How NTT DATA & are revolutionizing medical imaging

  • November 09, 2023

As artificial intelligence (AI) has crept its way into various industries, healthcare is no exception. Today, healthcare providers can leverage the power of AI to advance their practices and create better patient outcomes. Use cases in healthcare include medical imaging, disease diagnosis, remote patient monitoring and triage/workflow.

AI can speed up the interpretation of medical images. Using algorithms — a finite sequence of instructions to identify normal or abnormal anatomical characteristics — AI can improve operational efficiencies and expedite patient care. For example, AI can triage normal chest x-rays over chest x-rays with critical findings. This allows radiologists to interpret patient exams with critical findings at a higher priority.

However, people remain apprehensive about the idea of AI, much less using it as a decision-making tool for their health. How are we sure that we can trust the results of AI? The answer to that question is validation.

In simple terms, validation is the act of confirming the accuracy or validity of something. Validating an algorithm means checking to make sure that the output of the algorithm is accurate, which in turn gives us the confidence to trust the results of the algorithm in the future.

Statistical and clinical validation

For AI in healthcare, there are two types of validation: statistical and clinical. Statistical validation confirms that the model's statistical metrics, such as the Area Under the Curve (AUC) or Receiver Operating Characteristic (ROC) fall within a certain threshold of reliability. These metrics show how well the model can distinguish between distinct categories; a high AUC lets us know that we can be more confident in the model’s predictions.

Clinical validation is performed at the hospital level when key decision makers compare the AI results to the original image to verify that the AI results are correct. For example, during the validation process, AI might identify certain subtle patterns as potential diseases, which upon human review, turn out to be benign or normal anatomical variations. This collaborative approach ensures that the AI's capabilities complement the expertise of medical professionals, leading to more accurate diagnoses and better patient care.

Finding the right data ‘fit’

There are countless AI models used in healthcare today — some are FDA-approved while others aren't. However, FDA approval doesn't absolve the user from validating the algorithm's results. Using AI in healthcare means that a computer program, called a model, can be good at predicting or diagnosing things for one group of people.

The model may not work well with a different group of people. The model learns from data about a specific group of individuals, and if that data doesn't represent the group’s interest, the model's predictions can be off. As a result, the population in the data used to train the model is one of the most critical factors affecting how well the model performs.

Therefore, for a clinician to implement an AI solution safely and successfully, they must evaluate the results of the algorithm before deployment. Often, the clinician may need to 'try out' multiple AI solutions to find the right fit for their data.

While it can help create better patient outcomes, the amount of effort needed to adopt AI remains a large barrier to entry for many healthcare providers. To work with an AI vendor today, the clinician must provide a deidentified sample population to the vendor to determine how well a particular algorithm accommodates to their data and use case.


To improve healthcare quality, accessibility and affordability, NTT DATA has partnered with to accelerate the adoption of AI in clinical imaging for healthcare providers.'s single user interface, procurement and integration platform for medical imaging AI empowers healthcare providers to seamlessly discover, explore, validate and deploy third-party radiology AI solutions. This allows healthcare providers to focus on the clinical side of AI by providing a single platform for AI solutions, simplifying the administrative process with a single bill across all their vendors for AI.'s platform, combined with NTT DATA’s advanced data and analytics capabilities, helps clinicians harness data assets and deliver valuable insights for personalized, data-driven care, and increase the value of diagnostic imaging. Read the fact sheet for more information on how you can accelerate your radiology capabilities with AI today.

Subscribe to our blog

Vedrana Vujic

Vedrana is a consultant in the healthcare space with a master’s degree in Business Analytics. Leveraging her experience in data science, she works with clients in managing and mining clinical databases for AI research projects in healthcare and life sciences. Vedrana consults with a team working with healthcare organizations to gain insights into clinical operations and data quality. As a member of the Society for Imaging Informatics in Medicine, she actively contributes to industry advancements as a committee member on artificial intelligence, exploring data science and informatics in healthcare and life sciences.


Related Blog Posts