Multi-Disease Screening Tool

Our solution is positioned to overcome the problem of limited access to sufficiently sensitive and specific disease screening. Initially, we are focused on alleviating these problems for the TB epidemic. From which an estimated 10 million people fall ill from TB annually, with roughly 1.4 million dying from the disease. The current diagnostic efforts miss an estimated 3 million patients annually.

Our solution is a consumable free screening tool that is designed for primary health care facilities and active screening efforts. The screening tool uses AI to analyse clinical sensor inputs (first of which is a digital stethoscope) to make disease status predictions, outputting a suggested diagnostic response. 

Our estimated increase in screening sensitivity will roughly half the number of TB patients missed by the current diagnostic algorithm. Thus improving treatment outcomes and reducing transmission. The increased screening specificity will save states 1/3 of their TB diagnosis budgets.

Our solution is positioned to overcome the problem of limited access to sufficiently sensitive and specific disease screening. Initially, we are focused on alleviating these problems for the TB epidemic. 

An estimated 10 million people fall ill from TB annually, with roughly 1.4 million dying from the disease. The current diagnostic efforts miss an estimated 3 million patients annually.

South Africa (Third-World countries alike) have limited access to lung X-ray screening facilities and personal and are therefore unable to fulfil this WHO’s TB screening recommendations. Unfortunately, this is especially true for the rural low-income regions where TB is most prevalent. Instead, a symptom-based questionnaire (77% sensitive and 65% specific) is used to screen for TB.

In South Africa, 360000 people fall ill from TB annually, of which an estimated 24% are left undiagnosed, predominantly due to poor screening sensitivity. This significantly contributes to the 60000 annual TB deaths.

South Africa spends an estimated $50,84 million USD on TB diagnosis. The insufficient screening specificity results in superfluous secondary diagnostic testing on TB-Negative patients, incurring unnecessary costs and capacity congesting. 

Our solution is a consumable free screening tool that is designed for primary health care facilities and active screening efforts. The screening tool uses AI to analyse clinical sensor inputs to make disease status predictions, outputting a suggested diagnostic response.

• Local Processing Unit and GUI, currently powered by the Jetson Nano.
• Clinical Sensors, currently utilising a digital stethoscope for TB screening.
• Cloud-Based Registry/database, currently utilising AWS.
• Predictive Algorithms, currently primarily using ResNet34 (Convolutional Neural Network) and gradient boosting techniques.

Guided by the touchscreen GUI on the Local Processing Unit, the health worker acquires the patient’s input dataset. (Personal details, chest auscultation recordings and some relevant clinical questions and vital measurements). This input data is fed into the latest Predictive Algorithm (this is run locally to avoid intermittent connectivity barriers). The predictive algorithm result is compared to a threshold (potentially tailored to the needs of the region) deeming the patient either screened Positive or Negative dictating whether the patient is sent for further TB diagnostic testing. When an internet connection is established this is synced to our cloud-based database.

There are three categories in which our solution is designed to serve.

• Beneficiary: TB sufferers and low-income and underserviced communities.
• Customer: National Departments of Health.
• User: Nurses and public health workers.

We will provide access to sufficiently sensitive screening, in the regions where it is most needed. These patients are typically significantly affected by the poverty pandemic, which means they rely on the state for medical care and live in underserviced, malnutrition, congested environments, primed for the spread of TB and future pandemics. These patients typically only have access to facilities that are void of x-ray units and/or those trained to take x-rays let alone interpret them. Chest x-rays, when used properly, are sufficiently sensitive for TB screening, yet due to access, these underserviced communities are instead screened for TB using a symptom-based questionnaire. From the get-go this symptom-based questionnaire misses 23% of self-presenting TB positive patients. Ultimately improved screening will result in earlier TB diagnosis for infected patients and thus earlier treatment. This will improve treatment outcomes while protecting the community by reducing disease transmission. We’ve followed both TB advocacy and the international organisation’s prescribed needs, however, more actual patient engagement could be incorporated. 

An increase in screening specificity will serve the National Departments of Health by reducing the percentage of TB-Negative patients being sent for secondary TB diagnosis testing. Preserving the state TB budget will in turn also benefit TB sufferers as funds can be redirected to improving TB treatment etc. We have maintained an ongoing conversation with the state to ensure that we aline with their needs. 

The screening tool has been designed to empower the user by extending their diagnostic input while ensuring the tool remains fast and easy to use. Nurse input played an important role in the development of our product User Needs.

When future pandemics strike, these well-distributed screening tools (primarily positioned in the most valuable communities) and national health laboratory collaborations will allow for the development of new disease training databases and subsequent remote model updates allowing for screening of new disease. Imagine if we had deployed prior to that Covid-19 pandemic, what disease insights would we have been able to share from our training dataset and how would the pandemic look today with earlier detection and more efficient use of secondary testing resources.

An effective response to a new pandemic or existing epidemics (TB) must ensure infected patients are identified early, to improve treatment outcomes and reduce transmission. Accessible, sensitive and specific disease screening plays an essential part in identifying infected patients early. We are equipping the last-mile primary healthcare with a tool to respond in such a manner.

Our product is specifically designed for low-income countries, where specialised diagnostic capabilities are scarce. The consumable free nature (extremely low cost), offline functionality and portable nature of our product are intentionally designed for poverty-stricken communities who typically make use of underserviced clinical environments.

Our Local Processing Unit prototype is complete as can be seen in figure 3.  The developed digital stethoscope frequency response has been tested - superior to Littmann 2300.  The developed predictive algorithm architecture has been trained and tested on an open-source database for an analogous application (abnormal lung sounds detection). Our developed database pipeline is being piloted during our data collection study which makes use of our in-house digital data collection and consent system. 

We are in the process of populating our TB training database from 3200 participants in the high TB prevalent areas near Cape Town. 

The concept of detecting TB and other diseases like COVID-19 using auscultation and cough sounds has previously been proven. 

We have developed our business plan which indicates a convincing win-win situation where our customer (The DoH) saves money while the beneficiaries(TB sufferers) are detected and treated earlier. 

Instead of using visual, or biomarkers, our system uniquely uses audio markers or signatures to detect TB. TB infected lungs alter the structure and content within the lungs. In a first-world context, these structural changes would typically be seen on a screening chest x-ray. Instead, our screening tool is used to detect TB by identifying these same structural changes by their specific lung sounds.

The supporting operational model contributes to the solution's uniqueness. Digital diagnostic stethoscopes have been developed for private use or for specialist augmentation. However, the truly vulnerable population are unlikely to warrant the purchase of such devices. Our semi-decentralised operational model, which uses public clinics provides an accessible, free and accurate screening alternative. 

In terms of future pandemic readiness, our unique digital data collection and consent system can be deployed on the widespread TB screening tool itself. Our national laboratory service collaboration enables us to validate each new dataset is validated according to the PCR test. Therefore collecting data to populate the new disease AI training databases becomes a quick seamless clinic-based exercise. From the new disease training database, we create the new disease-specific screening AI model and remotely upload it to our widespread screening tools.

Currently: 0

1 year: 0

3 years: South African public clinic using population (48 million patients). Each patient, by the RSA standard of care, is to be screened for TB upon clinic visitation. Roughly 1000 000 TB patients are missed each year.

5 Years: We would like to service the public health sectors of India, China, Indonesia, Nigeria Pakistan, the Philippines and Bangladesh.

Maximising the capture and efficiency ratio will be the main metric indicators of success. (linked to detection sensitivity and specificity)

Capture Ratio: Number of bacteriological confirmation test conducted on TB-Positive patients vs the WHO's national TB incidence rate. 

Efficiency Ratio: Number of bacteriological confirmation test conducted on TB-Positive patients vs Number of bacteriological confirmation test conducted on TB-Negative patients.

Fulltime - Braden van Breda, CEO, PI

Fulltime - Johan Coetzee, CTO, Investigator

Fulltime - Mark van Breda, COO

Fulltime - Avuyisiwe Godlani, Study Nurse

Fulltime - Arnold Godlani, Security Gaurd

Contractors - Malcolm Applewhite, Regulatory Consultant, regZAmed

Contractors - Caryn Upton, Medical Doctor, Scientific Officer, Task

The AI Diagnosisis based team has 15 years of medical device development experience.

 I completed a BSc in Mechatronic and MSc in Mech. Eng. After which I developed novel heart valve replacements and delivery systems that were designed primarily to serve patients suffering from rheumatic heart disease. 

Johan completed his BTech Chem Eng. and MSc in Biomaterials. After working in the medical product development field he switched careers and trained as a data scientist, specialising in AI. He has previously worked on multiple audio interpretation projects. 

Mark completed his BSc in Electrical Eng. and has 33 years of experience launching and managing business ventures, the last of which is a care facility for Alzheimer’s patients. 

We are aware of our shortcomings as a team and will continue to seek consultation in the areas we lack expertise. Most importantly the team shares a desire to change the circumstances of those in need.

Each team member grew up or lives in areas where the TB prevalence is of the highest in the world. We have all witnessed the perpetuating devastation of TB in the communities around us. Although this experience doesn't necessarily provide design input it certainly provides an abundance of motivation.

Admittedly we do not currently have an officially documented approach towards this goal, so instead I will present our current leaderships intentions.

Firstly, our business outputs are centred around supporting those most in need, which is made evident by our slogan, Equity in Medical Care. This is a shared ethos amongst the current leadership team. Therefore, it is as important to us that the internal business leadership, hiring policies and appointments mirror that ethos. In attempt to make South Africa a more equitable landscape, we will stive to appoint, hire and partner with those who are specifically previously disadvantaged, this way those who are truly in need are supported.

From a diversity perspective, we fully recognise the importance of diversity in leadership teams as it provides multiple perspectives and avoids blind spots. How exactly we will implement this is yet to be determined.

None of the leadership member are currently paid. However, I believe making all salaries public would keep us all accountable.

We are primarily applying to Solve for financial support. 

However, we believe this platform may open up the opportunity to partner with country governments and essential stakeholders.

The Global Fund.

Product guidance, financial support and policy driving.

Our solutions leveraging data science, artificial intelligence, and machine learning to benefit humanity. 

I believe the response to the "core technology" question clearly indicates our use of AI, to recap:

The core technology utilised in this solution is the disease predictive algorithm, primarily developed using ResNet34 (Convolutional Neural Network) and gradient boosting techniques. To leverage the core technology an adequate training database is required. In a supervised training algorithm such as this, both the data inputs and output need to be present in each participant's dataset. It is this training dataset that we are currently populating.

I believe the response to the "Who does it serve" question clearly indicates the way in which we will benefit humanity, to briefly recap: 

The improved screening will result in earlier TB diagnosis for infected patients and thus earlier treatment. This will improve treatment outcomes while protecting these typically impoverished communities by reducing disease transmission.

The prize money will be used to complete the data collection study and product development.