Ovarian cancer is the leading cause of death from gynecologic cancer in the US. For most patients, the disease is first diagnosed at an advanced stage, and the 5-year survival rate is low (< 30%). Despite incremental improvement in chemotherapy, the cure rate has not improved significantly in the past decades, The dramatic difference in long-term survival between patients with local disease (80-90%) and those with distant metastases (5-20%) suggests the need for a non-invasive, yet effective test applicable to at-risk population groups to detect ovarian cancer in early stages. Building upon prior research in differentiating malignant from benign ovarian masses, this project seeks to apply artificial neural network (ANN) technology to the problem of screening for early-stage ovarian cancer based on a variety of serum markers and other clinical inputs. The recently completed Phase I pilot project validated the feasibility by (1) assembling existing data from collaborating Organizations, (2) analyzing the predictive value of relevant biomarkers, (3) developing a preliminary ANN, and (4) validating the ANN using independent test data. The successful completion of Phase I now facilitates the commencement of Phase II activities to develop a production version of the screening system and to initiate broad-scale validation through multiple clinical studies.
ANN software capable of detecting early-stage ovarian cancer with sufficient improvement in specificity, sensitivity, and predictive value over alternative techniques would have clear commercial value in screening high-risk populations. Horns presently offers as a commercial product an Internet-based clinical information processing service, called ProstAsure, developed using ANN technology, for the detection of prostate cancer.