Multimodal AI-Based Renal Cancer Patient Care
Abstract
Kidney cancer is considered one of the most common cancers in the United States (approximately 80,000 new cases per year) and worldwide (450,000 new cases per year) and one of the most rapidly increasing cancers in American adults between 25 and 50 years old. Most kidney cancers, if caught early, can be treated and cured with surgery, however in about 30%-40% of patients, the cancer will metastasize. When kidney cancer spreads to other organs, about 85% of those patients will die within 5 years. Fortunately, most of the recent increase in kidney tumors have been diagnosed as small renal masses (tumor size less than 4 cm), which often turn out to be either benign (30%) or indolent (50%). However, with existing tools in clinical practice, it is difficult to sort out which kidney tumors are dangerous and which tumors can be left alone. This has led to overdiagnosis and over-treatment in recent years with little reduction in the number of people dying from kidney cancer. It is estimated that tens of thousands of patients each year may undergo surgery for a concerning kidney tumor that turns out to be benign or indolent, leading to an estimated national cost of $153 million ($55,573/individual) on kidney cancer treatments that may not help a patient live longer or better. Unfortunately, doctors often cannot differentiate between a benign kidney mass and kidney cancer on the patient s preoperative imaging. Biopsy for diagnosis of the kidney mass is not routinely used by most physicians, as it is an invasive procedure with a risk of complications and is sometimes inaccurate or non-conclusive. Thus, we aim to develop reliable, accurate, non-invasive tools that use artificial intelligence (AI) to evaluate patient information, including medical images, blood tests and other details like age, gender, etc., to accurately differentiate the benign or indolent masses from the potentially aggressive tumors in the preoperative setting. Our proposed tools will fill in this missing knowledge gap to assist the doctor and patient in choosing personalized treatment, and we will compare them in terms of time, efficiency, and accuracy to existing clinical tools. Current treatments for kidney cancer can be as significant a threat to patient outcome as the risk of a metastatic tumor. In the case of early kidney cancer, the main non-cancer related risk is the loss of kidney function during partial or total kidney removal surgery to take out the tumor. It is important to be able to predict what the kidney function will be after the surgery to help plan whether to remove all or part of the kidney when removing the tumor. We aim to create an automated tool using AI and preexisting information that all kidney tumor patients have collected during routine care, a patient s baseline kidney function plus preoperative computed tomography (CT) scans to accurately estimate kidney function. If we are successful, patients and doctors will have more accurate ways to predict kidney function so they can balance cancer and non-cancer risks and personalize the decision to remove all or just part of the kidney when operating on kidney tumors. To help advance the field rapidly, we will make some of our models and data publicly available so researchers around the world can evaluate our work, generate their own models from our data, and build on it and implement it in practice. When we complete our project, some patients can experience clinical benefit immediately, as the tool will be validated and ready to implement in the clinic. If successful, it will serve to increase confidence in detecting benign or indolent kidney masses preoperatively, potentially sparing thousands of patients from unnecessary surgery and treatment. Instead, patients with benign/indolent masses will be safely watched, potentially saving more than $150 million dollars and sparing patients the complications of unnecessary interventions. Furthermore, patients with malignant kidn
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 04, 2024
- Source ID
- HT94252310918
Entities
People
- Christopher Weight
Organizations
- Cleveland Clinic
- United States Army