OPTIMIZE Diagnosis of Cervical Cancer Using Order Parameter Based Image Based Analysis

Cervical cancer is preventable, yet it ranks as the fourth most common cancer in women globally, with over 500,000 cases annually. The incidence varies greatly between high and low resource settings; 85% of cases occur in low- and middle-income countries (LMICs), particularly among under-screened populations in rural and low-income areas. Kenya, a low-income country, is among those with the highest cervical cancer rates, notably in Western Kenya where HIV prevalence ranges from 20.9% to 28.5%. For instance, cervical pre-cancer incidence in developed nations is around 0.6-0.8%, but can climb to 8% among HIV-positive women in Kenya. The World Health Organization (WHO) aims to eliminate cervical cancer globally by vaccinating 90% of girls against human papillomavirus (HPV), screening 70% of women, and treating 90% of cervical disease.

Current cervical cancer prevention strategies vary by location and resources, with screening methods ranging from pap smears and HPV tests to visual inspections like VIA/VILI. These approaches necessitate trained providers, exam rooms, and equipment for colposcopy, biopsy, and treatment. However, navigating multiple visits and potential pathologic reviews poses challenges, deterring some patients from completing recommended interventions. Obtaining cervical specimens through biopsies during colposcopy, while usually brief and minimally painful, can be distressing and lead to avoidance of follow-up due to anxiety and fear.

While WHO supports simplified screening and treatment in resource-limited settings, histopathologic analysis of clinician-guided biopsies remains the diagnostic standard for cervical pre-cancer. Clinicians base biopsies on specific visual cues during colposcopy, which may be subtle and mimic benign conditions, resulting in detection rates ranging from 30% to 70%. Incorporating HPV testing can enhance sensitivity but risks overtreating benign conditions, potentially harming patients. Treatments like cervical conization can affect reproductive health by shortening the cervix, limiting future options, and increasing preterm birth risk. Objective, affordable triage tests are needed to enhance the accuracy and value of screen-and-treat protocols.

We have a working proof of concept, noninvasive software program that uses completely innovative technology of image-based analysis using x-ray diffraction to determine the degree of disorder for the immediate diagnosis of cervical pre-cancer. This technology has the ability to assist the eradication of cervical cancer by improving detection of precancerous lesions and reducing barriers to care.

Our image-based analysis diagnosis when applied to colposcopy images can provide an immediate and rapid diagnosis based on a magnified image of the cervix. This software program has the capability to be incorporated into existing hardware during a colposcopy procedure. This expeditious result allows a clinician to treat immediately if results are positive, or provide a negative result right away to patients without needing cervical biopsies. 

The program can also function to provide a spatial map of disorder for the lesion to guide biopsy location to improve accuracy, and to determine the size and extent of the lesion. This information can be used to determine eligibility for ablative procedures (cryotherapy, laser or thermal ablation, or borders needed for an excisional procedure (loop electrocautery excision of the transformation zone or cold knife cone procedure) for immediate or subsequent treatment. This technology greatly improves the detection of precancerous lesions.

Based on our retrospective review of over 400 colposcopy images, our image analysis method has perfect (100%) accuracy. This high-level of accuracy is attributed to our quantitative methodology which measures the degree of structural disorder through image analysis. This purely objective method of diagnosis is superior to current qualitative methods of assessment. Currently, colposcopy performed by clinicians observes the presence during an exam of acetowhite changes with acetic acid, vascular abnormalities and non-staining Lugol's areas under a microscope. 

Our solution is designed to address the significant need for cervical cancer screening in under-screened populations, particularly in Kenya, aiding both professionals and patients. In this country, fewer than 20% of women have ever been screened for cervical cancer, and less than 13% have been screened in the last 5 years, despite cervical cancer being the second most common cancer among Kenyan women, with 5,250 cases reported annually. Traditional screening methods used in developed nations have proven ineffective in low-resource settings due to limited pathologists, healthcare providers, and challenges in patient follow-up. To make a real impact and improve existing models, technology must be specifically tailored for success in these high-risk populations.

Western Kenya faces numerous challenges including scorching heat, poverty, lack of health insurance, limited infrastructure, cultural and language barriers, high rates of immunosuppression due to the HIV/AIDS epidemic, and shortage of medical staff including clinicians and pathologists. Within western Kenya, women typically undergo HPV self-sampling or receive VIA/VILI tests performed by nurses. Mobile colposcopy with camera units is a common method employed by nurses in rural and limited resource areas. Our technology streamlines this process by providing immediate interpretation at the time of screening. Within minutes, both the nurse and patient can ascertain if the patient has normal, low-grade, or high-grade changes. If pre-cancer is detected, the cervical mapping feature can inform the nurse whether immediate thermal ablation is suitable or if referral to higher-level care is necessary for an excisional procedure based on lesion size and extent. This technology empowers nurses in the field to conduct screening and treatment, easing the burden on higher-level clinicians.

For higher-level clinicians such as medical officers or obstetrician-gynecologists, our technology can assist in performing guided biopsies or mapping excisional procedure borders, thereby enhancing clinical performance and intervention effectiveness.

Women who have access to this technology may avoid the cost and discomfort of a cervical biopsy and the need for a return visit by receiving a definitive diagnosis and/or treatment during a single visit. Our technology is mobile, affordable, reusable, and can be integrated into existing widely-used tools, potentially reducing barriers for patients. This innovation holds promise in transforming cervical cancer care and reducing its burden in low-resource settings.

To execute this project, we have assembled a team of Kenyan and international experts in physical science, clinical care, epidemiology and innovation surrounding the diagnosis of cervical pre cancer. 

Robert Makin, our team lead, is an Assistant Professor in the School of Computer Science at Western Michigan University. He is the innovator of this software technology. He previously patented this technology for the diagnosis of melanomas and has completed preliminary studies with his sister, Jennifer Makin, demonstrating its diagnostic accuracy for detecting cervical precancer.

Drs Jennifer Makin, Jackton Omoto and Stephen Gwer are affiliated with Jaramogi Oginga Odinga Teaching and Referral Hospital (JOOTRH) in Kisumu, Kenya and the sub-county hospital Lumumba. JOOTRH is the center for referrals of abnormal cervical cancer screening and cares for an average of 30 patients weekly at their colposcopy clinic. Lumumba is currently a center for screening and outreach and is established within the region as a women’s hospital. They screen an average of 200 individuals every month.

Jennifer Makin is an Assistant Professor of Obstetrics and Gynecology and Director of the Global Health Program and Lower Genital Tract Disease at UPMC. She completed a global health fellowship where she worked for two years at a rural hospital in Western Kenya. She has served the Luo Community in Kisumu and Siaya County for over ten years. She currently spends between 8-12 weeks annually in Western Kenya.

Dr Omoto, Director of the Colposcopy Clinic at JOOTRH, and the Department Chair of Obstetrics and Gynecology has completed extensive research regarding safety and feasibility of several cervical dysplasia screening and treatment methods in western Kenya including. Our innovation is a tool that has been strongly desired, as it many perceive artificial intelligence as the next step necessary to overcome the barriers of understaffing and cost in Kenya.

Dr Gwer is the Division Chair of Maseno University Department of Obstetrics and Gynecologists based in Kisumu, Kenya. He has collaborated with Dr Omoto in the past and has expertise in piloting novel innovations within women's health. He has sought and received a patent for a reuseable syringe extender designed to provide anesthesia for gynecologic procedures. Reducing painful gynecologic procedures is an important component to improving patient seeking behaviors and our noninvasive image-based diagnosis would be a highly attractive service.

Dr Fredrick Otieno is an epidemiologist and the Director of the Nyanza Reproductive Health Center, a nongovernmental research institute, located within the Lumumba Hospital. This organization has current and ongoing research collaborations in women’s health and cervical cancer screening/treatment.

Laura Makin is a health care administrator currently working within the Women's Health Service Line within a health care system in New Jersey. She is an expert in financial analysis and planning.

We would be an excellent choice for funding for this innovation. There is a strong partnership between UPMC and our Kenyan collaborators. Our family network and long-standing partnership makes our team cohesive and strong. We are well established within the community we intend to serve. 

We have executed studies demonstrating our image-based analysis is highly accurate for melanoma. The software program has been patented for this purpose. We have performed preliminary studies that have demonstrated perfect accuracy using order parameter image-based analysis for the diagnosis of cervical precancer based on retrospective studies. We performed retrospective study of over 4,000 images of cervical cytology slides to calibrate our technology and determine the order ranges for each diagnosis of cervical dysplasia ranging from normal to low grade, high grade and cancerous changes. We then applied these order parameters to colposcopy images of a cervix and demonstrated our technology remains perfectly accurate for the diagnosis of cervical cancer. We have yet to build the software prototype that would be downloaded on colposcopy camera unit. We do not yet have any customers.

We are applying to Solve for assistance in forwarding our technology. We are currently executing retrospective and prospective studies of diagnostic accuracy, but we have yet to pilot the impact of this technology on a population. If we are selected, we would use this funding to continue software development and complete our prototype. Funding is needed for a larger GPU server for storage and running the computer algorithms. We also require assistance with exposure to and networking with the companies and leaders of existing colposcopy camera units which could incorporate our software technology. Once we are able to partner with a colposcopy camera unit we can begin to complete feasibility and impact studies. The publicity and exposure through this award would be useful for advertising our program to other universities and nongovernmental institutions working in limited resource settings.

Our solution offers innovative advancements in cervical cancer screening and triage protocols. There is a need for more objective, affordable, and accessible triage tests to improve the positive predictive value of screen and treat protocols. Traditional methods often rely on subjective assessments, leading to challenges in accuracy and accessibility. Our approach utilizes sophisticated image analysis techniques, particularly a novel concept known as the Bragg-Williams order parameter (S), originally developed for characterizing structural ordering in material systems.

It has been shown theoretically and confirmed experimentally that S has a direct relationship to system level properties and can be used to predict key properties of material systems. The Bragg-Williams measure of structural ordering can be adapted to measure the degree of ordering in other physical systems, such as polymers or organic systems. It is possible to extend the concept of S to cellular systems. Underlying the methodology is the concept of structural motifs, which describe the nearest neighbor environment of each protein. These order parameters can be used to distinguish benign, pre-cancerous and malignant conditions using a numerical value of an order parameter (S) extracted from cellular images using a numerical analysis.Through rigorous retrospective testing, we have identified specific order parameter values that correlate with normal, low-grade, and high-grade cervical dysplastic changes, achieving perfect accuracy when compared to histopathology—the gold standard.

Unlike deep learning approaches that require extensive datasets and computational power, our technology delivers rapid, accurate results within minutes using minimal computing resources. While deep learning can achieve accuracies around 90%, our method boasts flawless performance without these resource-intensive demands.

Practically, our innovation eliminates the need for invasive cervical biopsies and reliance on pathologists, thereby reducing costs and enabling immediate diagnosis and treatment by nurses in underserved areas. By minimizing biopsy-related trauma, our solution also enhances patient compliance and engagement with screening programs. Additionally, our technology enhances the accuracy of diagnosis and biopsy mapping, improving the effectiveness of targeted therapies. Through all of these avenues, our technology has the capacity to increase screening, improve identification and treatment of cervical pre-cancer thereby reducing the incidence of cervical cancer.

Our technology has the ability to target under screened populations worldwide in order to make a meaningful impact on cervical cancer screening, treatment of cervical pre-cancer and ultimately the elimination of cervical cancer.

Our technology has the potential to reduce multiple barriers to care for diagnosis and treatment of cervical pre-cancer in four ways. 

First, by providing the diagnosis based on an image this eliminates the requirement of a cervical biopsy, which is costly and requires a pathologist. 

Second, this technology can support task shifting and provide immediate and accurate diagnosis in the field for screening and treatment performed by nurses in limited resource areas. 

Third, by minimizing trauma from reducing the number of cervical biopsies it can reduce fear and anxiety surrounding the procedure and improve patient seeking behavior. 

Finally, by improving the accuracy of the exam fewer cases of cervical precancer will be missed. By improving the accuracy of biopsies and mapping the borders of the region clinicians can improve the effectiveness of a selected therapy. 

Through all of these avenues, our technology has the capacity to increase screening, improve identification and treatment of cervical pre-cancer thereby reducing the incidence of cervical cancer.

Our highest priority is ensuring healthy lives and promoting well-being for all, at all ages. Our impact goals are centered around improving cervical cancer screening and treatment outcomes, enhancing patient experiences, and ultimately reducing the incidence of cervical cancer. We are measuring our progress towards these goals through several key initiatives:

1. Tracking Cervical Pre-cancer Detection: We are conducting prospective cohort studies to track the detection of cervical pre-cancer using our image-based analysis compared to the gold standard of cervical biopsies. This allows us to assess the accuracy and effectiveness of our technology in identifying pre-cancerous lesions.

2. Comparing Biopsy Performance: We are evaluating the performance of clinician-guided biopsies during colposcopy versus biopsies guided by our artificial intelligence. This comparison helps us understand how our technology enhances the precision and efficacy of biopsy procedures.

3. Feasibility Studies: We are conducting feasibility studies to integrate our software into existing colposcopy microscopes and smartphones. This involves assessing technical compatibility and practical implementation of our solution in real-world clinical settings.

4. Usability Assessment: We are assessing the ease of use and human-computer interaction of our software. This includes evaluating the usability of our interface to ensure it is intuitive and user-friendly for healthcare providers.

5. Patient Experience Evaluation: We are studying the impact of our software on patient experience and satisfaction. This involves gathering feedback from patients undergoing screening and treatment facilitated by our technology to identify areas of improvement.

Long-Term Outcome Tracking: We are tracking long-term outcomes to monitor the progression of disease to cervical cancer among individuals screened with our software. This helps us gauge the effectiveness of our solution in preventing the development of cervical cancer over time.

In the 1930s, Bragg and Williams characterized structural ordering in crystalline lattice systems with a single numerical value, the Bragg-Williams order parameter (S). It has been shown theoretically and confirmed experimentally that S has a direct relationship to system level properties and can be used to predict key properties of material systems. The Bragg-Williams measure of structural ordering can be adapted to measure the degree of ordering in other physical systems, such as polymers or organic systems. It is possible to extend the concept of S to cellular systems. Underlying the methodology is the concept of structural motifs, which describe the nearest neighbor environment of each protein. These order parameters can be used to distinguish benign, pre-cancerous and malignant conditions using a numerical value of an order parameter (S) extracted from cellular images using a numerical analysis.

The Bragg-Williams method originally was developed to measure the ordering of the arrangement of atoms in lattice materials using the intensities of peaks in x-ray diffraction patterns; the modified approach extends this framework to non-lattice-based physical systems using any wave-based imaging technique. The images of the cells are processed by contrast analysis technique using the pixel intensity histogram. Images of cells are blindly interpreted based on their order with 0 set as complete disorder, correlating with atypical cells and 1 defined as perfect order and correlating with normal cytology. Our results were cross-referenced with the tissue confirmed diagnosis and the mean and standard deviations were determined for each cell classification. 

We have an abstract as the American Society for Colposcopy and Cervical Pathology at the 2024 annual meeting in New Orleans titled "Detection of Cervical Precancerous Cells through Order Parameter Based Image Analysis."

We have submitted two papers for publication respectively to the Journal of Lower Genital Tract Disease and Frontiers titled "Diagnostic Accuracy of Order Parameter Based Image Analysis of Cervical Precancerous Cells" and "Diagnostic Accurancy of Order Parameter Based Image Analysis of Colposcopy Images."  

Robert Makin has previously validated this technology method for diagnosis of melanoma and patented this technology. US20230000426A1 - Quantitative image-based disorder analysis for early detection of melanoma type features - Google Patents

Robert Makin and his team in Western Michigan have previously published this novel technology using a modified Bragg-Williams approach to measure the ordering of the arrangement of atoms in lattice materials using the intensities of peaks in x-ray diffraction patterns; the modified approach extends this framework to non-lattice-based physical systems using any wave-based imaging technique.

R.A. Makin, K. York, S.M. Durbin, N. Senabulya, J. Mathis, R. Clarke, N. Feldberg, P. Miska, C.M. Jones, Z. Deng, L. Williams, E. Kioupakis, and R.J. Reeves, “Alloy-Free Band Gap Tuning across the Visible Spectrum,” Phys. Rev. Lett. 122(25), (2019). PhysRevLett.122.256403 (aps.org)

R.A. Makin, K. York, S.M. Durbin, and R.J. Reeves, “Revisiting semiconductor band gaps through structural motifs: An Ising model perspective,” Phys. Rev. B 102(11), (2020). 10195979 (nsf.gov)

 R.A. Makin, K.R. York, A.S. Messecar, and S.M. Durbin, “Quantitative Disorder Analysis and Particle Removal Efficiency of Polypropylene-Based Masks,” (2020). div-class-title-quantitative-disorder-analysis-and-particle-removal-efficiency-of-polypropylene-based-masks-div.pdf (cambridge.org)

R. A. Makin, S. N. Hanumantharao, S. Rao, and S. M. Durbin, MRS Advances 8, 386 (2023). 10405450 (nsf.gov)

Full time staff:

Robert Allen Makin II and Jennifer Makin

Part Time Staff: 

Lauran Makin, Jackton Omoto, Stephen Gwer, and Fredrick Otieno

Robert Makin has been developing this technology since 2020. He has experimented and refined the image-based analysis around biologic systems, patenting first his technology with diagnosis of melanoma. He and Jennifer Makin began calibrating this technology for the diagnosis of cervical precancer in September of 2023.

We have an international and diverse team. We have a close collaboration with clinicians and researchers in Kenya, with three key team members who are Kenyan. Our team spans a range of career stages and ages with a pHd student, a health care administrator in the early stages of her carreer, two assistant and two associate professors. We have two women on the team. One who is progressively moving up leadership within her institution and recently took the leadership role of Director of the Lower Genital Tract Center at UPMC Magee in Pittsburgh. We are dedicated to diversity, equity and inclusion and actively working to create a diverse and welcoming environment where all team members are respected, supported and valued.

Our business model is designed to address the critical need for cervical cancer screening among underserved populations, particularly those of low socioeconomic status in rural areas of low and middle-income countries. Our initial target region is Western Kenya, leveraging our current team's expertise and network in this area.

Our value proposition centers on improving patient satisfaction by reducing cervical biopsies and consolidating patient visits. We anticipate that clinicians will benefit from the technology's accuracy, facilitating more targeted treatment decisions and biopsy locations, ultimately enhancing their confidence and effectiveness in providing care.

To effectively implement our solution, we will forge strategic partnerships with leading companies that specialize in colposcopy equipment, such as Cooper, Wallach, Olympus, or providers of mobile units like Lutech, ColpPhon, or MobileODT, which are compatible with our software platform.

We plan to utilize a subscription-based service for our revenue stream, which offers several key advantages. It provides a predictable revenue stream allowing for better financial planning to continue ongoing innovations in software. It will also allow us to scale our model to different levels of usage and need, and ensure we can align pricing with the volume of access required ensuring affordability. This model will also allow for lower barriers to customers price-wise encouraging adoption, while creating long-term customer relationships.

Costs associated with our business plan are fixed labor and purchasing of GPU servers.

Key stakeholders for investment in our innovation include Kenyan non-governmental organizations, the Ministry of Health, County Governments, regional universities such as Maseno, the Kenyan Medical Research Institute, and USAID.

Furthermore, we will build a donor network targeting organizations like the World Bank, Gates Foundation, AbbVie, and the International Foundation to secure funding to support our initiative.

Critical activities and key milestones in our roadmap include developing the prototype, conducting impactful studies to assess feasibility, user experience, accuracy, and patient satisfaction. As we progress towards market readiness, we will prioritize patenting our technology and establishing a compelling brand identity through logo and website development.

While other artificial intelligence programs exist, our offering stands out for its perfect accuracy achieved through a quantitative focus, enabling significantly quicker responses.

Our plan for achieving financial sustainability revolves around strategic funding, partnerships, and targeted marketing of our software program for cervical cancer screening.

Initially, our research exploring order parameter image-based analysis was indirectly funded through the National Science Foundation, highlighting our ability to secure research funding for innovative projects.

To fund the development of our prototype and impact studies, we are pursuing research grants and preparing an NIH application for prospective diagnostic accuracy studies in Kenya. This demonstrates our proactive approach to securing funding for critical development phases.

Our software program is dependent on partnerships with companies producing colposcopy hardware incorporating camera units. We intend to collaborate with these companies, sharing profits from the service of their instruments integrated with our software.

Upon developing our prototype, our software will be marketed through partnerships with institutions like the Nyanza Reproductive Health Center in Kisumu, Kenya, offering subscription-based access tailored to the number of colposcopy units utilizing the program. 

A key benefit of our business model includes low overhead costs for software maintenance, with initial profits reinvested into advertising and further software development. During the initial service years, we anticipate the need for subsidization and investment from healthcare stakeholders such as national or county governments, or organizations like USAID, given our focus on serving low-income populations.

To demonstrate the success of our financial sustainability plan, we will track key metrics including revenue growth from subscription fees and partnerships, successful grant applications securing funding for development phases, and evidence of impactful partnerships with colposcopy hardware companies. Additionally, successful adoption and utilization of our software by healthcare institutions will serve as tangible evidence of our business model's viability and sustainability.