Pengukuran ketinggian pesakit diperlukan bagi menentukan keperluan tenaga asas, langkah-langkah penyeragaman kapasiti fizikal dan untuk menyesuaikan dos dadah. Ramalan ketinggian pesakit warga tua dengan mengukur panjang span lengan mereka bukanlah suatu perkara yang baru dalam bidang perubatan. Melalui konsep Rangkaian Neural Buatan (ANN), latihan rangkaian dengan data input yang diketahui boleh menjana data output yang diingini. Di sini, data input terdiri daripada panjang span lengan dan umur pesakit manakala ketinggian diberikan sebagai data output. Kedua-dua jantina iaitu lelaki dan perempuan dijadikan faktor, memberi 2 set eksperimen. Ketinggian tetap dan panjang span lengan 205 pesakit lelaki dan 126 pesakit wanita yang berumur di antara 60 dan 80 telah diukur. MATLAB Neural Network Toolbox digunakan sebagai perisian antara muka (API) untuk ramalan ketinggian dalam eksperimen. Teori Rangkaian Neural Buatan dikaji sebelum nilai untuk bias berat awal dan parameter rangkaian diberi. Selepas itu, data untuk panjang span lengan dan umur adalah input kepada rangkaian untuk menjalani pembelajaran diselia. Rangkaian dilatih sehingga zaman maksimum dicapai atau berhenti pengesahan, menjana data output. Data output yang dijana, juga dikenali sebagai output sebenar berbanding dengan output yang diingini dari segi Ketepatan, Kadar Ralat dan Regresi. Eksperimen kedua-dua jantina dilakukan dengan meningkatkan bilangan neuron tersembunyi dalam rangkaian, mengakibatkan 8 set rangkaian masing-masing. Daripada itu, analisis prestasi untuk rangkaian dilakukan. Hasil perbandingan menunjukkan bahawa terdapat nombor optimum neuron tersembunyi untuk menjana ketepatan maxima. Jantina bukan suatu faktor penting dalam ramalan ketinggian menggunakan panjang span lengan.
_______________________________________________________________________________________________________
Measurement of the height of patients is required for determination of basic energy requirements, standardization of measures of physical capacity and for adjusting drug dosage. Prediction of elderly patient’s height by measuring their arm span length is not a new thing in medical field. Through the concept of Artificial Neural Network (ANN), training a network with known input of data can generate desired output data. Here, the input data consists of arm span length and age of patients while height is given as output data. Both gender of male and female is considered, resulting in 2 sets of experiment. Standing height and arm span lengths of 205 male patients and 126 female patients between the ages of 60 and 80 were measured. MATLAB Neural Network Toolbox is used as the application programming interface (API) for height prediction in these experiments. The theory of Artificial Neural Network is studied before the initial weight bias and parameters of network are conducted. After that, given data of arm span length and ages are input to the network to undergo supervised learning. Network is trained until maximum epoch is reached or validation stops, generating result data. The generated result data, also known as actual output is compared with the desired output in terms of Accuracy, Error Rate and Regression. Experiments of both gender is done by increasing the number of hidden neurons in network, resulting in 8 set of network respectively. From that, analysis of performance for network is done. The comparison result shows that there will be an optimum number of hidden neurons to generate highest accuracy result. Gender is not an important factor in height prediction using arm span length.
Prediction of standing height for hospitalized elderly using artificial neural network / Kelvin Ng Kai Ming