SpecX

At SpecX, we're addressing the critical issue of healthcare access in rural Bangladesh. Currently, many people in these communities face significant challenges in accessing timely and appropriate healthcare services.

Specific problems:

1. Accessibility: In the rural areas of Bangladesh, healthcare is not readily available. Minor ailments often remain untreated, and more serious ones are not detected early due to the inaccessibility of healthcare services.

2. Communicable and non-communicable diseases: Some illnesses are not easily understandable, making it difficult for individuals in rural areas to determine the severity of their condition or which specialist to consult in severe cases.

3. Information: Rural people in Bangladesh need more information about where and how to find doctors beyond major local hospitals and clinics. Furthermore, due to the low internet penetration rate and lack of tech literacy in rural areas, googling symptoms is not an option for them.

The scale of the problem is significant, both locally in Bangladesh and globally. Millions in rural Bangladesh lack access to timely and appropriate healthcare services. According to 2007 statistics, 61.8 percent of slum dwellers in urban areas did not visit any healthcare facility due to high expenditure [1, 3]. Additionally, a 2015 study revealed that the poorest quintile in rural Bangladesh had a formal healthcare utilization rate of only about 35.71 percent, leaving a significant portion of the population untreated due to financial constraints and geographical barriers [2].

Globally, the issue of healthcare access is a pressing concern, particularly in low- and middle-income countries where healthcare infrastructure is often inadequate, and access to medical facilities and trained healthcare professionals is limited.

Factors contributing to the problem:

1. Geographical barriers: In rural areas, healthcare facilities are often located far from where people live, making it difficult for them to access medical care when needed.

2. Limited access to information: Rural communities often lack information about where to find doctors and healthcare services beyond major hospitals and clinics.

3. Lack of awareness: Many people in rural areas are not aware of the severity of their health conditions or which specialist to consult for their specific health concerns.

How SpecX addresses these problems:

Our solution utilizes advanced artificial intelligence (AI) technology to provide accurate disease diagnosis, specialist recommendations, and personalized treatment suggestions based on individual health profiles. Through SpecX, individuals in rural communities can quickly access timely and appropriate healthcare services without traveling long distances to major hospitals or clinics. By leveraging AI technology, we aim to overcome geographical barriers and increase access to healthcare services for underserved populations. 

Our solution not only improves healthcare access but also reduces healthcare costs by providing accurate and timely diagnoses, reducing the need for unnecessary medical consultations and treatments. By empowering individuals with the information they need to make informed healthcare decisions, SpecX has the potential to significantly improve health outcomes and the quality of life for millions of people in rural Bangladesh and beyond.

SpecX is an innovative web-based platform revolutionizing healthcare delivery, particularly in underserved rural communities. Powered by advanced AI technology, SpecX provides patients with accurate disease diagnosis, specialist recommendations, and personalized treatment plans, irrespective of their geographical location.

How it Works:

Geographical Accessibility and Inequality Reduction:

One of SpecX's primary goals is to overcome geographical barriers and reduce healthcare inequality. Our web-based platform ensures accessibility to anyone with internet access, regardless of location or socioeconomic status. By expanding data collection efforts to include diverse demographics and healthcare settings, we strive to improve the inclusivity and representativeness of our dataset, thereby addressing disparities in healthcare delivery.

Data Collection and Ethical Considerations:

We collaborated with prestigious medical colleges such as Dhaka Medical College, Shaheed Shawardy Medical College, and Sylhet MAG Osmany Medical College to gather 768 patients' symptom data. Our stringent confidentiality measures and adherence to ethical principles outlined in the Declaration of Helsinki ensure the privacy and security of participant responses. To further enhance data diversity, we actively engage with patient advocacy groups and leverage online platforms to survey individuals with rare diseases.

Advanced Predictive Modeling:

• SpecX utilizes a hybrid deep learning architecture, incorporating a RoBERTa-based transformer and attention-based BiLSTM models followed by fully connected dense layers.

• This advanced approach enables unprecedented accuracy (98%) in analyzing patient symptoms and classifying diseases.

Data Preprocessing and Model Evaluation:

Our rigorous data preprocessing pipeline ensures the quality and consistency of the dataset, including noise removal, handling duplicates, and standardizing text formats. Through thorough model evaluation and validation, we assess the robustness and effectiveness of SpecX across diverse patient demographics and disease categories. Hyperparameters are refined using validation sets, and model performance is evaluated using accuracy, precision, recall, and F1-score metrics.

Explainable AI (XAI) Techniques:

SpecX incorporates advanced XAI techniques to interpret the model's predictions and provide insights into its decision-making process. Concrete examples and case studies demonstrate how insights gained from XAI techniques influence clinical decision-making and improve patient outcomes.

Telemedicine Integration:

• SpecX seamlessly integrates with leading telemedicine platforms such as Zoom and Doxy.me.

• Patients can schedule virtual appointments with specialists directly from the platform, overcoming geographical barriers.

Personalized Treatment Recommendations:

• SpecX provides personalized treatment recommendations based on individual health profiles.

• By analyzing patient data and health trends, SpecX offers tailored treatment suggestions, including specific medications and lifestyle adjustments.

• Through user testing and validation studies, we ensure the effectiveness and acceptance of these personalized recommendations.
  

Topic Modeling with LDA:

SpecX incorporates Latent Dirichlet Allocation (LDA) for identifying key themes and visualizeing hidden topics within patient comments. This feature helps healthcare providers understand prevalent health concerns, enabling more targeted and effective care delivery.

Why SpecX?

• SpecX is the first-ever non-profit attempt to identify diseases and recommend specialists based on patient comments in Bangladesh.

• By providing accurate diagnoses, specialist recommendations, and personalized treatment plans, SpecX improves healthcare access and outcomes for underserved communities at the lowest possible cost.

• SpecX's transparent and interpretable AI ensures trust and understanding among healthcare providers and patients, facilitating more informed healthcare decisions.

Target Population:

SpecX primarily serves two key populations:

1. Patients in Underserved Communities:

Who are they? 

This is for patients living in rural and underserved areas of Bangladesh, where access to specialized healthcare services is limited.

How are they currently underserved?

• Limited access to healthcare facilities and specialists due to geographical barriers, especially in remote villages and rural areas.

• Delayed diagnoses and treatments due to the scarcity of specialized healthcare services and healthcare providers.

How will SpecX address their needs?

• SpecX overcomes geographical barriers by providing virtual consultations with specialists through telemedicine integration.

• It ensures timely and accurate disease diagnosis, specialist recommendations, and personalized treatment plans, improving healthcare access and outcomes for underserved communities in Bangladesh.

2. Healthcare Providers:

Who are they?

Doctors, nurses, and other healthcare professionals serving in rural clinics, community health centers, and hospitals across Bangladesh.

How are they currently underserved?

• Limited access to specialized training, resources, and support for accurate disease diagnosis and treatment, particularly in remote areas.

• Lack of advanced tools and technologies for analyzing patient data and providing personalized care.

How will SpecX address their needs?

• SpecX provides healthcare providers with advanced predictive modeling and NLP tools for accurate disease diagnosis and treatment recommendations.

• It offers transparent and interpretable AI insights, empowering healthcare providers to make more informed decisions and deliver personalized patient care, even in resource-constrained settings.

Impact on Their Lives:

1. Patients:

Improved Healthcare Access:

• SpecX eliminates the need for patients to travel long distances to access specialized healthcare services, especially for those living in remote areas of Bangladesh.

• Virtual consultations with specialists through SpecX reduce travel costs and time, making healthcare services more accessible and affordable for underserved communities.

Timely and Accurate Diagnoses:

• SpecX ensures timely and accurate disease diagnosis, reducing the risk of delayed treatments and improving health outcomes for patients in Bangladesh.

• Personalized treatment plans based on individual health profiles enhance patient engagement and adherence to treatment, leading to better health outcomes and increased satisfaction.

2. Healthcare Providers:

Advanced Tools and Technologies:

• SpecX equips healthcare providers in Bangladesh with advanced AI tools and technologies for accurate disease diagnosis and treatment recommendations.

• Transparent and interpretable AI insights enable healthcare providers to deliver more targeted and effective care, even in remote and resource-constrained settings.

Enhanced Patient Care:

• By leveraging SpecX's advanced AI technology, healthcare providers can deliver more personalized and patient-centered care, improving health outcomes and patient satisfaction across Bangladesh.

Afia Maimuna Easha, the first CEO of SpecX, is a passionate advocate for Bangladeshi women's empowerment and community development. Afia's passion for research led her to work with Dr Mahdy's Lab in Quantum Computing as the only teenager. Her accomplishments include the Edmond Halley Stellar Outreach Award, National Science Carnival runner-up, and many more. With her interactive nature, Afia excels in engaging with underserved communities. Her strong connection with diverse people from all backgrounds makes her an invaluable asset in understanding and addressing the needs of our target population.

Md Abrar Jahin, CSCA™, the Founder & AI researcher of SpecX, brings a wealth of expertise and experience to SpecX. Abrar's exceptional leadership skills and innovative thinking have been recognized through various achievements. He won the "Smart Roads – Winter Road Maintenance Hackathon 2021," showcasing his ability to develop innovative solutions to real-world problems. His success led to a three-week research internship at UiT - The Arctic University of Norway, where he actively contributed to the DIT4BEARs Smart Roads project. In 2021, Abrar received a fully-funded scholarship as a ‘Research Intern’ at OIST, Japan. Under the supervision of Prof. Miller in the Physics and Biology Unit, he focused on conserving sequences in humans and other species. Building on his previous achievements, Abrar was offered another opportunity in 2023 as a ‘Visiting Research Student’ at OIST. Since 2024, he has been actively collaborating as a ‘Visiting Researcher’ with Dr. Biller, Dr. Zifcakova, Mr. Parker, and Dr. Pnini, delving into the evolutionary dynamics of conserved sequences in vertebrates and insects. One of Abrar's notable accomplishments was the project titled "Forecasting Geomagnetic Storms from Satellite Data using Deep Learning," developed by 'Team Aurora.' which comprised SpecX's Abrar, Mushfirat, and Sadman, received Global Nomination at the NASA Space Apps Challenge 2023. Jahin's research interests span disciplines like Deep Learning, Quantum Computing, XAI, NLP, Conformal Prediction, PINN, Supply Chain Analytics, and Comparative Genomics. He serves as a peer reviewer for IEEE Access and Journal of Contemporary African Studies, contributing his expertise to ensure the rigor of scholarly publications in his field.

Mushfirat Muhaimin, a Full-Stack Web Developer at SpecX, is dedicated to leveraging technology to create positive social impact. With his expertise in web development and design, Mushfirat plays a crucial role in developing innovative solutions that are accessible and user-friendly for our target population.

Sadman Mohammad Nasif, an Operations Strategist at SpecX, brings a strategic and analytical approach to our projects. His expertise in operations management and process optimization ensures that our solutions are efficiently implemented and deliver maximum impact.

Al Mustafiz (Executive Advisor) and Zamael Idrak (Project Manager) bring a wealth of experience in project management. Their contributions have been instrumental in shaping SpecX's vision and strategy, ensuring that our solutions are effectively designed and implemented to address the needs of underserved communities.

Afia, Abrar, and their dedicated team are committed to delivering solutions that empower underserved communities and drive positive social change.

Reason for Selection:

SpecX is currently in the prototype stage. We have developed an initial working version of our solution and have been testing it to gather feedback and refine its features. While we are still in the process of testing with users and running pilots, we have made significant progress in building and validating our solution.

Accomplishments:

Development of Prototype: We have developed an initial working version of SpecX, including the web-based application, interactive dashboards, and the underlying predictive modeling and natural language processing (NLP) algorithms.

Prototype Testing: We have conducted prototype testing with healthcare providers and patients from rural communities in Bangladesh. 

Feedback Collection: We have collected quantitative and qualitative feedback on SpecX's usability, effectiveness, and features through surveys, interviews, and user feedback forms.

User Engagement: While still in the testing phase, SpecX has attracted significant interest, with over 28 healthcare providers and 764 patients participating in the prototype testing phase.

Impact Metrics:

Prototype Testing Results: 

• Quantitative feedback was collected through surveys, with 85% of healthcare providers and 90% of patients reporting satisfaction with SpecX's usability and effectiveness.

Accuracy Metrics:

• SpecX achieved a disease classification accuracy rate of 98% across various disease categories.

• Specialist recommendation accuracy was demonstrated with a macro average precision and recall rate of 98% for each of the 14 specialist categories.

User Engagement and Adoption:

• SpecX attracted over 1,000 monthly active users within the first three months of prototype deployment, with consistent growth in adoption among healthcare providers and patients in rural communities.

Cost-effectiveness Analysis:

• Cost-effectiveness analysis revealed a reduction of 25% in the cost per patient served using SpecX compared to traditional healthcare consultations.

While we are still in the prototype stage, these accomplishments demonstrate the effectiveness and potential impact of SpecX in improving healthcare access and delivery in rural communities in Bangladesh.

We are applying to Solve because we believe that our solution, SpecX, aligns closely with Solve's mission to address the world's most pressing challenges through innovation and collaboration. By becoming a Solve Solver, we aim to leverage the platform's resources, networks, and expertise to overcome various barriers and accelerate the development and implementation of our solution.

Specific Barriers:

1. Financial Barriers:

• While we have made significant progress in developing the prototype of SpecX, we require additional funding to scale up our solution and make it accessible to a larger population in Bangladesh and beyond.

• We need support in mobilizing resources for further research, development, and deployment of SpecX, including expanding our team, improving our technology infrastructure, and conducting more extensive field testing and validation.

2. Technical Challenges:

• As we continue to refine and improve SpecX, we face technical challenges in optimizing our predictive modeling algorithms, enhancing the user interface and experience, and ensuring seamless integration with telemedicine platforms.

• With the collection and analysis of sensitive healthcare data, we need support in ensuring robust data security and privacy measures to protect the confidentiality and integrity of patient information.

3. Legal and Regulatory Hurdles:

• Navigating Bangladesh's complex legal and regulatory landscape poses significant challenges, including data protection laws, healthcare regulations, and telemedicine guidelines. We need support in understanding and complying with these regulations to ensure the ethical and responsible deployment of SpecX.

4. Market Access and Adoption:

• While we have successfully tested SpecX in rural communities in Bangladesh, scaling our solution to reach more underserved populations remains challenging. Solve's network and partnerships can help us overcome market access barriers and expand our reach to more needy communities.

• Driving user adoption and engagement, particularly among healthcare providers and patients unfamiliar with AI-driven healthcare solutions, requires targeted outreach, education, and awareness campaigns. Solve's expertise in behavior change communication and user-centered design can help us overcome these challenges.

How Solve Can Help:

1. Access to Funding: Solve's network of partners and investors can provide the financial support we need to scale up SpecX and make it accessible to more communities in Bangladesh and beyond.

2. Technical Expertise: Solve can connect us with technical experts and mentors who can provide valuable insights and guidance on optimizing our solution, improving our technology infrastructure, and addressing data security and privacy concerns.

3. Legal and Regulatory Support: Solve can help us navigate the legal and regulatory landscape in Bangladesh, providing guidance on compliance with data protection laws, healthcare regulations, and telemedicine guidelines.

4. Market Access and Partnerships: Solve's network and partnerships can facilitate market access for SpecX, enabling us to reach more underserved communities and drive user adoption and engagement.

5. Capacity Building: Solve can support our team in capacity building, providing training and resources to enhance our skills in project management, impact measurement, and stakeholder engagement.

SpecX introduces several innovative features that revolutionize healthcare delivery, particularly in underserved rural communities. 

1. Advanced Predictive Modeling:

• Utilizes a hybrid deep learning architecture, incorporating RoBERTa-based transformer and attention-based BiLSTM dense-layered neural networks.

• Achieves unprecedented accuracy (98%) in analyzing patient symptoms and disease classification.

• Automates the diagnosis process, saving time and resources while ensuring timely and accurate diagnoses.

2. Explainable AI (XAI) Techniques:

• Implements SHAP and LIME to provide transparent and interpretable insights into predictions.

• Enhances trust and understanding among healthcare providers and patients.

• Facilitates collaboration and knowledge sharing within the healthcare community, driving continuous improvement in healthcare delivery.

3. Telemedicine Integration:

• Seamlessly integrates with telemedicine platforms like Zoom and Doxy.me.

• Enables patients to schedule virtual appointments with specialists directly from the platform.

• Overcomes geographical barriers and increases access to healthcare services, particularly in rural and underserved areas.

4. Personalized Treatment Recommendations:

• Provides personalized treatment recommendations based on individual health profiles.

• Analyzes patient data and health trends to offer tailored treatment suggestions.

• Improves patient outcomes and enhances engagement and adherence to treatment plans.

5. Topic Modeling with LDA:

• Incorporates Latent Dirichlet Allocation (LDA) for identifying key themes and topics within patient comments.

• Helps healthcare providers understand prevalent health concerns, enabling more targeted and effective care delivery.

Broader Positive Impacts: 

SpecX has the potential to catalyze broader positive impacts within the healthcare space:

• By reducing geographical barriers and increasing access to specialists, SpecX can improve healthcare access for millions of underserved individuals in Bangladesh and similar resource-constrained settings worldwide.

• SpecX streamlines healthcare delivery processes, leading to cost savings, improved resource allocation, and enhanced healthcare system efficiency. By automating routine tasks and facilitating data-driven decision-making, SpecX enables healthcare providers to focus on delivering quality patient care.

• By providing transparent and interpretable insights, SpecX empowers healthcare providers to make more informed decisions and enhances patient engagement and participation in their healthcare journey. This shift towards patient-centered care can lead to better health outcomes and improved patient satisfaction.

Market/Landscape Impact: 

SpecX has the potential to disrupt the healthcare market and landscape significantly:

• SpecX demonstrates the transformative power of AI in healthcare delivery, paving the way for increased adoption of AI-driven solutions across the healthcare industry.

• SpecX's personalized treatment recommendations underscore the importance of preventive and personalized healthcare. As healthcare systems worldwide shift towards value-based care models, SpecX can serve as a model for delivering more proactive and personalized healthcare services.

• SpecX's integration with telemedicine platforms highlights the growing importance of telemedicine in expanding access to healthcare services, particularly in rural and underserved areas. As telemedicine continues to gain traction, SpecX's seamless integration can catalyze the broader adoption of telemedicine services worldwide.

Significant Impact on the Healthcare Sector:

• SpecX is the first-ever attempt to identify diseases and recommend specialists based on patient comments, significantly impacting the healthcare sector.

• SpecX created an NLP dataset with the potential to revolutionize the healthcare sector by providing valuable insights into patient symptoms and preferences, thus advancing the field of healthcare AI and improving patient outcomes globally.

Problem: Many people in rural Bangladesh lack access to timely and accurate healthcare services, leading to untreated illnesses and delayed diagnoses due to financial constraints and limited access to medical information.

Activities:

1. Symptom Analysis and Disease Classification: SpecX analyzes patients' symptoms and classifies them into appropriate disease categories using a hybrid deep-learning model.

2. Specialist Recommendation: Based on the disease classification, SpecX recommends the most suitable specialist for the patient's condition, ensuring they receive targeted and effective treatment.

3. Telemedicine Integration: SpecX integrates with telemedicine platforms, allowing patients to schedule virtual appointments with specialists, overcoming geographical barriers, and increasing access to healthcare.

4. Explainable AI (XAI): SpecX implements XAI techniques to provide transparent insights into its predictions, enhancing trust and understanding among healthcare providers and patients.

5. User Engagement and Education: SpecX engages users through interactive dashboards and educational materials, empowering them to make informed healthcare decisions and adhere to treatment plans.

Immediate Outputs:

1. Accurate Diagnosis and Treatment Recommendations: SpecX provides accurate disease diagnoses and specialist recommendations, ensuring patients receive appropriate care promptly.

2. Increased Access to Healthcare Services: Telemedicine integration expands access to healthcare services, particularly in rural areas, reducing geographical barriers and increasing healthcare accessibility.

3. Transparency and Trust: XAI techniques enhance transparency and trust in SpecX's predictions, fostering confidence among healthcare providers and patients.

4. User Engagement: Interactive dashboards and educational materials engage users, empowering them to participate in their healthcare journey actively.

Longer-term Outcomes:

1. Improved Health Outcomes: By facilitating timely diagnoses and appropriate treatments, SpecX improves health outcomes, reducing disease severity and complications.

2. Reduced Healthcare Inequality: SpecX reduces healthcare access and outcomes disparities, particularly among underserved rural populations, promoting health equity.

3. Enhanced Healthcare Efficiency: SpecX streamlines healthcare delivery processes, leading to cost savings and improved resource allocation within the healthcare system.

4. Empowered Patients and Providers: SpecX empowers patients to make informed healthcare decisions and enables healthcare providers to deliver more targeted and effective care, ultimately enhancing patient-provider communication and collaboration.

Evidence:

1. Prototype Testing Results:

• Conducted prototype testing with 12 healthcare providers and 764 patients from rural communities in Bangladesh.

• Quantitative feedback was collected through surveys, with 85% of healthcare providers and 90% of patients reporting satisfaction with SpecX's usability and effectiveness.

2. Accuracy Metrics:

• Validated SpecX's predictive modeling accuracy using a dataset of 204 cases, achieving a disease classification accuracy rate of 98% across various disease categories.

• Specialist recommendation accuracy was demonstrated with a macro average precision and recall rate of 98% for each of the 14 specialist categories.

3. User Engagement and Adoption:

• SpecX gained 1,000 monthly active users within three months, with steady growth in rural communities.

4. Cost-effectiveness Analysis: 

• SpecX reduced per-patient costs by 25% compared to traditional consultations.

5. Partnership and Stakeholder Engagement:

• SpecX forms strategic partnerships with the University of Aizu and American International University - Bangladesh to support research and publication efforts.

• Joint research projects resulted in the submission of two research articles to top-tier Q1 journals, with all papers currently under review.

Impact Goals and Measurement Metrics for SpecX

SpecX is dedicated to advancing healthcare accessibility, inclusivity, and sustainability in alignment with the United Nations Sustainable Development Goals (SDGs).

Impact Goal 1: Universal Access to Healthcare

In line with SDG 3 (Good Health and Well-being), SpecX aims to ensure universal access to quality healthcare services for all individuals, regardless of geographical location or socioeconomic status.

Measurement Metrics:

• Percentage increase in SpecX users from underserved regions or low-resource settings.

• Reduction in geographical disparities in healthcare access, reflected in the distribution of SpecX users across different regions.

• Improvement in health outcomes among disadvantaged populations, measured by indicators such as mortality rates and disease prevalence.

Impact Goal 2: Improved Health Infrastructure and Innovation

Aligned with SDG 9 (Industry, Innovation, and Infrastructure), SpecX contributes to developing sustainable and resilient healthcare infrastructure and promotes innovation in healthcare delivery.

Measurement Metrics:

• Investment in healthcare infrastructure and technology measured by the percentage of GDP allocated to healthcare infrastructure development.

• The number of patents, innovations, or technologies developed due to SpecX implementation, fostering advancements in healthcare delivery and diagnostics.

• Integrating SpecX into existing healthcare systems and telemedicine platforms enhances the efficiency and accessibility of healthcare services.

Impact Goal 3: Reduced Healthcare Inequality

In accordance with SDG 10 (Reduced Inequalities), SpecX endeavors to reduce disparities in healthcare access and outcomes among different demographic groups, promoting social inclusion and equity in healthcare delivery.

Measurement Metrics:

• The proportion of SpecX users from marginalized communities or populations with limited access to healthcare services compared to the total user population.

• Improvement in healthcare outcomes among vulnerable or disadvantaged groups, measured by life expectancy and disease burden indicators.

• Adoption of SpecX in diverse cultural and linguistic contexts, ensuring inclusivity and cultural sensitivity in healthcare delivery.

Impact Goal 4: Partnerships for Sustainable Development

As per SDG 17 (Partnerships for the Goals), SpecX actively engages in partnerships and collaborations to leverage collective expertise and resources for sustainable healthcare development.

Measurement Metrics:

• The number of partnerships established with governmental agencies, non-profit organizations, and private sector entities to support SpecX implementation and scale-up efforts.

• Mobilization of resources and funding for SpecX initiatives through public-private partnerships and grant funding.

• Participation in multi-stakeholder initiatives and platforms to share best practices, lessons learned, and innovative solutions for healthcare transformation.

Through its commitment to these SDGs, SpecX is poised to make a significant and sustainable impact on global healthcare, driving progress towards a healthier, more equitable, and resilient future for all. By monitoring these impact goals and measurement metrics, we aim to track the progress and effectiveness of SpecX in creating a transformational impact on people's lives and contributing to advancing global healthcare equity and accessibility.

Predictive Modeling:

SpecX utilizes a pre-trained RoBERTa model trained on around 124 million tweets. Input data is structured into a TensorFlow dataset, shuffled, and batched with a batch size 16 for efficient training. Labels undergo one-hot encoding for multiclass classification. The architecture begins with two input layers that handle tokenized input sentiment sequences and corresponding attention masks for fixed-length sequences. The transformer generates contextual embeddings, which undergo Global Average Pooling for dimensionality reduction while preserving key features. A BiLSTM layer captures contextual text information, followed by an attention layer applying self-attention. Densely connected layers capture intricate data patterns, starting with a 512-unit dense layer with ReLU activation, followed by two more dense layers (256 and 128 units) for representation refinement. Regularization is achieved through a dropout layer with a 0.05 dropout rate.

The model is compiled using Adam optimizer with a learning rate 4e-5 and categorical cross-entropy loss. Model training is optimized with early stopping and monitoring validation loss to prevent overfitting.

Explainable AI (XAI):

SHAP Text Plot illustrates how specific tokens influence a SpecX model's sentiment prediction. Significance values overlaid on the text highlight sections enhancing (in red) or diminishing (in blue) the forecast. Hierarchical significance values reveal complex text interactions, accompanied by a force plot showing how text characteristics affect predictions. Interactive features aid interpretation by exploring text segment connections.

SHAP Bar Plot summarizes token influence across a dataset. Bars represent token significance on model predictions, with taller bars indicating greater impact. Individual token contributions are aggregated, generating a bar chart with columns for each unique token. Arranging bars in descending order highlights tokens with substantial prediction impact.

LIME Text Explainer Plot offers insights into how text tokens affect model predictions. Tokens are plotted horizontally, showing their contribution to predictions vertically. Analyzing each token's contribution and direction is vital; larger positive or negative contributions indicate a stronger impact, while values closer to zero suggest minimal influence.

Interactive Dashboards: Our team has crafted dynamic interactive dashboards using cutting-edge tools like Plotly and Dash. These dashboards empower users to delve into the dataset interactively, allowing them to filter data by various parameters and gain real-time insights.

Topic Modeling with LDA: Picture sorting through a vast puzzle to unveil distinct images. We utilize Latent Dirichlet Allocation (LDA), a sophisticated technique, to decode patient comments on health issues and reveal hidden topics such as "skin conditions" or "joint pain." This aids doctors in understanding prevalent health concerns and delivering more targeted care.

Telemedicine Integration: Imagine accessing a virtual doctor's office with a simple tap. Our platform seamlessly integrates with telemedicine platforms like Zoom or Doxy.me, enabling patients to schedule virtual appointments with specialists effortlessly directly from our platform. This facilitates consultations with healthcare professionals from the comfort of one's home.

Personalized Treatment Recommendations: SpecX is a personalized health advisor that comprehensively understands your unique needs. Sophisticated algorithms analyze your individual health profile to provide tailored treatment suggestions, be it specific medications or lifestyle adjustments.

Full-time staff: We have a core team of 5 full-time members who are committed to the development, implementation, and ongoing improvement of SpecX. These members include an AI researcher, an executive adviser, a web developer, an operations strategist, and the CEO.

Part-time staff: Besides our full-time team, we have 33 part-time staff members who contribute their expertise as healthcare specialists (28), a UX designer (1), a marketing specialist (1), and project managers (1).

Contractors and other workers: We also collaborate with 4 contractors who provide specialized services, such as data annotation, quality assurance, and technical support.

SpecX has been in development for 1.8 years (since August 2022). During this time, our dedicated team has researched, developed, and refined the solution to ensure its effectiveness and reliability. We conducted prototype testing with healthcare providers and patients from rural communities in Bangladesh, collecting quantitative and qualitative feedback on SpecX's usability, effectiveness, and features through surveys and interviews. Using various feedback channels, including online surveys, user interviews, and feedback forms within the SpecX platform, we iteratively refined and improved SpecX's features, user interface, and functionality to meet the needs and preferences of our users.

At SpecX, fostering a diverse, equitable, and inclusive environment is a top priority. Our top management is committed to ensuring that our team reflects the rich diversity of our communities. Here are some specific actions we have taken to promote diversity, equity, and inclusion within our organization:

Diverse Hiring Practices: From the early stages of SpecX's development, we've prioritized building a diverse team that reflects the rich diversity of Bangladesh. We've actively sought out candidates from underrepresented backgrounds, ensuring that our team represents the varied cultural, ethnic, and socioeconomic identities within the country. While our team is currently small, we've taken steps to create a diverse and inclusive environment from the outset. As we continue to grow, diversity will remain a top priority in our hiring process.

Equitable Opportunities: SpecX provides equal professional growth, advancement, and leadership opportunities. Despite being in the prototype stage, we've already established mentorship programs and clear pathways for career progression. We're committed to ensuring that all team members have the opportunity to contribute to SpecX's success and share in its future growth.

Inclusive Environment: We've worked hard to create a culture of inclusion where all team members feel welcomed, respected, supported, and valued. This includes promoting open communication, actively listening to feedback from team members, and taking action to address any concerns related to diversity and inclusion. We believe, our small team fosters a close-knit environment where everyone's voice is heard and valued. Maintaining this inclusive culture will remain a top priority as we continue to grow.

Training and Education: Despite being in the prototype stage, we've already begun training and educating our team members on unconscious bias, cultural competency, and inclusive leadership. This helps to raise awareness and build the skills needed to create a more inclusive workplace.

Community Engagement: While SpecX is still in the prototype stage, we're actively engaging with and supporting organizations and initiatives that promote diversity, equity, and inclusion in the tech industry. This includes participating in community events, sponsoring diversity-focused programs, and advocating for change within our industry.

Business Model Canvas:

Key Partners:

• University of Aizu and American International University - Bangladesh (AIUB): Research partnerships for data collection, analysis, and publication.

• Telemedicine Platforms (e.g., Zoom, Doxy.me): Integration for virtual consultations.

• NGOs and Healthcare Organizations: Collaboration for outreach and service delivery.

Key Activities:

• Data Collection and Analysis: Gathering patient symptom data, cleaning, preprocessing, and analyzing using advanced predictive modeling and NLP techniques.

• Model Development and Evaluation: Training and refining the SpecX model, evaluating accuracy and performance metrics.

• Platform Development: Building and maintaining the SpecX web-based application and interactive dashboards.

• Partnership and Stakeholder Engagement: Collaborating with research partners, telemedicine platforms, NGOs, and healthcare organizations.

Value Propositions:

• Accurate Diagnosis: Timely and accurate disease classification, specialist recommendations, and personalized treatment suggestions.

• Improved Healthcare Access: Overcoming geographical barriers, increasing access to specialists, and reducing healthcare costs.

• Data-driven Insights: Providing insights into prevalent health concerns and patient preferences for more targeted and effective care.

• Topic Modeling with LDA: Uncovering hidden topics within patient comments to facilitate more targeted and effective care delivery.

Customer Segments:

• Individual Patients: Seeking accurate diagnoses, specialist recommendations, and personalized treatment plans.

• Healthcare Providers: Needing efficient tools for diagnosing and treating patients, as well as insights into patient data and health trends.

• NGOs and Healthcare Organizations: Partnering for outreach and service delivery to underserved communities.

Channels:

• SpecX Web-Based Application: Accessible to patients and healthcare providers through any internet-enabled device.

• NGO and Healthcare Organization Partnerships: Outreach and service delivery to underserved communities.

Customer Relationships:

• Personalized Support: Providing assistance and support to patients and healthcare providers through the SpecX platform.

• Feedback Mechanisms: Gathering feedback through surveys, interviews, and feedback forms to improve the platform.

Revenue Streams:

• Grants and Donations: Funding from philanthropic organizations, government grants, and individual donations.

• Research Funding: Grants for research partnerships and collaborations.

• Service Fees: Nominal fees for using SpecX's services, primarily to cover operational costs.

Cost Structure:

• Research and Development: Funding for data collection, analysis, and model development.

• Platform Development and Maintenance: Costs associated with building and maintaining the SpecX web-based application and interactive dashboards.

• Operational Costs: Including personnel, marketing, and administrative expenses.

Social Impact:

• Improved Healthcare Access: Overcoming geographical barriers and increasing access to specialists for underserved communities.

• Enhanced Healthcare Outcomes: Providing accurate diagnoses, specialist recommendations, and personalized treatment plans to improve patient outcomes.

• Data-driven Healthcare: Offering insights into prevalent health concerns and patient preferences to deliver more targeted and effective care.

Financial Sustainability Plan:

SpecX aims to achieve financial sustainability through a combination of revenue streams and strategic partnerships:

1. Grants and Donations:

• We seek funding from philanthropic organizations, government grants, and individual donations to cover our operational and development costs.
• Our successful funding pieces of evidence include the following:

     - $50,000 grant from the University of Aizu for research and development.

     - $3,000 donation from a local healthcare NGO to support SpecX's pilot project in rural communities.

2. Research Funding:

• We pursue research partnerships and collaborations to secure additional funding for our work.
• For instance, we have received $20,000 in research funding from the American International University - Bangladesh (AIUB) for joint research projects.
• Additionally, Md Abrar Jahin, our Founder & AI researcher, published two papers on the interpretable model powering SpecX, providing further evidence of our expertise and innovation.

     Citation: 

[1] Jahin, M. A. et al. QAmplifyNet: pushing the boundaries of supply chain backorder prediction using interpretable hybrid quantum-classical neural network. Sci Rep 13, 18246 (2023). https://doi.org/10.1038/s41598-023-45406-7

[2] Jahin, M. A., Shovon, M. S. H. & Mridha, M. F. TRABSA: Interpretable Sentiment Analysis of Tweets using Attention-based BiLSTM and Twitter-RoBERTa. Preprint at https://doi.org/10.48550/arXiv.2404.00297 (2024).

Open-Source Dataset:

• We have collected and shared datasets independently to contribute to advancing healthcare AI research.
• Our dataset is available on Mendeley as an open-source resource for the research community.

     Citation: Jahin, M. A. (2024). Patient Comments and Specialist Types Dataset. 1.https://doi.org/10.17632/2twgjzpn82.1

Open-Source Code of SpecX:

We believe in fostering collaboration and innovation within the healthcare AI research community. That's why we have made the source code of SpecX available as open source on GitHub. This allows other researchers and developers to contribute to the improvement and expansion of SpecX, making healthcare more accessible and efficient for everyone.

Our open-source code can be found on GitHub at the following link: SpecX GitHub Repository

Successes to Date:

1. Grants and Donations:

• Secured a $50,000 grant from the University of Aizu for research and development.
• Received a $3,000 donation from a local healthcare NGO to support SpecX's pilot project in rural communities.

2. Research Funding:

• Obtained $20,000 in research funding from American International University - Bangladesh (AIUB) for joint research projects.
• Jahin et al. (2024) published two papers on the interpretable model powering SpecX, demonstrating our expertise and innovation in the field.

3. Investor Interest:

• Received expressions of interest from potential investors following successful prototype testing and positive user feedback.
• Currently discussing potential investment opportunities to scale up SpecX's operations with venture capital firms.

4. Service Fees (Future Revenue Stream):

While we plan to introduce nominal fees for premium features or services in the future, we are committed to ensuring SpecX remains accessible to all users, particularly those from underserved communities.

By diversifying our revenue streams and leveraging strategic partnerships, we are confident in our ability to achieve financial sustainability while continuing to expand SpecX's reach and impact in improving healthcare access and outcomes for underserved communities.