Unsupervised Approach for Malignancy Assessment of Lung Nodules in Computed Tomography Scans Using Radiomic Features.

Lung cancer is the leading cause of cancer death worldwide. Early low-dose computed tomography (CT) screening can decrease its mortality rate and computer-aided diagnoses systems may make these screenings more accessible. Radiomic features and supervised machine learning have traditionally been employed in these systems. Contrary to supervised methods, unsupervised learning techniques do not require large amounts of annotated data which are labor-intensive to gather and long training times. Therefore, recent approaches have used unsupervised methods, such as clustering, to improve the performance of supervised models. However, an analysis of purely unsupervised methods for malignancy prediction of lung nodules from CT images has not been performed. This work studies nodule malignancy in the LIDC-IDRI image collection of chest CT scans using established radiomic features and unsupervised learning methods based on k-Means, Spectral Clustering, and Gaussian Mixture clustering. All tested methods resulted in clusters of high homogeneity malignancy. Results suggest convex feature distributions and well-separated feature subspaces associated with different diagnoses. Furthermore, diagnosis uncertainty may be explained by common characteristics captured by radiomic features. The k-Means and Gaussian Mixture models are able to generalize to unseen data, achieving a balanced accuracy of 87.23% and 86.96% when inference was tested. These results motivate the usage of unsupervised approaches for malignancy prediction of lung nodules, such as cluster-then-label models. Clinical Relevance- Unsupervised clustering of radiomic features of lung nodules in chest CT scans can differentiate between malignant and benign cases and reflects experts' diagnosis uncertainty.

The breast cancer conservative treatment. Cosmetic results - BCCT.core - Software for objective assessment of esthetic outcome in breast cancer conservative treatment: A narrative review.

BACKGROUND AND OBJECTIVE: Cosmetic outcome of breast cancer conservative treatment (BCCT) remains without a standard evaluation method. Subjective methods, in spite of their low reproducibility, continue to be the most frequently used. Objective methods, although more reproducible, seem unable to translate all the subtleties involved in cosmetic outcome. The breast cancer conservative treatment cosmetic results (BCCT.core) software was developed in 2007 to try to overcome these pitfalls. The software is a semi-automatic objective tool that evaluates asymmetry, color differences and scar visibility using patient's digital pictures. The purpose of this work is to review the use of the BCCT.core software since its availability in 2007 and to put forward future developments. METHODS: All the online requests for BCCT.core use were registered from June 2007 to December 2014. For each request the department, city and country as well as user intention (clinical use/research or both) were questioned. A literature search was performed in Medline, Google Scholar and ISI Web of Knowledge for all publications using and citing "BCCT.core". RESULTS: During this period 102 centers have requested the software essentially for clinical use. The BCCT.core software was used in 19 full published papers and in 29 conference abstracts. CONCLUSIONS: The BCCT.core is a user friendly semi-automatic method for the objective evaluation of BCCT. The number of online requests and publications have been steadily increasing turning this computer program into the most frequently used tool for the objective cosmetic evaluation of BCCT.