In oncology, the most critical decisions often hinge on a millimeter. A tumor's size, its volume, the wasting of muscle around it,these are the metrics that determine treatment plans, gauge drug efficacy, and predict survival. Measuring them manually from a CT scan is a painstaking, time-consuming task for radiologists, prone to human error and inconsistency. Nucleo, a startup from Y Combinator's Fall 2025 batch, is betting that an AI agent can do it faster, more consistently, and with a level of precision that could reshape clinical workflows [Y Combinator, Unknown].
The Clinical Workflow Wedge
Nucleo's software is designed to ingest a standard CT scan and automatically perform a suite of analyses that are currently manual, clerical burdens. Its core functions target three areas critical in cancer care: body composition analysis for conditions like sarcopenia (muscle wasting), precise tumor sizing and volumetric calculations, and the classification of target versus non-target lesions [Nucleo, Unknown]. The company claims its system can segment these features 2,500 times faster than a human, with 98% agreement to expert radiologist assessments [Y Combinator, Unknown]. For an oncologist tracking a patient's response to therapy across multiple scans, that speed could translate from days of waiting for a formal read to near-instantaneous quantitative feedback.
The startup's early validation appears to be happening within prestigious academic medical centers, a common and prudent path for clinical AI. According to a Forbes roundup of YC's top startups, Nucleo is working with leading hospitals including Stanford, Cedars-Sinai, and Weill Cornell [Forbes, Unknown]. These are not commercial deployments, but rather research collaborations or pilot studies essential for refining the algorithm and gathering the real-world evidence needed for any future regulatory submission. The company has disclosed a $4 million seed round to fund this development phase, though its investors and founding team remain unnamed in the public record [Y Combinator, Unknown].
An Honest Counterfactual on the Path to the Clinic
The ambition is clear, but the path from academic pilot to adopted clinical tool is fraught with hurdles that have tripped up many promising AI startups. The market for radiology AI is crowded, with numerous companies offering tools for specific organs or diseases. Nucleo's focus on a broader oncology workflow is its differentiator, but it also means competing for budget and attention within complex hospital IT systems.
- The regulatory gate. For software that provides diagnostic information, even as an aid, the FDA's 510(k) or De Novo classification is a significant milestone. Nucleo has not announced any regulatory filings, which suggests its current hospital work is firmly in the research category.
- The clinical evidence gap. A 98% agreement rate in a controlled study is one thing; proving that the software improves patient outcomes or clinician efficiency in a busy hospital is another. That proof requires large, prospective trials.
- The integration burden. The most accurate AI is useless if it doesn't fit seamlessly into a radiologist's existing PACS (Picture Archiving and Communication System) workflow. Deep integration with major vendors like Epic or Cerner is a long, costly endeavor.
Nucleo's most plausible answer to these challenges is its reported hospital partnerships. By embedding its research within institutions like Stanford, it can build the necessary evidence and refine its product in the exact environment where it must eventually function. Success will be measured not by algorithm accuracy alone, but by whether those pilot programs evolve into paid, enterprise-wide contracts.
For patients with solid tumors, from lung cancer to colorectal cancer, the standard of care today involves a radiologist manually measuring lesions on a CT scan with digital calipers, a process that can take 30 minutes or more per case and introduces variability. Sarcopenia, a powerful predictor of chemotherapy toxicity and survival, is rarely routinely assessed because measuring muscle mass is so labor-intensive. Nucleo's bet is that automating this quantitative drudgery will free clinicians to focus on interpretation and patient care, while providing more consistent, trackable data to guide therapy. The next 12 months will be telling; the transition from a research project cited in a YC demo day to a product with published clinical validation and a clear commercial motion is the steep climb ahead.
Sources
- [Forbes, Unknown] The Top Startups To Watch From Y Combinator’s Fall 2025 Batch | https://www.forbes.com/sites/dariashunina/2025/11/13/the-top-startups-to-watch-from-y-combinators-fall-2025-batch/
- [Nucleo, Unknown] Nucleo | https://nucleoresearch.com
- [Y Combinator, Unknown] Nucleo | Y Combinator | https://www.ycombinator.com/companies/nucleo