• There are no suggestions because the search field is empty.

Overview of digital pathology in developing countries

Developing countries carry the greatest disease burden in the world but AI has the potential to assist the pathology workflow.
Written by Aiforia

Introduction

Developing countries carry the greatest disease burden in the world yet resources for diagnostics and health care are limited and hard to access. This is particularly prominent within pathology, a field critical to all other healthcare workflows. There is a scarcity of pathologists worldwide and developing countries are particularly affected.

According to the World Health Organization, the density of pathologists in high income countries is approximately 1 in 15,000, but only 1 in 1 million in some countries in Africa (1). Meanwhile in Pakistan, there are less than 1,000 pathologists specialized in cancer diagnostics covering a population of over 225 million people.

Pathologists are critical to diagnostics, directly affecting treatment and patient outcomes. About 95% of clinical pathways rely on access to pathologists while pathology services are only available to less than 30% of patients in low-income countries (2). As physician shortages are acute and resources are lacking, low-income and middle-income countries, hosting 87% of the global population, need more efficient ways to combat pathologists’ overwhelming workloads (3).

Graph of estimated number of pathologists per 100,000 people in selected countries

Number of pathologists per 100,000 people-1

Data gathered from a collection of sources (11-16)

As remote work has increased, particularly during the COVID-19 pandemic, pathology and medical diagnostics have increasingly begun to rely on digital and virtual tools, such as telepathology, to keep up with demand. However, this is not new. Digital pathology and artificial intelligence (AI) technologies have been adopted in medical fields worldwide for years, promoting speed and efficiency, while tapping into the vast knowledge base of the worldwide medical community. These technologies also provide a possible solution for remote areas with limited medical choices or low-resource communities unequipped to handle their population’s pathology needs.

Challenges

In Pakistan, for example, there is an absence of pathology slide scanners, digital microscopes, and whole slide image (WSI) software. Often, remote pathologists are asked to visit their colleagues for second opinions because their digital images are not enough for a reliable consultation, particularly in difficult cases. And when scanners and digital microscopes are available, they are used in research and education, not for routine clinical diagnosis.

However, healthcare is at a turning point where digital pathology and AI are becoming increasingly common and available. With a widening knowledge base from worldwide contributors, pathologists can test out these softwares without the need for expensive equipment. Talat Zehra, a pathologist promoting the use of AI-assisted diagnostics work in Pakistan, carried out successful projects on chorionic villi and malarial parasite detection with the help of Aiforia without a scanner or digital microscope (4).

“I have no words to express my gratitude to Aiforia. Knowing that I don’t have a scanner or digital microscope, not only did they give me their demo version but also provided me with training very patiently.” - Dr. Talat Zehra, Aiforia User (5)

 

Blog Post Photos-2

Benefits

Unknown to most, the benefits of telehealth and digital pathology have been around for decades as remote collaboration becomes more common. Despite slow adoption in many countries, digital Whole Slide Images (WSIs) are replacing the glass slide in light of their ease for online sharing and remote analysis. Implementing WSIs into telemedicine and digital pathology enable pathologists to collaborate on projects worldwide, gathering second opinions and validations with ease. These digital versions of slides also open possibilities for AI to be incorporated. AI is particularly applicable in pathology sample analysis, which depends largely on pattern recognition.

Aided by AI-based image analysis software, it is possible to assist healthcare in developing countries. AI models achieve faster and more accurate diagnoses than pathologists due to decreased observer biases and errors. Through simple annotation to train the AI Model, they are able to identify, quantify, and measure features faster and more accurately than human-based microscopy.

This reduces turnaround time, improves diagnostic performance, and provides consistent and accurate care to patients in a timely manner. While delegating the time-consuming aspect of sample analysis to a computer, pathologists can optimize their work and spend more time verifying abnormal results and treating patients. In countries such as Kenya, which holds one pathologist for every million people, this is lifesaving.

AI speeds up cervical cancer screenings

Cervical cancer is a major cause of death for women worldwide despite its high preventability with early detection. An estimated 65% of Kenyan women diagnosed with cervical cancer in 2006 died (6). Pap smears, a method of cervical screening for cancerous cells, significantly decreases the risk of developing further disease but limited resources means there is a severe lack of regular screenings in Kenya.

Regular pap smear sample analyses are labor intensive, prone to variations in sensitivity and reproducibility, and require medical experts to analyze the samples. In an effort to combat this issue an international research team from the University of Helsinki and Karolinska Institutet, led by Aiforia co-founder Professor Johan Lundin, created “MoMic”, a handheld microscope made from low-cost equipment (7). It scans pap smears that are uploaded to a cloud-based server. AI models developed using Aiforia Create are then used to analyze the samples and identify abnormal cells before cancer has time to develop.

Blog Post Photos copy

Lab manager Martin Muinde scans pap smears at the Kinondo Clinic in Kenya. Photo: Oscar Holmström.

The team worked with Kinondo Hospital in the rural area of Kenya’s southeast coast, collecting pap smears from 740 women. The AI model was trained to identify precancerous lesions and smears without lesions with an accuracy of 96-100% and 78-85%, respectively. Such technological advancements are aiding the women in Kenya receive better diagnostic and oncological care. Read the publication here.

Expanding the technology to malaria

AI models are not constrained to any certain type of disease. With sufficient annotation and training, AI models are flexible and able to cover any pathology need. In 2019, nearly half of the world's population was at risk of malaria, a disease particularly dangerous for children under the age of five (8).

Most malaria cases and deaths occur in resource-poor countries in sub-Saharan Africa. Detecting malaria parasites in blood smears is time-consuming, laborious, and error-prone with conventional microscopy. Microscopic examination of Giemsa-stained blood smears, taken from a finger prick, is the easiest and most reliable test for malaria and remains the gold standard for laboratory confirmation of parasites (9). However, accuracy depends on the quality of the reagents, the microscope, and the experience of a trained microscopist.

Combined with limited resources, current diagnostic methods are insufficient for the large number of malaria cases in developing countries. AI models can be trained using annotated datasets to find parasites in blood smears, as demonstrated in this case study using Aiforia’s AI software. Automated image analysis is particularly beneficial for malaria diagnosis where pathologists are overloaded. Healthcare professionals can instead focus on preparing the slides of blood samples, verifying the results, and tending to patients.

Demonstration of Aiforia's AI model to find parasites in blood smears for detecting malaria

Educational advantages

The benefits of digitization and AI extend beyond hospitals and pathology labs. Medical education in pathology and histology in low-resource countries face various obstacles including limited equipment and telecommunication challenges. Many educational institutions in high-resource countries are adopting virtual and digitized technology.

The benefits of technologically advanced education are also possible in developing countries. Virtual slides, accessible through a local server or portable drive may be a solution to the high bandwidth requirement of digital pathology or in places where internet is unavailable. In a report on student experience at the University of the Philippines with virtual slides on a local network and a remote image server, digital pathology has been shown to provide advantages over the usual method of teaching histology and pathology (10). Furthermore, collaborations with universities in developed countries can help build image collections for teaching within institutions as well as sharing with others.

The future

Digital pathology and AI methods for medical diagnosis and research are being adopted in developing countries. While these technologies can be resource intensive, digitally aided diagnosis is a possibility in all regions of the world and may be the key to providing sufficient healthcare worldwide. AI also has the potential to increase quality, efficacy and democratization of care.

The benefits of AI-assisted analysis far outweigh traditional microscopy, and its use will become inevitable. AI is transforming decades-old practices and the beginning of this is already seen in the integration of image analysis and machine learning into routine histopathology. This is an exciting development that could enable predictive clinical knowledge and potentially address international pathology work-force shortages. The role of a pathologist will not only be reactionary, rather they will have the option to work on preemptive health, validation of diagnoses, and improving patient treatment plans. Thus AI-enabled digital pathology use will transform pathologists’ traditional clinical practices, which has been unchanged for decades.

References

  1. WHO Report on Cancer, 2020
  2. The Royal College of Pathologists
  3. Nabi J, The Lancet, 2018 
  4. Healthcare in Europe, 2021
  5. Aiforia Interview with Dr Talat Zehra
  6. Ng’ang’a A, et al. BMC Public Health, 2018 
  7. Holmström O, et al. JAMA Network Open, 2021 
  8. WHO Malaria Fact Sheet, 2021
  9. WHO World Malaria Report, 2020
  10. Fontelo P, et al. Analytical Cellular Pathology, 2012
  11. Pathologists In Kenya, Labtestzote.com
  12. The Eagle Online, 2019
  13. Sawai T, et al. Journal of Pathology Informatics, 2010
  14. Pathologist, Careers.govt.nz
  15. Eurostat, Physicians by medical speciality [data: HLTH_RS_SPEC, MED_PAT], 2021
  16. WHO, Global Health Observatory, Medical and Pathology Laboratory Personnel