Artificial Intelligence-based Medical Image Automatic Diagnosis and Prognosis Prediction

Purpose: We aim to develop and validate PET/ CT image-based radiomics to determine the Ki-67 status of high-grade serous ovarian cancer (HGSOC), in which we use the metabolic subregion evolution to improve the prediction ability of the model. At the same time, the stratified effect of the radiomics model on the progression-free survival rate of ovarian cancer patients was illustrated.

Materials and methods: We retrospectively reviewed 161 patients with HGSOC from April 2013 to January 2019. 18F-FDG PET/ CT images before treatment, pathological reports, and follow-up data were analyzed. A randomized grouping method was used to divide ovarian cancer patients into a training group and validation group. PET/ CT images were fused to extract radiomics features of the whole tumor region and radiomics features based on the Habitat method. The feature is dimensionality reduced, and meaningful features are screened to form a signature for predicting the Ki-67 status of ovarian cancer. Meanwhile, survival analysis was conducted to explore the hierarchical guidance significance of radiomics in the prognosis of patients with ovarian cancer.

Results: Compared with texture features extracted from the whole tumor, the texture features generated by the Habitat method can better predict the Ki-67 state (p < 0.001). Radiomics based on Habitat can predict the Ki-67 expression accurately and has the potential to become a new marker instead of Ki-67. At the same time, the Habitat model can better stratify the prognosis (p < 0.05).

Conclusion: We found a noninvasive imaging predictor that could guide the stratification of prognosis in ovarian cancer patients, which is related to the expression of Ki-67 in tumor tissues. This method is of great significance for the diagnosis and treatment of ovarian cancer.

Object detection technology has been widely used in medical field, such as detecting the images of blood cell to count the changes and distribution for assisting the diagnosis of diseases. However, detecting small objects is one of the most challenging and important problems especially in medical scenarios. Most of the objects in medical images are very small but influential. Improving the detection performance of small objects is a very meaningful topic for medical detection. Current researches mainly focus on the extraction of small object features and data augmentation for small object samples, all of these researches focus on extracting the feature space of small objects better. However, in the training process of a detection model, objects of different sizes are mixed together, which may interfere with each other and affect the performance of small object detection. In this paper, we propose a method called pixel level balancing (PLB), which takes into account the number of pixels contained in the detection box as an impact factor to characterize the size of the inspected objects, and uses this as an impact factor. The training loss of each object of different size is adjusted by a weight dynamically, so as to improve the accuracy of small object detection. Finally, through experiments, we demonstrate that the size of objects in object detection interfere with each other. So that we can improve the accuracy of small object detection through PLB operation. This method can perform well with blood cell detection in our experiments.