Related Experiment Video
Updated: Aug 5, 2026

Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
[Metastatic cáncer presentation. Validation of a diagnostic algorithm with 221 consecutive patients]
F Losa Gaspà1, J R Germá, J M Albareda
1Servicio de Oncología Médica, Institut Català d'Oncologia (ICO), Hospital Duràn i Reynals, L'Hospitalet, Barcelona, Spain. Ferran.Losa@chcr.scs.es
Background:
The use of a diagnostic algorithm for metastatic cancer presentation (MCP) might enhance the diagnosis of primary tumors amenable to treatment with considerable savings both in time and diagnostic examinations.
Materials And Methods:
From January 1992 to April 1997, all patients admitted with the diagnosis of MCP were prospectively studied. From each patient, a basic study consisting in a clinical interview, complete physical examination, standard blood testing with tumoral markers and chest X-ray were obtained. Patients with a negative basic study were classified as having a metastatic cancer of unknown origin (MUO); in these patients, a protocolized study (abdominal CT scan and mammography among women) were performed. Patients who after the application of the basic and protocolized studies had no primary tumor detected underwent an exhaustive investigation in order to validate the efficiency of the diagnostic algorithm.
Results:
Two hundred twenty-one patients were included in the study. The mean age of patients was 63 years (range: 23-82). The main symptom was of bone (30%), neurological (24%), thoracic (16%) and abdominal (16%) origin. The basic study was positive for 138 patients (62.4%), with chest X-ray and physical examination yielding the highest number of diagnoses among these patients. The histology of metastases contributed to the definite diagnosis in 31 patients. Only PSA had a high sensitivity and specificity. Eighty-three patients were classified as MUO. The protocolized study diagnosed the primary tumor in 24 patients (30%), 20 by abdominal CT scan and four by mammography; eight of these patients were deemed to be amenable to treatment. The remaining 59 patients underwent an exhaustive study, and a diagnosis was made in 13; nevertheless, none of them was considered candidate for a specific treatment. Finally, 47 patients (21%) remained undiagnosed. The predominant primary tumors included sites at the lung (42%), prostate (6%) and breast (6%). The most common metastatic locations included bone (42%), central nervous system and liver (24%), and the most common histological types were adenocarcinoma (61%) and undifferentiated carcinoma (15%).
Conclusions:
Lung cancer and MUO represented 62% of MCP. The basic study oriented in two thirds of cases, and the physical examination and chest X-ray showed the highest diagnostic yield. The histology of metastases and PSA had a key, diagnostic relevance. A protocolized study based on abdominal CT scan and mammography (females) can identify the remaining treatable tumors.
More Related Videos
08:54Pathological Analysis of Lung Metastasis Following Lateral Tail-Vein Injection of Tumor Cells
Published on: May 20, 2020
07:13Comparison of Predictive Performance of Three Lymph Node Staging Systems in Colorectal Signet Ring Cell Carcinoma Based on Machine Learning Model
Published on: April 18, 2025