Deep learning predicts patients outcome and mutations from digitized histology slides in gastrointestinal stromal tumor.

Risk assessment of gastrointestinal stromal tumor (GIST) according to the AFIP/Miettinen classification and mutational profiling are major tools for patient management. However, the AFIP/Miettinen classification depends heavily on mitotic counts, which is laborious and sometimes inconsistent between pathologists. It has also been shown to be imperfect in stratifying patients. Molecular testing is costly and time-consuming, therefore, not systematically performed in all countries. New methods to improve risk and molecular predictions are hence crucial to improve the tailoring of adjuvant therapy. We have built deep learning (DL) models on digitized HES-stained whole slide images (WSI) to predict patients' outcome and mutations. Models were trained with a cohort of 1233 GIST and validated on an independent cohort of 286 GIST. DL models yielded comparable results to the Miettinen classification for relapse-free-survival prediction in localized GIST without adjuvant Imatinib (C-index=0.83 in cross-validation and 0.72 for independent testing). DL splitted Miettinen intermediate risk GIST into high/low-risk groups (p value = 0.002 in the training set and p value = 0.29 in the testing set). DL models achieved an area under the receiver operating characteristic curve (AUC) of 0.81, 0.91, and 0.71 for predicting mutations in KIT, PDGFRA and wild type, respectively, in cross-validation and 0.76, 0.90, and 0.55 in independent testing. Notably, PDGFRA exon18 D842V mutation, which is resistant to Imatinib, was predicted with an AUC of 0.87 and 0.90 in cross-validation and independent testing, respectively. Additionally, novel histological criteria predictive of patients' outcome and mutations were identified by reviewing the tiles selected by the models. As a proof of concept, our study showed the possibility of implementing DL with digitized WSI and may represent a reproducible way to improve tailoring therapy and precision medicine for patients with GIST.

Nationwide incidence of sarcomas and connective tissue tumors of intermediate malignancy over four years using an expert pathology review network.

BACKGROUND: Since 2010, nationwide networks of reference centers for sarcomas (RREPS/NETSARC/RESOS) collected and prospectively reviewed all cases of sarcomas and connective tumors of intermediate malignancy (TIM) in France. METHODS: The nationwide incidence of sarcoma or TIM (2013-2016) was measured using the 2013 WHO classification and confirmed by a second independent review by expert pathologists. Simple clinical characteristics, yearly variations and correlation of incidence with published clinical trials are presented and analyzed. RESULTS: Over 150 different histological subtypes are reported from the 25172 patients with sarcomas (n = 18712, 74,3%) or TIM (n = 6460, 25.7%), with n = 5838, n = 6153, n = 6654, and n = 6527 yearly cases from 2013 to 2016. Over these 4 years, the yearly incidence of sarcomas and TIM was therefore 70.7 and 24.4 respectively, with a combined incidence of 95.1/106/year, higher than previously reported. GIST, liposarcoma, leiomyosarcomas, undifferentiated sarcomas represented 13%, 13%, 11% and 11% of tumors. Only GIST, as a single entity had a yearly incidence above 10/106/year. There were respectively 30, 64 and 66 different histological subtypes of sarcomas or TIM with an incidence ranging from 10 to 1/106, 1-0.1/106, or < 0.1/106/year respectively. The 2 latter incidence groups represented 21% of the patients with 130 histotypes. Published phase III and phase II clinical trials (p<10-6) are significantly higher with sarcomas subtypes with an incidence above 1/106 per. CONCLUSIONS: This nationwide registry of sarcoma patients, with exhaustive histology review by sarcoma experts, shows that the incidence of sarcoma and TIM is higher than reported, and that tumors with a very low incidence (1<106/year) are less likely to be included in clinical trials.

Lower Rate of CTNNB1 Mutations and Higher Rate of APC Mutations in Desmoid Fibromatosis of the Breast: A Series of 134 Tumors.

Desmoid fibromatosis (DF) is a rare, locally aggressive, nonmetastasizing fibroblastic/myofibroblastic tumor with a tendency to recur and an unpredictable clinical course. A "wait-and-see" policy is the new standard of care. DF are characterized by activating alterations of the wnt/ß-catenin pathway: CTNNB1 or adenomatous polyposis coli gene (APC) mutations (these mutations being mutually exclusive). Desmoid-type fibromatosis of the breast (DFB) is rare with an incidence of 0.2% of breast tumors. The diagnosis of DFB is difficult, as it must be distinguished from metaplastic carcinoma and other spindle cell lesions. Sequencing of 128 DFB identified a lower rate of CTNNB1 mutations using Sanger (65.6%) or Sanger+next-generation sequencing (77.7%) and a higher rate of APC mutations (11.8%) than in all-site DF. By excluding patients with familial adenomatous polyposis (n=2), the rate of APC mutations in DFB was high (10.7%). The distribution of CTNNB1 mutations in DFB was different from all-site DF, with a higher rate of T41A (68.9%), a lower rate of S45F (5.7%), and a similar rate of S45T (12.6%). By combining the 2 molecular techniques in a 2-step manner (Sanger, then next-generation sequencing), we increased the detection rate of CTNNB1 mutations and lowered the rate of wild-type tumors from 34.4% to 9.8%, therefore improving the diagnosis of DFB. The identification of the exon 3 CTNNB1 mutation in breast spindle cell lesions is a highly specific tool for the diagnosis of DFB, in addition to extensive immunohistochemical analysis. Our study also underlines the importance of APC in DFB tumorigenesis. These findings have significant implications for patient care and management.

Management of desmoid tumours: A nationwide survey of labelled reference centre networks in France.

PURPOSE: The optimal management of rare tumours (i.e. from accurate diagnosis to management in reference centres) is a public health challenge. In 2009, the French National Cancer Institute (INCa) identified and financially supported the two expert networks for pathological and clinical diagnosis and management of soft tissue tumours. METHODS: The activities of both networks were prospectively collected using a nationwide database (rreps.org). Data describing the diagnosis management of 863 successive cases of desmoids tumours (DT) were prospectively collected from 2010 to 2013 and analysed. RESULTS: The number of confirmed DT constantly improved from January 2010 to December 2013 (from 173 to 273 cases per year); the expected incidence ranged from 132 to 330 cases/year. The rate of cases diagnosed with core-needle biopsies and CTNNB1 mutational status analysis increased from 30.6 to 40.7% and from 87.8 to 94.1%, respectively. The mean delay for pathological diagnosis confirmation constantly decreased from 107 to 47 d. Among the 846 adult patients, 414 (48.9%) patients were treated by reference centres. The rate of patients managed by reference centres constantly increased with time from 36.9 to 49.5% since 2010. The median management time of the referral centres constantly decreased from 440 to 67 d. CONCLUSION: The two expert networks worked synergistically and improved diagnosis modalities of rare desmoid tumours at a national level. The impact of management by expert networks on the outcome will be prospectively analysed in the future.