Osteo-assist: Detecting Osteoporosis through CV

In India, approximately 61 million people are diagnosed with osteoporosis yet it remains a vastly underrecognized and undertreated health concern, particularly prevalent among post-menopausal women, individuals with prolonged steroid use, and the elderly population. Despite its significant impact on quality of life, the condition is often overlooked due to various systemic challenges. Access to crucial diagnostic tools like DEXA scans is severely limited, with only 0.26 machines per million people in the country due to its cost. Given that, almost 2/3 of India’s population lives in rural areas with limited access to good facilities these DEXA Scans pose a great barrier to proper healthcare.  Moreover, healthcare providers exhibit a concerning lack of awareness and knowledge about osteoporosis, as evidenced by a study revealing an average score of 22.5% among doctors in the rural areas of South India. This knowledge gap leads to delayed diagnosis and inadequate treatment, exacerbating the already substantial burden of osteoporosis-related fractures and disabilities. Given these circumstances, one can easily consider the number “61 million” to be much higher.

Osteo-assist has created a model that automates the process of analyzing the radiographs and classifies them into different grades of osteoporosis using AlexNet (CNN: Alex. K) for image classification. The model is based on a pre-established and tested grading system, the Singhs index (a grading system that provides a way to detect osteoporosis by analyzing simple X-ray scans).  There are 6 Grades of the Singhs Index and using AlexNet a well-known CNN for image classification is fed in with the “right answers” (Supervised Machine Learning) which helps it derive the Index of new images. The functionality of the Osteo-assist can be described in simple terms: when a doctor uploads a radiograph into the program, the Osteo-assist utilizes image processing algorithms to extract key features from the bone structure captured in the image. These features are then analyzed and interpreted using machine learning algorithms trained on a vast dataset of radiographs annotated with corresponding osteoporosis grades. Therefore, Oseto-assist can accurately classify the severity of osteoporosis in the given radiograph, providing healthcare professionals with valuable diagnostic information in a matter of seconds.

Osteoporosis depends on various factors such as sex, age, body size, race, family history, change of hormones, diet, and medications. Data shows that a minimum of 46 million women are affected by postmenopausal osteoporosis which is a type of primary osteoporosis(one of the four major types). Including the women who have had surgery to remove their uterus, this number may even be over doubled. Moreover, estimates from 2015(the most recent data set) show that 20% of the 230 million Indian women over age 50 have osteoporosis. Furthermore, a study from 2001 stated that the incidence of hip fractures in the Indian population was 361 women and 128 men per 100,000 people. Extrapolating these results we’ll see that the total number of hip fractures in India is approximately 440,000 with a female-to-male ratio of 3:1. Additionally a 2019 study also showed that India was the highest contributor to osteoporosis fracture-related deaths worldwide. Osteoporosis can be detected using DEXA or the bone mineral density scan which unfortunately is not accessible to the majority of the Indian population as there are only 0.26 DEXA machines per million of the population in India. Most of these DEXA machines are available only in urban areas and are way too expensive for the general population leading to multiple cases of undiagnosed osteoporosis. The fact that Indian doctors lack the knowledge and are ill-equipped to handle osteoporosis further adds to the severity of the situation.  Our solution, osteo-assist, is going to help doctors identify osteoporosis at early stages and prevent it from ever occurring. Our solution will be implemented in rural areas of India(which is where two-thirds of the population lives) and will offer a diagnosis for osteoporosis at a low cost or even for free in most cases.  

Our team is well-positioned to deliver the Oseto-assist solution to the target population due to our deep understanding and proximity to the healthcare community, especially those affected by osteoporosis. Our Team Lead has direct experience working in healthcare settings with orthopedics, providing insights into the challenges faced by healthcare professionals and patients too.  Additionally, several team members have personal and familial experiences with osteoporosis, providing a personal connection to the target population. Furthermore, throughout the design and implementation process, we actively seek input and feedback from healthcare professionals, patients, and advocacy groups representing diverse perspectives and experiences related to osteoporosis and the prototype. By incorporating community input and ideas into our solution, we ensure that Oseto-assist is tailored to the needs and preferences of those it serves, ultimately enhancing its relevance, effectiveness, and impact within the community.

As of now, our team has developed and tested an initial prototype of the Oseto-assist solution. The prototype represents a functional version of the software, capable of analyzing X-ray images and providing preliminary classifications of osteoporosis severity based on established grading systems. We have conducted internal testing on a large data set (approx 12000 images) and validation to ensure the accuracy and reliability of the algorithm's performance. While we have not yet served any external customers or direct beneficiaries our product has been tested by 2 orthopedics in a clinical setting, we are in the process of seeking partnerships with healthcare facilities and professionals to conduct pilot studies and have gathered feedback on the usability and effectiveness of Oseto-assist in real-world clinical settings. Our focus at this stage is on refining the prototype based on user feedback and preparing for wider deployment and adoption shortly.

We have applied to Solve because we believe that Osteo-assist has the potential to make an impact on the challenge of osteoporosis detection and management. While our team has made progress in developing the prototype of Osteo-assist, there are several barriers that we may face in scaling and implementing the solution effectively, including financial constraints related to further development, testing, and deployment of Osteo-assist at a larger scale, as well as technical challenges associated with integrating the solution into existing healthcare systems and hospitals. We would appreciate mentorship in fine-tuning our model to an increased level of accuracy. We also acknowledge the importance of community engagement in ensuring the success and acceptance of the solution. We hope to leverage the platform's network of partners and resources to overcome these barriers and accelerate the impact of Osteo-assist. Specifically, we are seeking support in the form of technical expertise, funding opportunities (for team expansion and further research and development), and strategic partnerships with healthcare institutions.

Our solution offers doctors an easier method to identify osteoporosis at early stages and prevent it from worsening. Osteo-assist makes use of AI and machine learning to easily identify osteoporosis at all stages and does not require a DEXA machine or a bone mineral density scan. This will be beneficial in rural areas where over 60% of the population resides. It will be easier to identify the type of osteoporosis and will also give the treatment accordingly in case the doctor does not know the treatment(there is a lack of knowledge regarding osteoporosis in India where when doctors were given a test on the topic, the mean score was 22.5% and all doctors got below a 50%). This is also a cheaper and better alternative to DEXA and bone mineral density scans as Osteo-assist identifies osteoporosis just from the X-rays.

Our theory of change for Osteo-assist is pretty straightforward,  by providing healthcare professionals with a more efficient and accurate tool for osteoporosis detection, we expect to achieve both short-term and long-term outcomes. The immediate consequences of Osteo-assist include the development of a user-friendly software solution capable of analyzing X-ray images and classifying osteoporosis. By training algorithms on a large dataset of radiographs and implementing the software in healthcare settings, we have and aim to provide doctors with timely and reliable diagnostic insights. The immediate outcome we expect is improved efficiency in osteoporosis detection, allowing healthcare professionals to quickly look at the risk and initiate appropriate treatment protocol. In the longer term, we expect several outcomes: firstly, enhanced accuracy in osteoporosis diagnosis overall, leading to reduced rates of misdiagnosis and increased rates of early detection, resulting in fewer fractures and other associated complications and ultimately, improved quality of life for patients living with osteoporosis. Our theory of change is supported by research demonstrating the effectiveness of the Singh index in osteoporosis detection, as well as feedback from healthcare professionals highlighting the need for efficient and reliable diagnostic tools like Osteo-Assist.

Our impact goals revolve around addressing the challenges of osteoporosis detection and management. Here are some key goals and how we have planned to measure progress towards them:

1. Our primary goal is to facilitate early detection of osteoporosis. We plan to measure progress by tracking the number of osteoporosis cases identified through Osteo-assist compared to traditional diagnostic methods. 

2. We aim to reduce the incidence of fractures among individuals diagnosed with osteoporosis. Progress can be easily measured by monitoring the number of fractures reported in patients diagnosed with osteoporosis using our solution, compared to those diagnosed through traditional methods.

3. Our ultimate goal is to improve patient outcomes and quality of life for individuals living with osteoporosis. Progress can be measured using standardized quality-of-life assessment tools for example the EQ-5D or SF-36.

So, by tracking these indicators, we can assess the effectiveness of Osteo-assist. Additionally, our impact goals align with relevant indicators associated with the UN Sustainable Development Goals, particularly Goal 3: Good Health and Well-being, ensuring that our efforts contribute to broader global health objectives.

The core tech that drives our solution is Artificial Intelligence , in particular Computer Vision and Deep Learning. We use AlexNet CNN which detects osteoporosis by analysing X-ray images and classifying them into their particular Singh Index grades(6 grade levels used to diagnose osteoporosis) accurately. This enables early intervention and improved healthcare outcomes, benefiting individuals globally with accessible osteoporosis screening.

Full-time staff - 3 

Part-time staff - 9

1 year

Our team includes male and female students from different religious and geographical backgrounds. We make sure to collaborate and celebrate one's background to the fullest at the same time opening avenues to innovative solutions.

OsteoAssist operates on a sustainable business model driven by impact and accessibility. We raise funds from awards, competitions, angel investors, and crowdfunding to provide our services free of cost to Hospitals and clinics so that they benefit from our advanced osteoporosis detection and management solutions without financial burden. 

Our donors ask us to report the impact and provide funds when required based on our growth and reach. We prioritize B2B partnerships with healthcare institutions, avoiding contract-based work with development partners to concentrate on expanding our impact through strategic collaborations and in-industry healthcare adoption.

1.⁠ ⁠Strategic Pharma Collaborations: Align with pharmaceutical firms and healthcare entities, leveraging synergies for innovation and market expansion. Pursue revenue-sharing agreements, statistically proven to enhance financial sustainability.

2.⁠ ⁠Individual Donor Cultivation with Recurring Micro-Donations: Employ donor segmentation strategies to identify individuals aligned with the organization's mission. Utilize CRM systems to nurture relationships and encourage recurring micro-donations, leveraging techniques such as peer-to-peer fundraising and personalized messaging for sustained support.

3.⁠ ⁠Strategic Grant Acquisition with Proposal Development and Compliance Management: Allocate resources to hire a dedicated grant writer proficient in RFP analysis and proposal development. Implement robust grant management software for tracking and compliance, ensuring efficient utilization of funds and accurate reporting to grantors.

4.⁠ ⁠Diverse Fundraising Strategies with Omnichannel Engagement and ROI Analysis: Implement an omnichannel fundraising approach, integrating online donation platforms, social media campaigns, and traditional fundraising events. Utilize data analytics tools for ROI analysis, optimizing fundraising efforts by identifying high-yield channels and adjusting strategies accordingly for maximum financial impact and organizational sustainability.

We are currently building relationships with individual donors as we didn't have a need for funding for the initial R&D. Our Initial Funding was from SK.Jajodia of Warden Group for getting access to certain API Keys. That was 5000 Rupees. We are currently applying for various grants and contacting Pharmas for collaboration. We are in touch with Eternal Hospital Sanaganer (EHCC chain, the biggest hospital chain in Rajasthan) for further support in the outreach program.