Artificial intelligence (AI) was used to predict response of high-grade extrauterine serous carcinoma (HGSC) to adjuvant platinum chemotherapy.
For the first time, researchers used artificial intelligence (AI) to predict response of high-grade extrauterine serous carcinoma (HGSC), a highly aggressive cancer with frequent chemotherapy resistance and a poor survival rate, to adjuvant platinum chemotherapy.
As explained in Scientific Reports, there are currently no validated tissue-based prognostic or predictive markers for primary platinum-based treatment in use for HGSC, a type of ovarian cancer; about 75% of ovarian cancer deaths are from HGSC. Typically presenting at stage III or stage IV, the 5-year survival rate for HGSC is less than 50% and about 25% in advanced disease.
These tumors are marked by genetic and morphologic heterogeneity and outcomes are uncertain, even when patients appear to have favorable characteristics, such as younger age and lower stage. About 15% of patients can survive for a decade or more. Standard treatment (debulking surgery and platinum-based chemotherapy) hasn’t changed for this disease in decades, and despite the arrival of poly ADP ribose polymerase inhibitors, treatment response is still unpredictable.
Patients with a progression-free interval (PFI) of less than 6 months are considered platinum resistant, while those with a PFI beyond 12 months are considered platinum sensitive; only 25% of women have a PFI of greater than 18 months.
The researchers, from Finland, said they believed the differences that exist in HGSC tumors—which ones are refractory to therapy immediately and which ones have later resistance—could be detectable using AI. The study aimed to see if a weakly supervised convolutional neural network trained on whole slide images (WSI) can accurately classify HGSC into outcome groups using the tumor morphology on the images. WSI refer to conventional slides that are scanned to create digital files.
The HGSC patient cohort (the test set) included 30 women with similar presentations who experienced very different treatment responses (PFIs of either ≤ 6 months or ≥ 18 months and 205 slide images).
In 3 steps, the neural network was trained to identify morphologic regions (digital biomarkers) that are highly associated with one or the other treatment response group. The classifier was tested using a separate 22-slide test set, and 18 of 22 slides were correctly classified.
The method was able to discern extremes in patient response to primary platinum-based chemotherapy with high sensitivity (73%) and specificity (91%).
The authors said the proof-of-concept results are novel since it is the first time prospective prognostic information is identified specifically within HGSC tumor morphology. The finding is important since unlike other tumor types, “pathologist evaluation of HGSC tumor tissue provides almost no predictive or prognostic information beyond the diagnosis itself,” the authors said.
Reference
Laury AR, Blom S, Ropponen T, Virtanen A, Carpén OM. Artificial intelligence-based image analysis can predict outcome in high-grade serous carcinoma via histology alone. Sci Rep. Published online September 27, 2021. doi:10.1038/s41598-021-98480-0
Standard Criteria for Loss of Ambulation Needed in DMD
April 19th 2024A recent study suggests the differences between ambulation definitions for patients with Duchenne muscular dystrophy (DMD) can impact the identification of ambulant vs nonambulant individuals, and standard criteria across settings are needed.
Read More
Navigating Health Literacy, Social Determinants, and Discrimination in National Health Plans
February 13th 2024On this episode of Managed Care Cast, we're talking with the authors of a study published in the February 2024 issue of The American Journal of Managed Care® about their findings on how health plans can screen for health literacy, social determinants of health, and perceived health care discrimination.
Listen
Early Involvement Critical in Treating Immunotherapy-Induced Overlap Syndrome
April 19th 2024A series of case studies reveals the importance of early diagnosis and involvement of special teams of clinicians when dealing with potential cases of overlap syndrome, which encompasses myocarditis, myasthenia gravis, and immune checkpoint inhibitor–related myositis.
Read More
Drs Raymond Thertulien, Joseph Mikhael on Racial Disparities in Multiple Myeloma Care Access
December 28th 2023In the wake of the 2023 American Society of Hematology Annual Meeting and Exposition, Raymond Thertulien, MD, PhD, of Novant Health, and Joseph Mikhael, MD, MEd, FRCPC, FACP, chief medical officer of the International Myeloma Foundation, discussed health equity research highlights from the meeting and drivers of racial disparities in multiple myeloma outcomes.
Listen
Government agencies have created an online portal for the public to report potential anticompetitive practices in health care; there are changes coming to the “boxed warning” section for chimeric antigen receptor T-cell therapies (CAR T) to highlight T-cell blood cancer risk; questions about the safety of obesity medications during pregnancy have arisen in women on them who previously struggled with fertility issues.
Read More