← Back to Home

How Are Cancer Cells Arranged, and How Far Away Are T Cells? AI Pathology Studies Seek Clues to Treatment Response

Using ovarian cancer slides and bladder cancer trial specimens, Lunit explores links between tissue morphology, immune cell location, and cancer biology. The findings offer directions for biomarker research, but further validation is needed before they can inform patient treatment decisions.

By SURL BioNews

A tumor slide can reveal not only cell morphology but also the distances between immune cells and cancer cells. Could these spatial clues help explain cancer, or even identify which patients may respond to immunotherapy? Lunit has announced two AI pathology studies seeking answers in ovarian and bladder cancer specimens. The findings so far show associations between tissue features, disease, and treatment response.

Both studies are scheduled for poster presentation on November 7 at the Society for Immunotherapy of Cancer annual meeting (SITC 2026) in Phoenix, United States. Lunit’s October announcement and reporting by The Asia Business Daily had confirmed that the studies were conducted in collaboration with Labcorp and CellCarta, respectively; materials released on November 4 disclosed further analytical results.

The ovarian cancer study compared tumors with homologous recombination deficiency (HRD) and those that were homologous recombination proficient (HRP). HRD involves cells’ ability to repair DNA damage and is also an important biomarker for PARP inhibitor treatment. The study sought to determine whether this molecular difference also leaves measurable traces on routine pathology slides.

The team used Lunit SCOPE IO to analyze 339 ovarian cancer whole-slide images, comprising 88 HRD and 251 HRP cases, and quantified 22 immune and morphological features. According to the company, mitotic cell density in the HRD group was approximately 1.5 times that in the HRP group (p = 0.002), and tumor cell density was also higher. These findings indicate tissue differences associated with cell proliferation, but they do not establish that AI can replace molecular testing.

The bladder cancer study used tissue specimens from the phase 2 ABACUS trial to investigate responses in patients with muscle-invasive bladder cancer who received the anti-PD-L1 drug atezolizumab before surgery. The team combined multiplex immunohistochemistry with spatial analysis, while CellCarta pathologists and SCOPE IO assessed tertiary lymphoid structures in the slides, placing AI interpretation among several analytical methods.

According to the company’s disclosures, responders had greater immune cell infiltration, and their T cells were closer to tumor cells; the proximity of some tertiary lymphoid structures to tumor cell nests showed only a trend. The study therefore suggests that, beyond counting immune cells, how they cluster within tissue and how far apart they are may also offer clues to understanding treatment response.

However, the press release did not specify the sample size for this bladder cancer analysis or provide accuracy measures for clinical prediction or independent validation results. Whether these associations can be reproduced in other patient populations and whether they can improve treatment selection remain questions for further research. Lunit also explicitly states that SCOPE products are currently for research use only and must not be used in diagnostic procedures or decisions.

References

  1. Lunit / PR Newswire
  2. Lunit
  3. The Asia Business Daily