Customer Intelligence Suite

Jul 2021 | Research Papers

Introduction 

Current  diagnostic technologies are often presented in 2D radiological or microscopic images.  These views are often susceptible to  multiple  errors  ranging  from  improper  staining  to  blurred images  resulting  from   patient   movement   during   radiological test.   However  with  the  use of image analysis and machine  intelligence, medical experts can be assisted in diagnosis and planning.  The  aim  is  to  develop  a  software  that  utilizes  deep  learning to automatically  detect patterns  from  medical images with  high accuracy. This will assist radiologists in concordance with disease diagnosis of disease using x-ray scans.

Lung  cancer  diagnosis diagnosis prognosis is primarily  dependent  on  the  size and stage of the  growth  upon  detection.    Early  diagnosis increases the patient’s chance of survival to up to 70% following surgical removal of the tumor.1    Studies have confirmed the advantage of low dose spiral CT over CXR in terms of early detection and accuracy.2, 3

Current   diagnostic  breast  imaging  technologies  are  often  presented in 2D.4 This limited view can result to risks such as tumor dosage overestimation.5      With  the  use of 3D  image rendering, segmentation and classification, medical experts can be assisted in diagnosis and planning.6

The study aims to develop a new and groundbreaking imaging software that will assist medical practitioners as concordance with lung, breast, and thyroid  cancer diagnosis using radiological and microscopic images.  The software will provide a low-cost, highly accurate, and readily available cancer detection method.  Historical CT, MRI, ultrasound, and microscopic images are needed as the training set for the software’s self-learning algorithm. There is no current risk to the health and privacy of the patient population as the information gathered will have no Personally Identifiable Information.

Radiological and microscopic images, age, gender, and diagnosis of patients who have undergone neck x-ray scan will be extracted for analysis. Infants and children under 18 are excluded as subjects in the study population.  Subjects will be anonymized and will be represented by unique ID key to preserve the personal identity of the patient.

Related Literature

Developing an automated cancer detection system is very challenging because lesion areas are only defined through intensity changes relative to surrounding tissues as seen on a radiological image.  Some of the factors that has to be taken into account in the system are as follows:

  • different imaging systems with varied image resolutions,
  • different settings for each patient (varied intensity values for each imaging),
  • obfuscation of lesion by partial volumes or artifacts,
  • varied tumor structures (size, extension, localization), and
  • mass effect: growing tumor displace normal tissues which limits reliability of spatial prior knowledge.

Due to the presence of the characteristic speckle noise from these images, image preprocessing, such as filtering, is necessary for accurate automatic detection of breast cancer. The information of the contrast and the texture of the different regions allow a precise isolation of the lesions. Finally, the automatic analysis of the diagnostic criteria by means of different shape and histogram analysis techniques supplies a deep characterization of the cancer nodules.

Medical specialists can identify lesions through experience by examining hundreds of scans.  However with these factors affecting the quality of the scans being examined, isolating lesions through visual inspection may not be enough.  In order to improve the images read by specialists, the objective is to develop a CAD system specifically designed in detecting and isolating possible lesions from radiological images.

Lung  Cancer

Lung cancer is one of the top cancers prevalent in the Philippines and causes of worldwide cancer mortality.  According to the World Health Organization,  there are approximately 1.59 million cases out of the 8.2 million cancer-related deaths reported  in 2012.  Moreover, it is the most common  cause of cancer-related mortality among men and the third most common mortality and morbidity  causes in the Philippines.7   Lung cancer prognosis is primarily dependent on the size and stage of the growth  upon detection.2    Early diagnosis increases the patient’s chance of survival to up to 70% following surgical removal of the tumor.1   Studies have confirmed the advantage of low dose spiral computed tomography (CT) over CXR in terms of detection accuracy.2, 3

Current  methods for lung cancer diagnosis have three initial steps prior to prognosis and treatment, namely: tissue diagnosis, staging, and functional evaluation.  Tissue diagnosis may be done via thoracotomy or bronchoscopy  following detection of the tumor.  For patients with suspected lung cancer but unclear detection results, noninvasive techniques such as sputum cytology and transthoracic needle aspiration are recommended.8

Apart from the high cost and inaccessibility of these tests ( particularly  for the lower classes), the  inaccuracy of lung cancer detection may cause over-diagnoses, unnecessary radiation exposure, and anxiety among patients.Hence, the need for a low-cost, highly accurate, and readily available lung cancer detection method is apparent.

Current  lung imaging techniques for early cancer detection include X-ray (Chest Radiograph), CT (Computed Tomography), PET ( Positron Emission Tomography), and MRI ( Magnetic Resonance Imaging) scans. CT scans are often the preferred input for feature extraction algorithms due to their high image quality and absence of distortion.31

ANNs have been used widely in medical image processing in the past decade.21 ANNs are used mostly in medical image preprocessing, segmentation, and object detection,.21   These techniques allow them to be used specifically in lung cancer detection.  Figure 1 describes the architecture of the ANN  developed by Dandil et al for identifying lung nodules and classifying them as benign or malignant.32  The hidden layer of the ANN  contains iterative training and prediction processes that enable nodule classification after learning from several examples.

Breast  Cancer

Breast cancer is the most diagnosed cancer in women and the second leading cause of cancer death. The chances of death due to breast cancer is 1 in 37. Death rates have decreased worldwide as a result of earlier screening and diagnosis. Diagnostic imaging techniques such as mammography  detect breast cancer before they become invasive. Cancer death, however, is still being observed and additional effort must be done for its further decline in the future.

 

The clinical procedure for breast cancer detection starts with the use of imaging technology, as mentioned, tools like mammography, CT and ultrasound (US), to capture the exact location of the nodules. From the image of the mass, the physician will then give the initial findings on the nature of the mass, whether Benign or Malignant, this judgement is based on the shape of the nodules and from past experience of the physician. However, patients are still often recommended to undergo needle biopsy. This procedure uses a needle device that will be used to take sample tissue of the mass for evaluation. While this guarantees 100% correct diagnosis based on the sample, the involvement of the needle could have been avoided if one can predict the benign tumor with high accuracy based on the image; and malignant tumors, given greater accuracy would have been diagnosed at surgery. Further, the limited view of the imaging technology presented in 2D9 can result to risks such as tumor dosage overestimation. This limitation initiated the application of deep learning to medical imaging diagnosis, specifically the use of 3D image rendering, segmentation, and classification, medical experts can be assisted in diagnosis and planning.6

The earliest sign of breast cancer is the presence of microcalcification and masses. These abnormal tissues can best be detected using modern techniques such as mammography,  MRI and ultrasound,  and further diagnosed through image-guided needle biopsy.37  Generally, detection of masses in breast tissues is more challenging compared to the detection of microcalcifications, not only due to the large variation in size and shape but also because masses often exhibit poor image contrast when using mammography.  The difficulty in classification of benign and malignant microcalcifications also causes a significant problem in medical image processing.20