The realization of targeted cancer therapy has driven the need to improve selection of patients with colorectal cancer for adjuvant therapy, leading to a search for potential new prognostic markers. There is accumulating evidence that immunosurveillance acts as an extrinsic tumor suppressor. As genetic instability is an early event in colorectal cancer, this can lead to altered expression of molecules conferring resistance to immune attack. Hence, molecules up or downregulated in this process may impact on patient survival. In our study, 449 colorectal tumors were screened for expression of the stress‐related protein MICA, which functions as a ligand for the NKG2D receptor and whose expression confers susceptibility to both T‐ and NK‐cell attack. Intensity of MICA expression was quantified using automated image analysis and MICA expression showed no correlation with conventional clinicopathological variables. In contrast, survival analysis showed a significant correlation between higher levels of MICA expression and improved disease‐specific survival, with independent prognostic significance in multivariate analysis. Thus, patients with low levels of MICA and a poor prognosis may be good candidates for aggressive chemotherapy. In contrast, patients with high expression of MICA may be candidates for the antibody therapies, as they should be susceptible to NK killing by antibody dependent cellular cytotoxicity.

Supervised machine learning techniques are increasingly being combined with ultrasonic sensor measurements owing to their strong performance. These techniques also offer advantages over calibration procedures of more complex fitting, improved generalisation, reduced development time, ability for continuous retraining, and the correlation of sensor data to important process information. However, their implementation requires expertise to extract and select appropriate features from the sensor measurements as model inputs, select the type of machine learning algorithm to use, and find a suitable set of model hyperparameters. The aim of this article is to facilitate implementation of machine learning techniques in combination with ultrasonic measurements for in-line and on-line monitoring of industrial processes and other similar applications. The article first reviews the use of ultrasonic sensors for monitoring processes, before reviewing the combination of ultrasonic measurements and machine learning. We include literature from other sectors such as structural health monitoring. This review covers feature extraction, feature selection, algorithm choice, hyperparameter selection, data augmentation, domain adaptation, semi-supervised learning and machine learning interpretability. Finally, recommendations for applying machine learning to the reviewed processes are made.

Introduction: There is sufficient evidence that blood group related Lewis antigens are tumour-associated molecules. The Lewis^y and Lewis^b antigens are complex carbohydrates that are over-expressed by breast, lung, colon and ovarian cancers. The SC101 mAb is a unique Lewis^y/b binding antibody that binds to native and extended Lewis^y and Lewis^b haptens, displaying no cross reactivity with H type 1, H type 2, Lewis^x or normal blood group antigens. Methods: Immunohistochemical detection of Lewis^y/b was performed on 660 formalin-fixed, paraffin embedded breast tumour specimens using a streptavidin-biotin peroxidase technique. Tissue from these patients had previously been included in tissue microarrays. This cohort comprises a well characterized series of patients with primary operable breast cancer diagnosed between 1987 and 1992, obtained from the Nottingham Tenovus Primary Breast Carcinoma Series. This includes patients 70 years of age or less, with a mean follow up of 7 years. Results: Of the breast carcinomas, 370 of 660 (56%) were negative for Lewis^y/b expression, 110 (17%) cases showed a low level of expression (75% of positive cells). We found significant positive associations between histological grade (p < 0.001), Nottingham Prognostic Index (p = 0.016), tumour type (p = 0.007) and the level of Lewis ^y/b expression. There was a significant correlation between the proportion of Lewis^y/b positive tumour cells and survival in lymph-node negative patients (p = 0.006). Conclusion: The unique epitope recognised by SC101 mAb on Lewis^y/b hapten is over-expressed on breast tumour tissue compared with normal breast. In this large series of invasive breast cancers, higher expression of Lewis^y/b was more often found in high grade and poor prognosis tumours compared to good prognosis cancers. Moreover, in lymph node negative breast carcinomas, over-expression of Lewis^y/b hapten was associated with significantly decreased patient survival.