Deep learning algorithms have been successfully used in medical image classification. In the next stage, the technology of acquiring explainable knowledge from medical images is highly desired. Here we show that deep learning algorithm enables automated acquisition of explainable features from diagnostic annotation-free histopathology images. We compare the prediction accuracy of prostate cancer recurrence using our algorithm-generated features with that of diagnosis by expert pathologists using established criteria on 13,188 whole-mount pathology images consisting of over 86 billion image patches. Our method not only reveals findings established by humans but also features that have not been recognized, showing higher accuracy than human in prognostic prediction. Combining both our algorithm-generated features and human-established criteria predicts the recurrence more accurately than using either method alone. We confirm robustness of our method using external validation datasets including 2276 pathology images. This study opens up fields of machine learning analysis for discovering uncharted knowledge.

We experienced marked efficacy with steroid treatment of three patients with jackhammer esophagus (JHE). An esophageal biopsy revealed eosinophilic esophagitis (EoE) in two patients. One of the patients without EoE had eosinophilia and an increased serum immunoglobulin E level, and endoscopic ultrasonography revealed thickening of the esophageal muscularis propria. Esophageal manometry was used to diagnose all cases of JHE. Treatment consisted of steroid administration, which improved the symptoms and resolved the esophageal muscularis propria thickening in all patients. The esophageal manometry findings also normalized following treatment. Allergic diseases, including EoE, were assumed to have caused JHE.

No abstract

1Deep learning algorithms have been successfully used in medical image classification and 2 cancer detection. In the next stage, the technology of acquiring explainable knowledge from 3 medical images is highly desired. Herein, fully automated acquisition of explainable features 4 from annotation-free histopathological images is achieved via revealing statistical distortions 5 in datasets by introducing the way of pathologists' examination into a set of deep neural 6 networks. As validation, we compared the prediction accuracy of prostate cancer recurrence 7 using our algorithm-generated features with that of diagnosis by an expert pathologist using 8 established criteria on 13,188 whole-mount pathology images. Our method found not only the 9 findings established by humans but also features that have not been recognized so far, and 10 showed higher accuracy than human in prognostic prediction. This study provides a new field 11 to the deep learning approach as a novel tool for discovering uncharted knowledge, leading to 12 effective treatments and drug discovery. 13 (149/ 150 word) 14 15

Segmental absence of the intestinal musculature (SAIM) can cause intestinal perforation in adults. However, its prevalence and clinicopathologic features have not been well-described. This study aimed to determine the prevalence of SAIM-associated perforation and characterize its clinicopathologic features. We retrospectively examined 109 cases of intestinal perforation that underwent surgical resection from January 2009 to December 2019. SAIM was defined as the complete absence of the muscularis propria without extensive inflammation and fibrinous exudation around the perforation. SAIM was the second most frequent cause of perforation (26 cases: 24%), the most frequent cause being related to diverticulitis (39 cases: 36%). The most common site was the sigmoid colon (12 cases: 46.2%). The younger group (aged below 65 y) exhibited more frequent perforation of the upper segments of the gastrointestinal tract (from the duodenum to the descending colon) than the older group (65 y and above) (P=0.0018). No patients developed recurrence. The most common gross features were well-defined circular or small punched-out lesions, and the histologic features were complete absence of the muscularis propria and absence of hemorrhage and necrosis around the area of perforation. The characteristic features of SAIM were unique and their prevalence was higher than previously reported. The precise recognition of SAIM can aid in understanding the cause of perforation and avoiding further unnecessary examinations.