Page 85 - sjsi

Research Article: Alsibai & Heydari 85

Research Article

The Effectiveness of CNN in Evaluating Ultrasound Image Datasets:
SJSI Diagnosing Polycystic Ovary Syndrome (PCOS) as an Example

Abstract
Authors: Deep learning has proved its potential and the vital role it can provide in
Mohammed Hayyan benefitting and assisting practitioners that use ultrasonography as a tool
Alsibai for diagnosis. However, medical datasets are often very difficult to
Seyed Youssef Heydari obtain. Moreover, erroneous or fake training datasets may lead to
inaccurate results. This paper is discussing the robustness of deep
learning in diagnosing polycystic ovary syndrome (PCOS). Two public
Affiliation: online data sets were used to train and test the ability of transfer learning
Department of using DenseNet201 architecture in detecting the infected ovaries. The
Computer and two datasets showed different results although the same model,
Informatics Engineering, architecture and parameters were used. Further investigation showed
International University that one of the data sets is extremely erroneous and of misleading
of Science and
information. In conclusion, Convolutional Neural Networks (CNN) and
Technology (IUST), Transfer Learning can be considered as a strong tool to evaluate datasets.
Daraa _ Syria
Keywords: Convolutional Neural Networks (CNNs); Transfer Learning;
Deep Learning; Polycystic ovary syndrome (PCOS); Medical Ultrasound
Submitted 14 Feb 2023
Accepted 10 Apr 2023 Image Processing; Classification; Misleading Datasets.
Correspondence:
mhdhayyan@gmail.com

P least two of the three following features were
Introduction
olycystic ovary syndrome (PCOS) is one
present at an expert meeting in Rotterdam in
of the most prevalent conditions
2003: (i) oligo- or anovulation, (ii) clinical and/or
affecting women of reproductive age. It
(5)
polycystic ovaries . The latter of which can be
affects 6%–20% of premenopausal women biochemical hyperandrogenism, or (iii)
(1)
globally . Ovarian dysfunction and an excess detected using ultrasonography which offers
of androgen are the two main symptoms of the highest contribution to the diagnosis of
(6)
PCOS. Menstrual abnormalities, hirsutism, PCOS . Early detection of PCOS is important
obesity, insulin resistance, cardiovascular because it can help to manage the symptoms
disease, in addition to emotional symptoms like and reduce the risk of long-term health issues.
depression (2) are all common among PCOS Although ultrasound images have some
patients (3) . As a result, it is crucial for the disadvantages of strong artifacts, noise and high
accurate diagnosis and treatment of PCOS. In dependence on the experience of doctors, they
(4)
1985, Adams et al. discovered that polycystic are still considered as one of the most widely
ovaries have an abnormally high number of used modalities in medical diagnosis. Many
follicles, also termed multifollicularity. It was artificial intelligence systems have been
suggested that PCOS be diagnosed when at developed to help doctors. Convolutional

SJSI – 2023: VOLUME 1-1