By Billy Candra 
The fight against breast cancer has seen remarkable progress over decades, largely thanks to advancements in screening technologies and therapeutic interventions. Mammography remains the cornerstone of early detection, capable of identifying tumors often before they are palpable. However, a significant challenge persists in the form of "interval cancers" – malignancies that emerge and are diagnosed between scheduled mammography screenings. These cancers are particularly insidious because they are often more aggressive, grow rapidly, and are associated with a poorer prognosis due to their later detection. A groundbreaking study from the UCLA Health Jonsson Comprehensive Cancer Center, published in the esteemed Journal of the National Cancer Institute, now presents compelling evidence that artificial intelligence (AI) could revolutionize the detection of these elusive tumors, potentially saving lives and mitigating the severity of treatment.
Understanding the Enigma of Interval Cancers
Interval breast cancers represent a critical blind spot in current screening protocols. They account for a significant proportion, estimated between 20% and 30%, of all breast cancer diagnoses. Their late detection, often when they manifest as a palpable lump or other symptoms, means they are typically more advanced than screen-detected cancers, with a higher likelihood of lymph node involvement and distant metastasis. This translates directly to more aggressive treatment regimens, including extensive surgery, chemotherapy, and radiation, and crucially, a lower five-year survival rate compared to cancers found during routine screenings.
There are several reasons why interval cancers go undetected. Some are "true interval cancers," meaning they are genuinely new growths that develop rapidly after a clear mammogram. Others, however, are present on the prior mammogram but are either missed by the interpreting radiologist ("missed reading error"), or present with such subtle signs that they are arguably below the threshold of human perception ("minimal signs"). A small fraction might also be "occult," meaning they are truly invisible on mammograms, or "missed due to technical error" during image acquisition or processing. The UCLA study specifically targeted the "mammographically-visible" types of interval cancers – those that, in hindsight, showed some indication on prior images but were overlooked.
The UCLA Study: A Deep Dive into AI’s Diagnostic Prowess
The research, spearheaded by Dr. Tiffany Yu, an assistant professor of Radiology at the David Geffen School of Medicine at UCLA and the study’s first author, and Dr. Hannah Milch, assistant professor of Radiology and senior author, represents a pivotal moment in the application of AI in medical diagnostics within the United States. While similar investigations have been conducted in Europe, the UCLA study is among the first to explore AI’s utility in detecting interval breast cancers within the unique context of American screening practices.
The methodology employed was a retrospective analysis, a critical approach that allowed researchers to examine past screening mammograms with the benefit of hindsight. The study meticulously reviewed data from nearly 185,000 past mammograms performed between 2010 and 2019. From this extensive dataset, the team focused on 148 cases where a woman was subsequently diagnosed with interval breast cancer. This detailed examination allowed radiologists to re-evaluate these cases and determine precisely why the cancer was not initially identified. To categorize these interval cancers, the researchers adapted a classification system widely used in European studies, encompassing categories such as Missed reading error, minimal signs-actionable, minimal signs-non-actionable, true interval cancer, occult, and missed due to a technical error. This systematic categorization provided a robust framework for assessing AI’s performance against established clinical realities.
AI as a "Second Set of Eyes": The Transpara Integration
The core of the UCLA study involved applying a commercially available AI software, Transpara, to the initial screening mammograms that preceded the interval cancer diagnoses. Transpara, developed by ScreenPoint Medical, is an AI-powered CAD (Computer-Aided Detection) system designed to assist radiologists by analyzing mammographic images for suspicious areas. The tool assigns a risk score to each mammogram, ranging from 1 to 10, indicating the likelihood of cancer. A score of 8 or higher was designated as a "flagged" mammogram, signaling a potentially concerning finding that warranted closer scrutiny.
The results were compelling. The AI software demonstrated a remarkable ability to identify "mammographically-visible" types of interval cancers earlier by flagging them at the time of screening. These included tumors with subtle signs that were either missed by human radiologists or presented with such faint indications that they fell below the human detection threshold. Researchers conservatively estimate that integrating AI into the standard screening workflow could potentially reduce the incidence of interval breast cancers by a significant 30%.
"This finding is important because these interval cancer types could be caught earlier when the cancer is easier to treat," emphasized Dr. Yu. "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." This sentiment underscores the profound impact early detection has on patient prognosis and quality of life. The ability to intervene when a tumor is smaller and localized often means less invasive surgery, fewer rounds of chemotherapy, and a higher probability of long-term survival.
Distinguishing U.S. and European Screening Paradigms
A crucial aspect of the UCLA study’s significance lies in its focus on the U.S. screening landscape, which differs markedly from European practices. In the United States, Digital Breast Tomosynthesis (DBT), commonly known as 3D mammography, has become the predominant screening modality. DBT offers a series of thin-slice images of the breast, effectively reducing tissue overlap and improving cancer detection rates compared to traditional 2D digital mammography (DM). Furthermore, U.S. guidelines typically recommend annual screenings for women at average risk.
Conversely, European screening programs have historically relied more on 2D digital mammography (DM), and screening intervals are often longer, typically every two to three years. These differences in technology and screening frequency mean that findings from European AI studies, while valuable, may not be directly transferable to the U.S. context. The UCLA study, by utilizing data that included both DM and DBT, and considering the annual screening cycle, provides critically relevant insights for the American healthcare system, paving the way for more tailored AI integration strategies.
Nuances and Limitations: The Path to Perfection
Despite the exciting potential, the study also provided a realistic assessment of AI’s current limitations. Dr. Hannah Milch highlighted that while the results were promising, they "uncovered a lot of AI inaccuracy and issues that need to be further explored in real-world settings." This candid acknowledgement is vital for the responsible development and deployment of AI in medicine.
One notable finding concerned occult cancers – those truly invisible on mammography. The AI tool surprisingly flagged 69% of the screening mammograms that had occult cancers. However, when researchers delved deeper to examine the specific areas on the images that the AI marked as suspicious, the accuracy dropped significantly, with the AI pinpointing the actual cancer location only 22% of the time. This suggests that while AI might detect subtle, non-specific changes that correlate with an occult cancer being present somewhere in the breast, it struggles to precisely localize these deeply hidden lesions. This distinction is crucial because for AI to be clinically actionable, it needs to not only detect an anomaly but also guide the radiologist to its precise location for further investigation.
These findings underscore that AI, in its current iteration, is not a perfect diagnostic tool and should not be used in isolation. Instead, its strength lies in its potential as an assistive technology. "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," Dr. Yu explained. "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."
Broader Implications and Future Directions
The implications of this study are far-reaching, touching upon various facets of breast cancer screening and care.
- Enhanced Radiologist Performance: AI’s ability to act as a "second reader" could significantly reduce radiologist fatigue and improve detection rates, particularly for subtle, easily missed lesions. This doesn’t suggest AI replacing radiologists, but rather augmenting their capabilities, allowing them to focus their expert attention on the most challenging cases flagged by the AI.
- Improved Patient Outcomes and Reduced Treatment Burden: Earlier detection of interval cancers means smaller tumors, less advanced disease, and consequently, less aggressive and invasive treatments. This translates directly to improved survival rates, fewer side effects, and a better quality of life for patients.
- Evolution of Screening Guidelines: As AI technology matures and its efficacy is further validated through larger prospective studies, professional organizations like the American College of Radiology may begin to incorporate AI into updated screening guidelines, potentially recommending its use as an adjunct tool.
- Economic Impact: While the initial investment in AI software might be a consideration, the long-term economic benefits from earlier detection are substantial. Less advanced cancers require less costly and complex treatments, reducing healthcare expenditures associated with advanced disease management.
- Ethical Considerations and Trust: The integration of AI also raises ethical questions about accountability, data privacy, and potential biases in AI algorithms. Establishing clear regulatory frameworks and ensuring transparency in AI’s decision-making process will be paramount to building trust among patients and clinicians.
- Need for Prospective Studies: The retrospective nature of the UCLA study, while powerful for identifying missed cases, needs to be followed by larger prospective studies. These "real-world" studies are crucial to understand how radiologists would practically integrate AI into their workflows, how AI’s flags would influence their diagnostic decisions, and how to manage cases where AI identifies suspicious areas invisible to the human eye, especially when the AI’s pinpointing accuracy is not yet perfect. Such studies will help refine AI algorithms, establish best practices, and develop robust protocols for handling equivocal AI findings.
The collaborative nature of this research is also noteworthy, with other contributing authors from UCLA including 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 interdisciplinary approach, pooling expertise from various fields, is characteristic of cutting-edge medical research. The work received crucial support from prominent organizations, including the National Institutes of Health, the National Cancer Institute, the Agency for Healthcare Research and Quality, and Early Diagnostics Inc., underscoring the national importance and collaborative investment in advancing cancer detection technologies.
In conclusion, the UCLA Health Jonsson Comprehensive Cancer Center study provides compelling evidence for the transformative potential of artificial intelligence in addressing one of the most challenging aspects of breast cancer screening: the detection of interval cancers. While AI is not yet a standalone solution and requires further refinement and validation, its demonstrated ability to act as an intelligent "second set of eyes" for radiologists marks a significant leap forward. This research not only promises to enhance diagnostic accuracy and improve patient outcomes but also signals a new era where human expertise and advanced AI capabilities converge to offer a more precise, proactive, and ultimately, more life-saving approach to breast cancer detection. The journey towards fully integrating AI into clinical practice is ongoing, but studies like this illuminate a promising path forward in the relentless pursuit of conquering cancer.
