經(jīng)過(guò)二輪審稿，第一輪大修三個(gè)人提出了不少意見。但是這個(gè)審稿人2是個(gè)奇葩，英語(yǔ)表示存在歧義，說(shuō)的不明不白的，到了第二輪才說(shuō)，這么點(diǎn)數(shù)據(jù)用深度學(xué)習(xí)效果不行，關(guān)鍵我也不是這么點(diǎn)數(shù)據(jù)，我們這個(gè)方向都是進(jìn)行樣本裁剪進(jìn)行增強(qiáng)后都有10w+的數(shù)據(jù)，他也不看我文中的。這是他兩次的審稿意見，很奇葩。
第一次意見：
in majority scenarios of medical image classification, acquiring sufficient amount of labeled images is very difficult. as mentioned in the manuscript (line 164, page 6), there are 468 images in total for this study. with this size of data, it could be questionable to justify the effectiveness of deep networks’ capability for feature extraction.

第二次意見：都是對(duì)應(yīng)點(diǎn)，就這個(gè)問題他不滿意
for example, in my first comment, i argued why to use the deep networks in this medical image classification, since the training samples are so little. in the authors’ response, i did not find any justification or arguments that can explain away this question. however, the deep networks are the main techniques used in the manuscript, and this should not be ignored.

overall, the manuscript is hard to follow to me. the manuscript’s motivation of using feature fusion also seems to be questionable.

therefore, i regret that i am not able to recommend this manuscript to be considered for publication in sensors. 返回小木蟲查看更多