In this article, we continue our interview series where we highlight urologists who have unique accomplishments outside the traditional clinical setting. We interviewed Dr Giovanni Cacciamani, faculty at the University of Southern California (USC), who has published several articles regarding artificial intelligence (AI); leads the UroGPT working group, the Young Academic Urologist working group in Uro-Technology and Digital Healthcare, and the global CANGARU (ChatGPT, Generative Artificial Intelligence and Natural Large Language Models for Accountable Reporting and Use) initiative to create AI guidelines for academia; and was recently featured in Time magazine and Nature News.

1. Can you tell us about your current academic position?

I graduated from medical school and completed my urology residency in Italy. I came to USC for a fellowship and now hold many roles here: I’m an associate professor of urology, vice chair of the research council, director of the resident research program, and director of the AI center. I work closely with fellows, residents, and students to bridge the gap between medicine and AI. Mentoring trainees is a crucial part of my work, and I have had the privilege of guiding nearly 50 medical students. I have 12 students working with me this year, which is really exciting. They’re passionate about AI, and I see them as a force of nature, motivated to create change in the world of urology. Ultimately, my passion for AI research is driven by a commitment to find innovative solutions to enhance patient care and outcomes.

2. How did you get involved in AI research?

Early on in my training—even before AI made its grand entrance as a mainstream buzzword—I saw the potential for AI to revolutionize how we diagnose and treat urological diseases. In fact, AI itself is not a new concept. It started 70 years ago with Alan Turing, who is considered the father of theoretical computer science and AI. The very first paragraph of his pivotal work examined “the imitation game,” where he asks the reader what would happen if a machine could potentially do the same tasks of a human.

Urology has always been a technologically driven specialty with a focus on lasers, endoscopes, and robotic surgeries. I saw many ways in which urology could embrace AI to help predict outcomes and improve staging. This was the beginning of my exploration into the vast potential of AI, and in 2021 at the USC urology department under Dr Inderbir Gill’s supervision, we took a significant step forward by establishing the first artificial intelligence center in urology in the country. This center has been instrumental in pushing the boundaries of AI applications in urology, and right now we’re working to develop international guidelines and to improve the validity of pathologic and radiological diagnosis for urological malignancies. Our collaboration extends to the department of engineering and department of radiology at USC. We are also working on frameworks that bridge the gap between complex technical jargon and patient-friendly language. Our goal is to ensure that patients receive digestible, reliable information and do not fall into the trap of misleading “fake news” on the internet.

3. How have you seen urologists using AI right now?

One critical challenge in urology is standardizing interpretation of clinical imaging modalities, especially in urological cancers. Fortunately, AI algorithms work to add reliability to diagnostics. For example, AI can identify subtle signs of cancer in a CT scan that might be missed by the human eye and can make a significant difference in patient outcomes. AI also proves valuable in the context of prostate multiparametric MRI. Based on the literature, we know that there is huge variability in PI-RADS (Prostate Imaging Reporting and Data System) scoring across radiologists. Even the same radiologist looking at the same MRI 6 months later may come up with a different score. Same patient, same radiologist, yet a different outcome. AI’s ability to read multiparametric MRIs smoothens variability between radiologists, which ultimately improves diagnostic accuracy and appropriate treatment for patients. Furthermore, AI can process and analyze findings in real time, providing more timely diagnosis. In fact, AI can analyze a provider’s video from a cystoscopy and provide insights by recognizing key structures and enabling a comprehensive view of the bladder.

AI can also interpret pathological slices, which is extremely time consuming. New scanners have implemented AI and can read slides immediately, reducing workload. And this is the main thing: reducing physician workload. Repetitive and time-consuming administrative tasks are a significant issue for all physicians, including urologists. Physicians often find themselves burdened with cumbersome documentation redundancies and other tedious tasks that eat away at their precious time. These tasks contribute to burnout, which is a concerning issue in modern medicine. AI offers an elegant solution: by automating administrative tasks such as clinical note-taking and discharge summary creation, AI can free up physicians’ time to focus more on humanistic care. This not only reduces the workload on urologists, but also ensures that records are maintained accurately and consistently.

4. What are some existing issues in urology that can be fixed with AI?

AI can help improve the accuracy of risk assessment. By analyzing a patient’s medical history, genetic factors, and lifestyle, AI can predict the likelihood of urological diseases. This can lead to more personalized and effective preventive measures. However, moving forward, we need to improve education. AI should be part of curricula in medical schools. One of my missions is not only teaching my own medical students and residents about AI, but also putting together a practical course on AI for the AUA that can educate the entire urology workforce about how to best utilize AI. We need to make sure that people understand that AI is just a tool that can help make our lives as physicians more efficient.

Think about AI as we think about our computers today—it would be archaic to practice medicine on paper now that we have electronic systems, and computers enable access to unlimited information at your fingertips. But at the end of the day, the computer could not replace you as a physician, in the same way that AI cannot go outside some borders. But we need to improve our knowledge of AI before diving deeper into how to best utilize machine learning to improve our daily practice.

5. Where do you see the future of AI?

The future of AI in health care is incredibly promising. AI can empower physicians to make their work more efficient and patient centered. One area where AI can be transformative is in surgical training. Imagine a scenario where an AI model is trained using the techniques of the most skilled surgeons in the world. This concept not only ensures consistent and top-level training, but also reduces the variability in surgical expertise. It will also be able to harness techniques from surgeons in different countries that otherwise would not be able to idea share. Furthermore, AI can play a critical role in surgery itself. During a surgical procedure, AI can act as a real-time assistant, offering insights and warnings, much like having an expert surgeon by your side. This kind of guidance can enhance surgical precision and safety.

AI also has a pivotal role to play in the evaluation and assessment of medical performance. Many of the metrics used in health care today are subject to human bias. AI, on the other hand, offers an objective, unbiased assessment of a physician’s performance. This could be a game changer, improving health care quality and patient safety. For example, it was reported a couple of years ago that surgeons actually perform worse on their birthdays,¹ potentially due to unconscious distraction. AI could check your operating schedule based on your birthday and alert you that human error is more common on birthdays, which then may make you more aware of your choices when (or if!) you operate that day.

6. What is your perspective on AI and ethics in health care?

AI ethics in health care is a topic of paramount importance. While AI has the potential to bring about transformative improvements in patient care, it also raises ethical concerns that need careful consideration. There is a researcher and engineer in Italy named Friar Paolo Benanti. He’s currently the advisor to Pope Francis on issues of AI and technology ethics, and he’s one of the most incredible people I’ve ever had the chance to meet. He coined the term “algorethics,” which is a new branch of psychoscience for evaluating and measuring the ethical performance and implications of AI. One fundamental ethical issue with AI is transparency. As providers, when we prescribe drugs to our patients, we know how they work and why we are giving the drug. But for the majority of deep machine learning, we don’t have an exclusive understanding of the entire knowledge acquisition or decision-making processes. We know the input and we know the output, but we don’t always know what AI is doing in the middle. Thus, making decisions as a physician based solely on AI’s output is limited by the fact that we can’t always explain the complex methodology behind AI. And explainability is the main thing in terms of liability. Thus, the same way that we supervise residents and medical students, we need to supervise AI.

I don’t see AI replacing physicians any time soon, and the main reason is that we see the patient holistically. We don’t only see the disease; we see the patient’s history, family, and the psychological impact of a disease. We see the follow-up and the consequences of our management plans. So there are many things that AI will never do, but AI could make our lives so much easier! The goal of AI isn’t to help us treat more patients, but rather to help us become more efficient, improve our ability to provide high-value care to our patients, and to elevate our own quality of life. AI won’t replace urologists; however, urologists who use AI will replace those who don’t.