integration of OncoSight into their diagnostic processes.
• Continuous engagement with and feedback collection from both healthcare professionals and patients to improve OncoSight.

Outputs:

• A fully functional AI tool capable of autonomously identifying early indicators of pancreatic cancer from EHR data.
• Increased awareness and adoption of OncoSight among healthcare professionals.
• A growing body of evidence, including case studies and clinical trial results we publish with UPMC, demonstrating OncoSight's efficacy in early detection.

Outcomes:

• Short-term: Enhanced capability of healthcare providers to detect pancreatic cancer at earlier stages, leading to an increase in early-stage diagnoses.
• Medium-term: Improved treatment outcomes and survival rates for patients diagnosed with pancreatic cancer, owing to earlier intervention.
• Long-term: A shift in the standard of care for pancreatic cancer towards early detection and preventative strategies, reducing the overall mortality rate associated with the disease.

Evidence Supporting the Theory of Change:

• Third-party research indicates that the key to improving pancreatic cancer outcomes is early detection, which significantly increases the feasibility of surgical interventions and the effectiveness of treatments (Source: Cancer Research UK).
• Findings from our prototype studies show that OncoSight can identify patterns in EHR data that are indicative of early-stage pancreatic cancer, with a high degree of accuracy compared to current diagnostic methods.
• Data from interviews with our target population, including healthcare professionals and patients, underscore the need for proactive detection tool like OncoSight. Feedback from these interviews has highlighted the potential impact of early, inclusive detection on treatment success rates and overall patient survival.

Our impact goals center around transforming the landscape of pancreatic cancer detection and treatment. Ultimately, we hope to significantly reduce mortality rates associated with PDAC and improve the quality of life for those diagnosed. To ensure we are on track to achieving these transformative impacts, we have the following specific, measurable indicators of success:

Impact Goals:

1. Increase Early Detection Rates: Facilitate the detection of pancreatic cancer at early stages (I & II), when treatment is more successful.
2. Improve Patient Survival Rates: Increase the 5-year survival rate for pancreatic cancer patients through earlier diagnosis and intervention.
3. Facilitate Adoption of AI in Healthcare: Drive the integration of AI-driven diagnostic tools in healthcare settings to improve disease detection and patient outcomes broadly.

Measuring Progress:

To measure our progress towards our goals, we will track the following key indicators:

1. Number of Early-Stage Diagnoses: We will monitor the change in number of patients diagnosed with stage I or II pancreatic cancer as a direct result of OncoSight's effectiveness in early detection. This will be measured through data collected from early healthcare partners deploying our AI, like UPMC. 

2. 5-Year Survival Rates: By comparing the 5-year survival rates of patients diagnosed through OncoSight to historical averages, we can gauge the impact of early detection on patient outcomes. This data will be collected from long-term follow-ups with diagnosed patients.

3. Adoption Rate by Healthcare Providers: We will track the number of healthcare systems and clinics adopting OncoSight as an indicator of both the solution's acceptance in the medical community and its potential impact on a broader scale. Here, adoption rates will be measured through sales and partnership agreements.

4. Patient and Healthcare Professional Feedback: Collecting and analyzing feedback from both patients and healthcare professionals using OncoSight provides qualitative data on its usability, efficacy, and impact on the diagnostic process. This feedback is gathered through surveys and interviews.

5. Reduction in Late-Stage Diagnoses: Related to #1, a decrease in the proportion of late-stage (stage III and IV) diagnoses among all pancreatic cancer cases diagnosed in facilities using OncoSight will indicate the tool's effectiveness in moving the detection window earlier.

These indicators not only provide a direct measure of our progress towards our impact goals but also align with several UN Sustainable Development Goals, particularly Goal 3: Good Health and Well-being. By tracking these specific indicators, we can quantitatively and qualitatively assess the transformative impact of OncoSight on the lives of those at risk for pancreatic cancer, ensuring that all aspects of our work are oriented towards our overarching mission of saving lives through innovation in healthcare.

Key Components of Our Technology:

1. Data Integration: OncoSight integrates with existing EHR systems, aggregating patient data across multiple sources, including clinical notes, lab results, imaging studies, and medication histories. This comprehensive dataset provides a rich foundation for our AI models to learn from.

2. Machine Learning Algorithms: At the heart of OncoSight are machine learning algorithms that have been trained on large datasets to recognize the early signs of pancreatic cancer. These algorithms analyze the aggregated EHR data, looking for specific patterns such as changes in blood markers, subtle symptoms noted in clinical documentation, or specific combinations of risk factors and patient history.

3. Predictive Analytics: The AI uses predictive analytics to assess the risk of pancreatic cancer, giving healthcare providers a probability score based on the patterns found in the data. This score helps prioritize patients for further diagnostic testing, ensuring early intervention for those at highest risk.

4. Continuous Learning: OncoSight's AI models are designed to continuously learn and improve over time. As more data becomes available from diagnosed cases, the AI refines its algorithms, increasing its accuracy and effectiveness in detecting early-stage pancreatic cancer.

Innovative Aspects and Benefits:

• Non-Invasive: Unlike traditional diagnostic methods that may require invasive procedures, OncoSight utilizes existing EHR data, making the screening process non-invasive and more patient-friendly.
• Scalable: The use of AI to analyze EHR data enables OncoSight to be scalable across healthcare systems, potentially benefiting a vast number of patients without the need for significant additional resources.
• Early Detection: By identifying pancreatic cancer at an earlier stage, OncoSight can significantly improve treatment outcomes and survival rates, addressing one of the most challenging aspects of managing this disease.
• Inclusive: The AI operates using only data points that virtually all patients have in their healthcare record. This allows it to digitally screen and protect nearly all patients in a given healthcare network simultaneously.

Our solution team is currently comprised of our two founders, Logan Nye and Kushagra Agarwal, who are deeply involved in all aspects of the project, from development to strategic planning. We are full-time graduate students at Carnegie Mellon's School of Computer Science, and work on this solution ~20 hours per week. Additionally, we collaborate closely with clinical partners at UPMC, working with them to refine and validate our AI-driven diagnostic tool.

We are a relatively new AI startup. We formed in October 2023 and have been working on this solution since then.

Diversity in Leadership & Backgrounds:
Our team comes from diverse academic and cultural backgrounds, bringing a broad spectrum of perspectives to our mission. We actively seek to expand this diversity as our team grows, ensuring representation across various dimensions including gender, ethnicity, age, sexual orientation, disability, and socioeconomic status.

Our Plans for Diversity, Equity, and Inclusion (DEI):

• Recruitment and Hiring: We plan to implement recruitment practices that reach underrepresented groups in tech and healthcare, using platforms and partnerships that cater to these communities. Our goal is to create a team that reflects the global diversity of the patients and healthcare professionals we serve.
• Professional Development and Advancement: Recognizing that equity is about access to opportunities, we plan to implement programs for mentorship, continuous learning, and career advancement tailored to the needs and aspirations of all team members, particularly those from marginalized groups.
• Inclusive Culture: We are committed to building an inclusive workplace where every team member feels valued, heard, and empowered to contribute. This includes regular DEI training, open forums for feedback, and policies that respect and accommodate the diverse needs of our staff.

We plan to create revenue by offering our AI solution to healthcare systems, specialized cancer centers, and research institutions.

Key Customers and Beneficiaries:

• Healthcare Providers (Hospitals, Specialized Cancer Centers, Primary Care Facilities): Our primary customers, who seek effective tools for early cancer detection to improve patient care and outcomes.
• Patients at Risk for Pancreatic Cancer: The ultimate beneficiaries of our solution, who receive a significant value in the form of potentially life-saving early diagnosis.

Products or Services Provided:

• OncoSight AI Diagnostic Tool: An enterprise-level software-as-a-service (SaaS) platform that integrates with existing Electronic Health Records (EHR) systems to analyze patient data for early indicators of pancreatic cancer.

How We Provide These Products or Services:

• Integration with EHR Systems: We work closely with healthcare providers to integrate OncoSight into their existing EHR systems, ensuring a seamless workflow and minimal disruption.
• Continuous Support and Training: We provide ongoing support and training for healthcare providers to maximize the effectiveness and utilization of OncoSight.

Why Customers Want or Need Our Solution:

• Healthcare Providers: They are in need of more effective tools for early cancer detection to enhance patient care, improve survival rates, and reduce the costs associated with late-stage cancer treatments. OncoSight offers a non-invasive, cost-effective, and scalable solution that fits these needs.
• Patients: Early detection of pancreatic cancer significantly improves treatment options and survival rates. Our solution directly addresses the lack of effective early detection methods for pancreatic cancer, offering hope to those at risk.

Revenue Generation:

• Licensing Model: Healthcare providers will pay a subscription fee to integrate the OncoSight platform, structured according to the size of the institution and the volume of patients served.
• Customization and Consulting Services: We will offer additional services, including customization of the AI model to suit specific institutional needs and consulting services for data analysis and integration, as an additional revenue stream.

Impact Generation:

• Improved Patient Outcomes: By enabling earlier detection of pancreatic cancer, we hope and expect that OncoSight will significantly improve treatment success rates and survival outcomes for patients.
• Healthcare Efficiency: Our solution will help healthcare providers and institutions optimize their diagnostic processes and resource expenditure, leading to more efficient use of resources and potentially reducing the overall burden on the healthcare system.

This will mainly be revenue from selling our AI diagnostic platform, securing grants for research and development, and exploring strategic partnerships and investment opportunities. This blend ensures enough financial foundation to support ongoing R&D and our mission to transform pancreatic cancer detection and care.

Revenue Streams:

1. Subscription Model for Healthcare Providers: Charging healthcare facilities a subscription fee for using OncoSight. This fee will be scaled based on the size of the institution and the volume of patients, making our solution accessible to a wide range of healthcare providers.

2. Customization and Consulting Services: Offering additional services, such as customizing the AI model for specific institutional needs and consulting on data integration and analysis, providing another revenue stream.

3. Grants and Research Funding: Pursuing grants and research funding opportunities, particularly those focused on cancer research and healthcare innovation, to support our development and validation efforts.

4. Strategic Partnerships and Investment: Seeking and forming strategic partnerships with healthcare technology companies and seeking investment from venture capital firms that specialize in healthcare innovation, to support scaling and further development.