A robot’s diagnosis: how AI is reshaping medicine

In today’s world, Artificial Intelligence (AI) is transforming every field, including medicine. The use of computer science dealing with intelligent computing is of great assistance in myriad domains of human existence.

In Pakistan, quality healthcare is a fundamental right but is unavailable to millions of people who do not have the resources to have access to private medical facilities. Quality of healthcare, even in the private sector, is not always up to the mark. One area that requires much more attention and resources is that of testing. Inaccuracy in tests and delay in reports often lead to misdiagnoses, wrong treatments and even loss of life.

Access to good healthcare is an issue that Pakistan has been facing for decades. Unavailability of specialists in the interior areas of Pakistan leads to harmful delays in diagnosing treatable diseases.

Many of these issues can be decreased by equipping essential medical equipment with AI. Some organizations are working to bring this technology to Pakistan. However, many barriers are yet to be removed to take full advantage of AI, particularly the use of AI in radiology.

Importance of AI

Radiology is the branch of medicine that through X-ray, CT, ultrasound, and MRI pictures creates medical imaging to find tumors and deformities. AI systems can automatically identify intricate abnormal patterns in visual data to help doctors diagnose patients.

There are several reasons why AI is important in radiology. For this, we talked to Umair bin Mansoor, Assistant Professor at the Electrical Engineering Department of DHA Suffa University. Mansoor with his team of students is working to bring AI in radiology in Pakistan.

Another reason is that AI can help to reduce the workload of radiologists. Medical imaging is a rapidly growing field, and the number of images that radiologists need to interpret can be overwhelming. By using AI to assist with image analysis, radiologists can focus on the most complex or unusual cases rather than on routine or straightforward ones.

Benefits of AI in Radiology

Inclusion of AI in radiology can greatly benefit Pakistan’s healthcare system. As Mansoor said, it could help improve accuracy, efficiency, and consistency, reduce workload, and save costs.

Accuracy: According to Mansoor, "AI algorithms can be trained to recognize patterns in medical images indicative of certain diseases or conditions. This can help radiologists to diagnose a patient’s condition more accurately, leading to more timely and effective treatment.”

Consistency: When different radiologists diagnose the same image, there can be a difference in each opinion. This not only confuses the patient but takes more time for treatment. "By using AI to standardize the interpretation of medical images, it is possible to reduce the variability that can occur when different radiologists interpret the same images. This can help to ensure that all patients receive the same high level of care, regardless of who is interpreting their images," Mansoor said.

Efficiency: AI in the field of radiology could increase the efficiency of medical imaging results. "AI algorithms can be trained to recognize patterns in medical images indicative of specific diseases or conditions, which can help radiologists to interpret images and diagnose a patient's condition more quickly and accurately. This can help to reduce turnaround times and improve patient care. However, it is essential to note that the impact of AI on the efficiency of medical imaging results will depend on several factors, including the quality and availability of training data, effectiveness of the AI algorithms being used, and overall skill and expertise of the radiologists interpreting the images.

"It is worth noting that while AI can improve the efficiency of medical imaging results, it is not a substitute for the skills and expertise of trained radiologists. Radiologists play a vital role in interpreting medical images and providing patient care, and their skills and knowledge will continue to be important even as AI technology make more advancements."

Reduced workload: In Pakistan, the number of good radiologists in proportion to the size of the population is not very high, and most of them spend a great deal of time diagnosing simple issues, leaving less time for them to work on more complex cases. AI can help to reduce the workload of radiologists by automating specific tasks and enabling them to focus on more complex or unusual cases.

Cost saving: When tests are more accurate and consistent, there will be no need for unnecessary tests and procedures, and this will eventually reduce the cost of healthcare.

Access: Use of AI in the field of radiology could increase access to X-rays and other medical imaging services in certain areas. "AI could be used to help interpret medical images in areas with a shortage of trained radiologists, which could help to fill gaps in care and improve access to medical imaging services. Impact of AI on access to medical imaging services will depend on several factors, including the availability of X-ray equipment, cost of using AI technology, and the overall health needs of the population," Mansoor said. He added that it is worth noting that while AI can improve access to medical imaging services in some areas, more is needed for the skills and expertise of trained radiologists.

Challenges

When it comes to adopting AI in radiology, Mansoor’s team and others working in the same field face several challenges that need immediate attention.

From the adoption and use of AI in the field of radiology to gathering the data, some of the main challenges include:

Data quality and availability: One of the main challenges in development of AI in radiology is the quality and availability of training data. For AI algorithms to be effective, they must be trained on a large and diverse dataset. However, medical images are often difficult to obtain, and there may be legal or ethical issues surrounding using certain types of images for training purposes.

As Mansoor explained, there might be privacy concerns related to the use of patient data for training purposes, or there may be issues regarding obtaining the necessary permissions to use certain types of images. Data quality can also be challenging, as the data may need to be completed or reliable, which can affect the accuracy of an AI algorithm. "There are also logistical challenges to gathering data, such as storing and managing large amounts of data and ensuring that the data is properly labelled and organized. These challenges can be time-consuming and resource-intensive, and to overcome them, specialized expertise may be required.

“Overall, while the challenges to gather data for AI in radiology can be significant, they are manageable. By addressing these challenges, it is possible to develop AI algorithms that are accurate and effective, and that can improve patient care and efficiency of the healthcare system," he said.

Regulation: Another challenge is clearng regulatory guidelines for using AI in medical imaging. While many AI algorithms have the potential to improve patient care, there is also a need to ensure that the use of these algorithms is safe and effective.

Talking about the general regulatory challenges that the world faces when adopting AI in radiology, Mansoor explained that the use of AI in healthcare is subject to various regulations and guidelines, which can vary by country. “These regulations may cover data privacy, patient safety, and development and testing of AI algorithms. Healthcare organizations need to understand and comply with these regulations to ensure AI's safe and effective use in patient care. It is worth noting that the use of AI in healthcare is a rapidly evolving field, and regulatory frameworks may be updated or changed over time to reflect technological advances and industry changes. Healthcare organizations must stay updated with these changes and ensure they comply with the latest regulatory requirements," he said.

Ethical concerns: There are also ethical concerns surrounding the use of AI in radiology, such as the potential for bias in algorithms and the impact on employment in the field. Using AI in radiology could lead to job loss in the industry. However, it is also important to note that AI is expected to replace radiologists partially and more to assist them in their work.

Mansoor said, “Use of AI in radiology will likely lead to changes in how work is done, and it may require radiologists to acquire new skills or specialize in certain areas. AI can improve the healthcare system's overall efficiency, which could lead to the creation of employment in other areas. Ultimately, AI's impact on employment in radiology will depend on several factors, including the rate at which AI technologies are adopted, willingness of radiologists to adapt to new technologies, and the overall demand for medical imaging services.”

Resistance to change: Some radiologists may resist adopting AI technologies due to concerns about job security or a preference for traditional image interpretation methods. It is common for doctors, including radiologists, to have questions or doubts about the results produced by AI systems. This is because AI algorithms are designed to assist with interpreting medical images, but they are not a substitute for the skills and expertise of trained radiologists.

Mansoor explained that radiologists make final diagnoses and treatment recommendations based on their interpretation of medical images. They need to have confidence in the AI systems' accuracy and reliability. "AI systems must be thoroughly tested and validated for radiologists to receive appropriate training to use these systems effectively. Adoption of AI in radiology is still in the early stages, and there is ongoing research and development to improve the accuracy and reliability of these systems. As AI technology continues to evolve, it is likely that doctors will become more comfortable using these systems and have fewer doubts about their results," he said.

This is not only limited to doctors; some hospitals are also hesitant. Several factors can influence a hospital’s willingness to adopt AI, such as availability of funding, perceived benefits of a particular technology, and level of support from hospital leadership. External factors, such as regulatory requirements or adoption of AI by other hospitals in the region, may also influence hospitals.

"Hospitals will likely be more open to adopting AI if they can see clear benefits, such as improved accuracy and efficiency in image interpretation, and if they have the necessary resources and infrastructure to support the successful implementation and use of these technologies. It is also important for hospitals to carefully consider the potential challenges and risks associated with AI adoption and to develop strategies to address these challenges," said Mansoor.

Cost: Implementation and use of AI technologies can be a barrier to their adoption in some settings. Cost can vary depending on several factors, such as the specific AI technologies being used, size and complexity of the healthcare organization, and availability of internal resources to support the implementation of AI.

"Cost of implementing AI in radiology can include the initial purchase or licensing fees for the AI software or hardware, as well as ongoing maintenance and support costs. There may also be costs associated with training staff on how to use the AI technology and adapting existing workflows and processes to incorporate AI. The potential benefits of these technologies, such as increased accuracy and efficiency in image interpretation and reduced workload for radiologists, may offset the cost of implementing AI in radiology. It is important for healthcare organizations to consider the costs and benefits of AI adoption carefully and to ensure that they have the resources and infrastructure to support the successful implementation and use of these technologies," said Mansoor.

Standardized image acquisition: It is essential to ensure that images are of sufficient quality and consistency to be accurately interpreted by radiologists. Several factors can affect the quality of medical images, including the type of imaging equipment being used, positioning of the patient, and technical expertise of the person performing the scan.

Mansoor explained that one way to address this challenge is to ensure that all medical imaging equipment is appropriately maintained and calibrated, and that all equipment staff is adequately trained and follows established protocols. "It can be helpful to establish guidelines or protocols for image acquisition, to ensure that images are consistently of high quality and are suitable for interpretation by radiologists," he said.

Mansoor added that while standardized image acquisition can be a challenge, it is essential to ensure the accuracy and effectiveness of medical imaging. Addressing this challenge can help improve patient care and efficiency of the healthcare system.