From counting to clustering: AI-defined “diffuse” vs. “focal” spatial patterns as the new precision biomarker for triple-negative breast cancer

The integration of immunotherapy into the neoadjuvant treatment has transformed the clinical management of triple-negative breast cancer (TNBC). However, the current "one-size-fits-all" approach to patient selection remains suboptimal. While the Nature Medicine forecasts for 2025 rightly emphasized the expansion of precision trials [1], clinical practice is still heavily reliant on quantifying the abundance of tumor-infiltrating lymphocytes (TILs) or PD-L1 expression (or Combined Positive Score, CPS). We write to highlight a pivotal shift emerging from late 2025 literature: the transition from "counting" immune cells to decoding their spatial clustering patterns. A groundbreaking multicenter study published in December 2025 has provided the first concrete evidence that the spatial architecture of TILs dictates therapeutic response in TNBC, independent of total cell count [2]. Utilizing an artificial intelligence (AI)-based spatial clustering classifier, the investigators identified two distinct immune landscapes: a "Diffuse Immune Subtype" and a "Focal Hotspot Subtype".Traditionally, a tumor with high TIL density is considered "immune-hot" and predicted to respond well to therapy. However, this new AI-driven analysis reveals a critical nuance. Patients with the "Focal Hotspot" subtype-characterized by dense but isolated clusters of immune cells-often exhibit poorer pathological complete response (pCR) rates compared to those with "Diffuse" infiltration, where immune cells are evenly spread throughout the tumor stroma [2]. This finding helps explain a persistent clinical paradox: why some patients with seemingly "high-TIL" tumors fail to respond to immune checkpoint inhibitors (ICIs). The AI analysis suggests that "Focal" patterns may represent an ineffective, contained immune response (potentially restricted by fibrosis or tertiary lymphoid structure dysregulation), whereas "Diffuse" patterns indicate a successful, pervasive anti-tumor engagement.The urgency of this shift was echoed in recent discussions at SABCS 2025 regarding the APHINITY trial analysis [3]. Although the SABCS data focused on HER2-positive disease, the conference consensus underscored a universal truth: manual scoring is reaching its limit. For TNBC specifically, integrating the AI-defined "Diffuse" vs. "Focal" metrics is no longer a futuristic concept but a tangible necessity to refine clinical workflows:1. Refining Prognosis: The "Diffuse" signature serves as an independent predictor of prolonged disease-free survival (DFS), offering a more granular risk stratification than current AJCC staging [2]. 2. Stratifying HER2-Low TNBC: The study further elucidated that the "Diffuse" pattern correlates with better outcomes specifically in the HER2-low TNBC subgroup, a population that is increasingly becoming a distinct therapeutic target [2].