AI inspired discovery of new biomarkers for clinical prognosis of liver cancer

Abstract

Tissue biomarkers are crucial for cancer diagnosis, prognosis assessment, and treatment planning. However, few of current biomarkers used in clinics are robust enough to show a true analytical and clinical value. Thus the search for additional tissue biomarkers, including the strategies to identify them, is imperative. Recently, the capabilities of deep learning (DL)-based computational pathology in cancer diagnosis and prognosis have been explored, but the limited interpretability and generalizability make the results difficult to be accepted in clinical practice. Here we present an interpretable human-centric DL-guided framework—PathFinder (Pathological-biomarker-finder)— that can inspire pathologists to discover new tissue biomarkers from well-performing DL models, which bridges the gap between DL and clinical prognosis. By combining sparse multi-class tissue spatial distribution information of whole slide images (WSIs) with attribution methods, PathFinder can achieve localization, characterization, and verification of potential biomarkers, while guaranteeing state-of-the-art prognostic performance. With the inspiration of PathFinder, we discovered that tumor necrosis in liver cancer, a long-neglected factor, has a strong relationship with patient prognosis. Thus we proposed two clinically independent indicators, including necrosis area fraction and tumor necrosis distribution, for practical prognosis, and verified their potentials in clinical prognosis according to Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK)-derived criteria. Our work demonstrates a successful example of introducing artificial intelligence (AI) into clinical practice in a knowledge discovery way, which can be adopted in identifying biomarkers in various cancer types and modalities.

Deep learning‐based accurate diagnosis and quantitative evaluation of microvascular invasion in hepatocellular carcinoma on whole‐slide histopathology images

“…The development of artificial intelligence (AI) in medicine has led to significant advancements in the diagnosis of tumors, assessment of drug efficacy, and prognosis prediction. [11][12][13][14][15] In HCC, deep-learning models based on histological whole-slide images (WSIs) have been widely applied in diagnosis, 16 pathological grading, 17 molecular characterization 16,17 and prognostic assessment. 15,18 However, AI studies of MVI in HCC have mainly focused on preoperative prediction using radiomics, 19,20 which results in underestimation of the detection rate of MVI.…”

“…[11][12][13][14][15] In HCC, deep-learning models based on histological whole-slide images (WSIs) have been widely applied in diagnosis, 16 pathological grading, 17 molecular characterization 16,17 and prognostic assessment. 15,18 However, AI studies of MVI in HCC have mainly focused on preoperative prediction using radiomics, 19,20 which results in underestimation of the detection rate of MVI. 21 As of now, there is no deep learning model that can rapidly and accurately assess postoperative histological MVI of HCC.…”

“…The development of artificial intelligence (AI) in medicine has led to significant advancements in the diagnosis of tumors, assessment of drug efficacy, and prognosis prediction 11–15 . In HCC, deep‐learning models based on histological whole‐slide images (WSIs) have been widely applied in diagnosis, 16 pathological grading, 17 molecular characterization 16,17 and prognostic assessment 15,18 .…”

PathEX: Make Good Choice for Whole Slide Image Extraction

“…Moreover, within the annotated regions, there are still some blank background area or area with lots of fat or others shown in Figure .1.c. If we donot treat three types of noise tile images carefully, we may result in inclusion of some irrelevant background information, noise tissue into DL training dataset, affecting the performance of neural network [21].…”

Multitask deep learning for prediction of microvascular invasion and recurrence‐free survival in hepatocellular carcinoma based on MRI images

“…Therefore, our study aimed to explore a non-invasive and user-friendly tool capable of reflecting the biological behaviour associated with HCC prognosis. Several recent studies have demonstrated the prognostic value of deep learning models in HCC, yielding promising results [30][31][32]. In our study, we suc-cessfully established a deep learning model for predicting RFS using multi-set data, achieving C-index values of 0.763, 0.716, 0.628, 0.675 and 0.728 in the training set, internal test set, external test set 1, external test set 2 and external test set 3, respectively.…”