By Billy Candra 
The advent of artificial intelligence is rapidly transforming various sectors, and its potential in healthcare, particularly in diagnostic imaging, is becoming increasingly apparent. A groundbreaking study from the UCLA Health Jonsson Comprehensive Cancer Center has recently illuminated the promising role AI could play in addressing one of the most challenging aspects of breast cancer detection: interval cancers. These are cancers that emerge and are diagnosed between scheduled mammography screenings, often presenting at a more advanced stage and carrying a poorer prognosis compared to screen-detected cancers. The UCLA research indicates that integrating AI into the screening process could significantly enhance early detection, leading to less aggressive treatment protocols and ultimately, improved survival rates for patients.
Understanding Interval Breast Cancers: A Critical Challenge in Oncology
Breast cancer remains one of the most common cancers globally, with millions of new cases diagnosed each year. In the United States alone, the American Cancer Society estimates hundreds of thousands of new invasive breast cancer cases annually. Early detection is paramount to successful treatment, significantly improving a patient’s chances of survival. For instance, the five-year survival rate for localized breast cancer is exceptionally high, often exceeding 90%, but this rate decreases substantially if the cancer has spread to regional lymph nodes or distant parts of the body.
Mammography is the cornerstone of breast cancer screening, widely recognized for its effectiveness in reducing breast cancer mortality. However, even with regular screening, a subset of cancers, known as interval cancers, invariably emerge. These cancers are particularly insidious because they manifest rapidly or possess characteristics that make them difficult to detect on prior mammograms, often slipping past human radiologists. They can be broadly categorized into several types: those truly invisible on a mammogram (occult), those with very subtle signs missed by radiologists (minimal signs), or those missed due to technical or interpretive errors.
The clinical significance of interval cancers cannot be overstated. Patients diagnosed with interval cancers often face a more aggressive disease course, requiring more intensive treatments such such as extensive surgery, chemotherapy, and radiation therapy. The emotional toll on patients and their families is also profound, as the delayed diagnosis can lead to feelings of frustration and anxiety. Consequently, strategies to reduce the incidence of interval cancers are a high priority in oncology research and public health initiatives.
The UCLA Health Jonsson Comprehensive Cancer Center Study: Methodology and Findings
Published in the esteemed Journal of the National Cancer Institute, the UCLA study represents a significant step forward in understanding how AI can augment existing screening paradigms. Led by Dr. Tiffany Yu, an assistant professor of Radiology at the David Geffen School of Medicine at UCLA and the study’s first author, the research focused on identifying "mammographically-visible" types of interval cancers earlier by flagging them at the time of initial screening. These are the tumors that, in retrospect, were visible on mammograms but either went undetected by human radiologists or presented with such subtle signs that they were arguably below the threshold of human detection.
Retrospective Analysis and Data Scope
The study was a comprehensive retrospective analysis, drawing upon a vast dataset of nearly 185,000 past mammograms performed between 2010 and 2019. This extensive period allowed researchers to examine a diverse range of screening practices, encompassing both digital mammography (DM), often referred to as 2D mammography, and digital breast tomosynthesis (DBT), commonly known as 3D mammography. From this large cohort, the research team meticulously identified 148 cases where a woman was subsequently diagnosed with interval breast cancer.
A critical phase of the methodology involved radiologists reviewing these 148 cases to pinpoint the reasons why the cancer was not initially detected. To achieve this, the study adapted a classification system predominantly used in European research, categorizing interval cancers into distinct types: "Missed reading error," where the signs were present but overlooked; "Minimal signs-actionable," indicating subtle signs that, upon review, should have prompted further investigation; "Minimal signs-non-actionable," where signs were exceedingly faint and difficult to act upon; "True interval cancer," referring to cancers that genuinely developed rapidly between screenings and were not visible retrospectively; "Occult," meaning truly invisible on the mammogram; and "Missed due to a technical error." This detailed classification was crucial for understanding the specific types of missed cancers that AI might be best positioned to address.
The Role of AI: Transpara Software
Following the human review, the researchers applied a commercially available AI software, Transpara, to the initial screening mammograms that preceded the interval cancer diagnoses. The objective was to ascertain whether the AI could detect the subtle signs of cancer that had eluded radiologists during their initial screenings, or at least flag them as suspicious. The AI tool assigned a cancer risk score to each mammogram, ranging from 1 to 10. A score of 8 or higher was designated as a potential concern, indicating an area that warranted closer scrutiny.
The findings from the AI analysis were compelling. Researchers estimated that integrating AI into routine screening protocols could potentially lead to a reduction of approximately 30% in the number of interval breast cancers. This figure represents a substantial clinical impact, translating to thousands of women who could receive an earlier diagnosis and, consequently, less aggressive and more effective treatment.
Dr. Yu emphasized the profound implications of this finding: "This finding is important because these interval cancer types could be caught earlier when the cancer is easier to treat. For patients, catching cancer early can make all the difference. It can lead to less aggressive treatment and improve the chances of a better outcome." The difference between an early-stage diagnosis and a later-stage one can mean the difference between lumpectomy and mastectomy, or between avoiding chemotherapy and enduring its harsh side effects.
Key Findings: Promise and Pitfalls
While the overall reduction in interval cancers was a major highlight, the study also provided a nuanced view of AI’s capabilities, revealing both its strengths and current limitations.
Dr. Hannah Milch, assistant professor of Radiology at the David Geffen School of Medicine and senior author of the study, highlighted these complexities: "While we had some exciting results, we also uncovered a lot of AI inaccuracy and issues that need to be further explored in real-world settings." One particularly interesting observation pertained to occult cancers—those truly invisible to the human eye on mammography. Despite their invisibility, the AI tool surprisingly flagged 69% of the screening mammograms associated with these occult cancers as suspicious. However, the AI’s ability to precisely pinpoint the cancerous area within these flagged images was less accurate, marking the actual cancer location only 22% of the time. This suggests that while AI can detect subtle patterns indicative of malignancy even in seemingly clear images, its spatial localization capabilities for extremely subtle lesions still require refinement.
Navigating the Differences: U.S. vs. European Screening Paradigms
A crucial aspect that sets this UCLA study apart is its focus on the U.S. healthcare context. While similar research exploring AI’s role in detecting interval breast cancers has been conducted in Europe, the distinct differences in screening practices between the two regions make the U.S.-based findings particularly relevant.
In the United States, the majority of mammograms are performed using digital breast tomosynthesis (DBT), or 3D mammography. DBT offers a significant advantage over traditional 2D digital mammography (DM) by providing a series of thin, high-resolution images, allowing radiologists to view breast tissue in "slices" and reduce the obscuring effects of overlapping tissue. This technology is particularly beneficial for women with dense breast tissue, who are at higher risk of having cancers missed on 2D mammograms. Furthermore, U.S. patients are typically screened annually.
In contrast, European screening programs have historically relied more heavily on 2D digital mammography (DM), and screening intervals are generally longer, often every two to three years. These differences in technology and frequency directly impact the types of cancers detected and the prevalence of interval cancers. The UCLA study’s use of a dataset predominantly featuring DBT, coupled with annual screening data, provides valuable insights into how AI would perform within the specific parameters of the U.S. system, making its conclusions more directly applicable to American clinical practice.
The Broader Landscape of AI in Medical Imaging
The integration of AI into medical imaging is part of a larger global trend toward leveraging advanced computational power to assist healthcare professionals. AI algorithms, particularly those based on deep learning, are proving adept at pattern recognition in vast image datasets, often identifying features that are too subtle or complex for the human eye to consistently discern. Beyond breast cancer, AI is being explored for its potential in detecting lung nodules, identifying diabetic retinopathy, diagnosing skin cancers, and even predicting disease progression in various conditions.
The promise of AI lies not in replacing human expertise but in augmenting it. Radiologists face immense pressure, interpreting hundreds of images daily, often under time constraints. AI tools are envisioned as "second readers" or intelligent assistants that can highlight suspicious areas, prioritize cases, and reduce the cognitive load on human experts, thereby improving diagnostic accuracy and efficiency.
Implications for Clinical Practice and Patient Outcomes
The UCLA study’s findings carry significant implications for the future of breast cancer screening, potentially reshaping clinical practices and profoundly impacting patient outcomes.
Enhancing Radiologist Capabilities
The most immediate implication is the potential for AI to serve as a powerful adjunct to radiologists. By automatically analyzing mammograms and flagging areas of concern, AI could act as a sophisticated safety net, drawing attention to subtle abnormalities that might otherwise be missed. This is particularly valuable for the "minimal signs" category of interval cancers, where the signs are present but faint. AI could help standardize detection across different radiologists, potentially reducing variability in interpretation. Furthermore, it could aid in managing the increasing volume of screenings, allowing radiologists to focus their expertise on the most challenging cases.
Potential for Improved Patient Care and Reduced Treatment Burden
For patients, the prospect of earlier detection of interval cancers is transformative. Catching cancer at an earlier stage typically means smaller tumors, less lymph node involvement, and a higher likelihood of successful treatment with less invasive procedures. This could translate to fewer mastectomies, less intensive chemotherapy regimens, and shorter recovery times, dramatically improving a patient’s quality of life during and after treatment. The reduction in the number of interval cancers by 30% projected by the study represents a tangible step towards achieving this goal, potentially saving countless lives and mitigating the long-term morbidity associated with advanced cancer treatments.
Economic and Healthcare System Impact
The economic implications of earlier detection are also substantial. Treating early-stage breast cancer is generally less costly than treating advanced, metastatic disease, which often requires a complex and prolonged regimen of therapies. A widespread reduction in interval cancers could lead to significant cost savings for healthcare systems, freeing up resources that can be reallocated to other areas of patient care or prevention. Moreover, by improving patient outcomes, AI-assisted screening could reduce the societal burden of cancer, including lost productivity and caregiving costs.
Addressing the Nuances: Challenges and Future Directions
Despite the exciting promise, the UCLA researchers are clear that AI is not a panacea and its integration into clinical practice requires careful consideration of its current limitations and the need for further research.
The Imperfection of AI: Accuracy and Localization
As Dr. Milch noted, AI isn’t perfect. The finding that AI flagged 69% of occult cancers but only pinpointed the exact location 22% of the time highlights a critical challenge: AI’s ability to identify suspicious patterns versus its precision in localization. This discrepancy can create dilemmas in clinical settings. If AI flags a mammogram as suspicious but radiologists cannot visually confirm an abnormality, what is the appropriate next step? Further imaging or biopsy could lead to increased patient anxiety and potentially unnecessary procedures (false positives), while ignoring the AI’s warning could mean missing a genuine cancer. This challenge underscores the need for continuous improvement in AI algorithms, particularly in their ability to provide clinically actionable spatial information.
The Need for Prospective Studies
The current study was retrospective, meaning it looked back at existing data. While valuable for identifying potential, prospective studies are now essential. These studies would involve integrating AI into real-time screening workflows and observing how radiologists interact with the AI tool, how it influences their decision-making, and what the actual impact on interval cancer rates is. Such studies would also help address key practical questions, such as optimal AI integration points, necessary radiologist training, and the management of AI-generated flags that are not immediately apparent to human interpretation.
Ethical and Practical Considerations for AI Integration
The widespread adoption of AI in diagnostics also raises several ethical and practical considerations. Questions of accountability (who is responsible if an AI-assisted diagnosis is missed?), potential biases in AI algorithms (if trained on unrepresentative datasets), data privacy, and the psychological impact on patients (e.g., anxiety from AI-flagged "suspicious" areas that turn out to be benign) must be thoroughly addressed. Regulatory bodies, such as the FDA in the U.S., are actively developing frameworks for approving and monitoring AI-powered medical devices, recognizing the need for robust validation and ongoing oversight.
Dr. Yu’s concluding remarks encapsulate the balanced perspective of the research team: "While AI isn’t perfect and shouldn’t be used on its own, these findings support the idea that AI could help shift interval breast cancers toward mostly true interval cancers. It shows potential to serve as a valuable second set of eyes, especially for the types of cancers that are the hardest to catch early. This is about giving radiologists better tools and giving patients the best chance at catching cancer early, which could lead to more lives saved." This vision positions AI not as a replacement, but as an indispensable partner in the ongoing fight against breast cancer.
Expert Perspectives and Collaborative Efforts
The comprehensive nature of this study is a testament to the collaborative spirit within the UCLA Health Jonsson Comprehensive Cancer Center. In addition to Dr. Yu and Dr. Milch, other distinguished authors from UCLA who contributed to this significant work include Dr. Anne Hoyt, Dr. Melissa Joines, Dr. Cheryce Fischer, Dr. Nazanin Yaghmai, Dr. James Chalfant, Dr. Lucy Chow, Dr. Shabnam Mortazavi, Christopher Sears, Dr. James Sayre, Dr. Joann Elmore, and Dr. William Hsu. This multidisciplinary team brought together expertise in radiology, oncology, data science, and clinical research, underscoring the complex and collaborative nature of cutting-edge medical investigations.
Funding and Support for Breakthrough Research
Research of this magnitude requires substantial support. The work was made possible, in part, by generous contributions and grants from several key organizations. These include the National Institutes of Health (NIH), a primary agency of the U.S. government responsible for biomedical and public health research; the National Cancer Institute (NCI), a component of the NIH and the principal agency for cancer research and training; the Agency for Healthcare Research and Quality (AHRQ), which works to improve the quality, safety, efficiency, and effectiveness of healthcare for all Americans; and Early Diagnostics Inc. Such funding is crucial for driving innovative research that has the potential to translate into real-world improvements in patient care and public health.
As AI continues to evolve, its role in refining breast cancer screening and detection promises to grow, offering a future where interval cancers become a rarity, and early, effective treatment is the standard for all.
