Background The backbone chemotherapy of first-line standard of care (SOC) for microsatellite stable (MSS) metastatic colorectal cancer (mCRC) combines 5-Fluorouracil to oxaliplatin and/or irinotecan. There are no biomarkers to predict response, which is complete or long-lasting (CR/LLR) in ∼25% of patients, while ∼15% are primary refractory. Our study aims at developing an accurate response predictor via a pathomics model based on Hematoxylin & Eosin (H&E)-stained digital slides (WSI). Methods We trained an unsupervised bag-of-words artificial intelligence (AI)-based model on 62 patients classified as response outliers as follow: i) super-sensitive if they achieved CR or LLR >10 months to any SOC (N=20); ii) refractory if progression occurred at first disease reassessment (N=42). WSIs of the resected primary tumors were tiled into patches of 224x224 pixel (0.5μm/pixel). First-order and texture features were subsequently extracted from all tumoral tiles, and grouped into homogenous clusters through a 3x3 self-organized map. For each patient, the percentage of tiles belonging to each tiles’ cluster was computed and used by a dendrogram to create clusters of similar patients. Patients’ survival was calculated by landmark analysis logrank test. Main clinicopathological features were matched to treatment response by Fisher’s exact test. Results Primary resistant patients had significantly worse survival [logrank HR 4.2 95% CI (2.7-7.3)]. While outliers showed similar clinicopathological features (including stage, RAS/BRAF status, histotype and sidedness), pathomics signatures were significantly associated to primary resistance (p=.005). According to the unsupervised clustering, we classified as resistant patients belonging to four out of eight clusters obtaining an 88% (23/26) negative predictive value. We observed that patients classified as belonging to one of the four resistant clusters were predominantly characterized by a specific pattern of patches. Conclusions We demonstrated that a pathomics signature has the potential to predict resistance to SOC in MSS mCRC. Validation of these preliminary findings on a further cohort of 125 patients is ongoing.
223P Artificial intelligence-based pathomics biomarker predict primary resistance to first-line treatment in metastatic colorectal cancers
Mauri, G.;Giannini, V.;Nicoletti, G.;Lazzari, L.;Berrino, E.;Arena, S.;Bardelli, A.;Regge, D.;
2023-01-01
Abstract
Background The backbone chemotherapy of first-line standard of care (SOC) for microsatellite stable (MSS) metastatic colorectal cancer (mCRC) combines 5-Fluorouracil to oxaliplatin and/or irinotecan. There are no biomarkers to predict response, which is complete or long-lasting (CR/LLR) in ∼25% of patients, while ∼15% are primary refractory. Our study aims at developing an accurate response predictor via a pathomics model based on Hematoxylin & Eosin (H&E)-stained digital slides (WSI). Methods We trained an unsupervised bag-of-words artificial intelligence (AI)-based model on 62 patients classified as response outliers as follow: i) super-sensitive if they achieved CR or LLR >10 months to any SOC (N=20); ii) refractory if progression occurred at first disease reassessment (N=42). WSIs of the resected primary tumors were tiled into patches of 224x224 pixel (0.5μm/pixel). First-order and texture features were subsequently extracted from all tumoral tiles, and grouped into homogenous clusters through a 3x3 self-organized map. For each patient, the percentage of tiles belonging to each tiles’ cluster was computed and used by a dendrogram to create clusters of similar patients. Patients’ survival was calculated by landmark analysis logrank test. Main clinicopathological features were matched to treatment response by Fisher’s exact test. Results Primary resistant patients had significantly worse survival [logrank HR 4.2 95% CI (2.7-7.3)]. While outliers showed similar clinicopathological features (including stage, RAS/BRAF status, histotype and sidedness), pathomics signatures were significantly associated to primary resistance (p=.005). According to the unsupervised clustering, we classified as resistant patients belonging to four out of eight clusters obtaining an 88% (23/26) negative predictive value. We observed that patients classified as belonging to one of the four resistant clusters were predominantly characterized by a specific pattern of patches. Conclusions We demonstrated that a pathomics signature has the potential to predict resistance to SOC in MSS mCRC. Validation of these preliminary findings on a further cohort of 125 patients is ongoing.File | Dimensione | Formato | |
---|---|---|---|
PIIS0923753423037833.pdf
Accesso riservato
Dimensione
137.63 kB
Formato
Adobe PDF
|
137.63 kB | Adobe PDF | Visualizza/Apri Richiedi una copia |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.