JMIR Publications

ABSTRACT

Background:

Colorectal and prostate cancer are the most dominant kind of cancer in men recorded all over the world. To diagnose colorectal and prostate, a pathologist performs a histological analysis on needle biopsy samples. This manual process is time-consuming and error-prone. This results in a high intra- and inter-observer variability, which affects the reliability of the diagnosis.

Objective:

The primary objective of our study is to develop a computerized automatic system for diagnosis of colorectal and prostate tumors using images of biopsy samples in order to reduce the human analysis time and the diagnosis error rates.

Methods:

In this work, we propose a CNN model (based on VGG16 [3] ) for classification of colorectal and prostate tumors from multispectral images of biopsy samples. The key idea is based on removing the last block of convolutional layers and halved the number of lters per layer.

Results:

The results show excellent performances achieving an average test accuracy of 99.8 % and 99.5% for the prostate and colorectal datasets, respectively. It shows excellent performance compared to pre-trained CNNs and to other classification methods as it avoids the preprocessing phase while using a single CNN model for the whole classification task. Over-all, the proposed CNN architecture was globally the best performing system for classifying colorectal and prostate tumor images.

Conclusions:

A proposed CNN architecture was detailed and compared to previously trained network models used as feature extractors. These CNNs were also compared to other classification techniques. As opposed to pre-trained CNNs and other classification approaches, the proposed CNN gave excellent results. The computational complexity of the CNNs was also investigated, and it was shown that the proposed CNN is better at classifying images than pre-trained networks because it does not require preprocessing. Thus, the overall analysis was that the proposed CNN architecture was globally the best performing system for classifying colorectal and prostate tumour images.